Sampling and Random Assignment

The sampling frame for SEED OK consisted of the birth records for all children born in Oklahoma in April through June and in August through October of 2007. (The second sampling period was necessary because the response rate in the first period was lower than expected.) Researchers oversampled African American, American Indian, and Hispanic infants, identifying 7,115 infants in total. The caregivers of 2,704 infants (SEED OK children) agreed to participate in the study and completed baseline interviews by telephone (response rate of 38%). Study participants were located by a survey research firm, provided their child’s Social Security number (required for the SEED OK account), and completed the telephone survey. Thus, the study participants differ from nonparticipants. To assess the extent of bias from caregiver self-selection into participation, Nam, Kim, Clancy, Zager, and Sherraden (2013) compared the birth records of participants’ children with the records of nonparticipants’ children. These records include information on the caregivers, enabling a direct comparison. Participants and nonparticipants were similar on most available variables, and the remaining differences were small. Weight variables were created to adjust for oversampling of minority groups and observed participation bias (Marks, Rhodes, & Scheffler, 2008). However, if participants differed systematically from nonparticipants in ways not captured by birth records, postsurvey adjustment with weight variables might not have eliminated bias due to selection into the study. Self-selection affects external validity but not the comparisons between treatment and control groups—that is, the impact analysis.

After study participants completed the baseline survey, the caregivers (and their infants) were randomly assigned to either the treatment or control group. For simplicity, we refer to all study participants as mothers, even though the sample included one father and five nonparents. Children in the treatment group (n = 1,358) received the CDA in SEED OK, and children in the control group (n = 1,346) did not. The use of random assignment enables researchers to attribute differences in outcomes to the CDA rather than to individual characteristics. Statistical analyses using chi-square tests (for categorical variables) and F tests (for continuous variables) showed that the baseline socioeconomic characteristics of the treatment group did not differ significantly from those of the control group (Kim & Nam, 2009). This finding indicated that random assignment created two comparable groups (see Table 1). Notably, both random assignment to treatment and control groups and sampling from a full state population are uncommon practices in social research, and these study elements make SEED OK an unusually rigorous policy test.

The CDA in SEED OK

The CDA in the SEED OK experiment was built on the existing Oklahoma 529 College Savings Plan (OK 529). Similar to other 529 plans, the OK 529 was created to encourage families to save for postsecondary education. The plan offers a small selection of funds with varying levels of risk. Contributions up to $10,000 per year (or $20,000 for married couples filing jointly) can be deducted from state income taxes, and qualified withdrawals are exempt from federal and state taxes.

As shown in Figure 1, the CDA in SEED OK involves two types of OK 529 accounts: (a) accounts opened automatically for SEED OK treatment children and (b) accounts that mothers (treatment or control) could open for their children. Unless a treatment-group mother opted out (which occurred in one case for religious reasons), every child in the treatment group received an OK 529 account automatically opened with an initial deposit of $1,000. (For simplicity, we say that a child holds or owns an account if she or he is the beneficiary.) We call these automatically opened OK 529 accounts SEED OK accounts, and these accounts hold deposits and incentives provided as part of the SEED OK CDA. Funds in these accounts are to be used for children’s postsecondary education expenses. For each calendar quarter, the OK 529 sends an account statement for SEED OK accounts to each treatment child.

The second type of account can be opened by mothers for SEED OK children. These individual OK 529 accounts hold savings set aside by family and friends. For administrative reasons discussed by Nam et al. (2013), deposits made by individuals are kept separate from most deposits made by SEED OK. The ideal CDA program would automatically open a single account to hold public, private, and individual contributions, so parents would not need to open individual accounts; however, this ideal was not possible in SEED OK. To encourage treatment mothers to open their own OK 529 accounts for their infants, the CDA in SEED OK offered a time-limited $100 account-opening incentive. The OK 529 requires a minimum initial deposit of $100. Because this requirement could have prevented some mothers from opening accounts, SEED OK deposited the required $100 initial contribution for treatment participants who opened an OK 529 account for a SEED OK child by April 15, 2009. Treatment participants who opened accounts after that date had to provide the initial deposit themselves.

For about 4 years, low- and moderate-income treatment families with individual OK 529 accounts were eligible for 1:1 or 0.5:1 matches on their contributions to these OK 529 accounts. Households with an annual adjusted gross income less than 150% of the estimated median in Oklahoma ($29,000) were offered a 1:1 match: Every dollar deposited was matched with one dollar, up to $250 per calendar year. Households with an adjusted gross income between 100% and 150% of the state median (between $29,000 and $43,499) were eligible for a 0.5:1 match: Every dollar deposited was matched with 50 cents, up to $125. The state determined savings-match eligibility through official state records, so parents did not have to submit tax returns or self-certify their adjusted gross income (Zager et al., 2010).

Mothers in the treatment group received letters, postcards, and brochures that described both types of OK 529 accounts, introduced SEED OK financial incentives, and communicated messages about the importance of education. Mothers occasionally received small gifts (e.g., storybooks or educational music compact discs) for treatment children. The Oklahoma state treasurer’s office sent the mailings on behalf of SEED OK researchers. (Communications with treatment families are summarized in Gray et al., 2012, Appendix A.) SEED OK materials were provided almost exclusively by mail because the intent was to reach treatment families without exposing the control group to any aspect of the treatment. Distributing SEED OK information by mail probably suppressed participation. For example, mothers who moved during the experiment might not have received all communications, and some mother might not have read or understood the written materials (Gray et al., 2012). A universal CDA policy implemented statewide or nationally could use multiple media outlets (e.g., television and the Internet) and multiple messengers (e.g., public officials, social workers, and businesses) to encourage people to save for college.

In sum, the SEED OK experiment gave treatment children an automatically opened OK 529 account with an automatic $1,000 initial deposit, and gave treatment mothers information about the OK 529 and financial incentives to open and save in an OK 529 account for their young children. These components together make up the CDA in SEED OK. This CDA has many of the characteristics recommended by CDA proponents: automatic account opening and initial deposits, restrictions on the use of funds, and progressive matches. A centralized savings platform—a 529 college savings plan, in this case—facilitates the use of these CDA features (Clancy, Sherraden, & Beverly, 2015).

Although individual OK 529 accounts are publicly accessible through the state’s plan, mothers in the control group did not receive any information from SEED OK about OK 529 accounts and were not offered any SEED OK financial incentives. Children in the control group were not eligible for the automatic SEED OK account with initial deposit.

Sources of Data and Methods of Analysis

As noted above, 2,704 mothers completed the SEED OK baseline survey between fall 2007 and spring 2008, soon after their children were born. In the spring of 2011, when children were about 4 years old, 2,272 (84%) of these mothers completed a follow-up telephone survey. In addition, 60 mothers completed face-to-face semistructured interviews when their children were between 2 and 3 years old (between mid-2009 and late 2010; Gray et al., 2012). These mothers (40 from the treatment group and 20 from the control group) were recruited from 295 SEED OK mothers in a subsample randomly extracted from the mothers who completed the baseline survey, after researchers stratified the sample by race in order to oversample non-Whites (Gray et al., 2012). In addition to survey and interview data, SEED OK has data from children’s birth records and receives quarterly account data for all OK 529 accounts with SEED OK children as beneficiaries. Because the account data are transmitted directly from the OK 529 program manager, TIAA-CREF, the quality and reliability of these data are high.

Most analyses summarized in this article use weighted data. As noted above, these weights adjust for oversampling of minority groups and observed participation bias. In analyses that used data from the follow-up survey, weights also adjust for sample attrition (Schreiner, 2012). Unless otherwise noted, the results of all statistical tests reported in this article are significant at the .05 level or lower.

Impact on Accounts, Assets, and Savings for Children’s Postsecondary Education

Impacts on the Nonfinancial Outcomes of Parents and Children

Early evidence from the SEED OK experiment has suggested that the CDA has positive effects on parents and children. As previously noted, an important hypothesis is that CDAs raise parents’ expectations for their children’s education. Preserving high expectations and raising low expectations are essential if CDAs are to improve long-term academic outcomes. Kim, Sherraden, Huang, and Clancy (2015) examined the CDA’s impact on mothers’ educational expectations for SEED OK children. The researchers’ measure of expectations captured change between the baseline and the follow-up survey, which was conducted when children were about 4 years old. Kim and colleagues found that, as compared with mothers in the control group, a significantly greater percentage of mothers in the treatment group reported their educational expectations for their child remained stable or increased. Further, Kim and colleagues reported that the CDA in SEED OK reduced by nearly 3 percentage points the probability that mothers’ expectations declined between surveys. Moreover, another study by those authors provided evidence that the CDA in SEED OK raised parental expectations by prompting mothers to open their own OK 529 account for their young children (Kim, Huang, Sherraden, & Clancy, 2014). This finding suggests the CDA in SEED OK offers families a way to begin early preparations for future college expenses (Kim et al., 2014).

If CDAs change maternal attitudes about the future, as has been suggested by theory and some early evidence, then CDAs might also influence maternal mental health. Huang, Sherraden, and Purnell (2014) found that the CDA in SEED OK reduced depressive symptoms among mothers. Measured about 4 years after the start of the SEED OK intervention, the effect is similar in size (.11) to the effect of Early Head Start (.10; Chazan-Cohen et al., 2007). The impact of the CDA on mothers’ depressive symptoms was greater among several disadvantaged subgroups (e.g., low-income mothers and less educated mothers). Another study found that the CDA in SEED OK reduced depressive symptoms more among mothers experiencing material hardship than among mothers who were not experiencing such hardship (Huang, Kim, Sherraden, & Clancy, 2014a). In addition, research has suggested that the CDA reduces maternal depressive symptoms regardless of whether a mother has opened and saved in an OK 529 account for her child (Huang, Sherraden, & Purnell, 2014).

Changes in parental attitudes might improve parent–child interactions, and therefore, have positive effects on child development. Several studies have demonstrated that the CDA in SEED OK positively affected the social-emotional development of children at about 4 years of age, especially in some disadvantaged groups. For example, Huang, Sherraden, Kim, and Clancy (2014) found that the CDA significantly improved social-emotional development for children whose mothers possessed low levels of education, had low income, received public assistance, or rented their homes. The effect sizes for these subsamples were between .20 and .25, roughly similar to at least one study’s estimates of the effect of the Head Start program on social-emotional development (U.S. Department of Health and Human Services, 2010; estimated effect sizes range from .10 to .21). However, the effect of the CDA on social-emotional development was not statistically significant in separate racial and ethnic subgroups (i.e., White, African American, American Indian, Hispanic), perhaps due to inadequate statistical power.

In other studies, Huang, Kim, Sherraden, and Clancy compared children from single-parent headed homes with children from homes with two caregivers, and demonstrated that the CDA in SEED OK had a greater impact on the social-emotional development of children from households headed by single mothers (Huang et al., 2014b). Further, these researchers found that the CDA in SEED OK has had a greater impact on the social-emotional development of children in households that experienced material hardship than on children in households without material hardship (Huang et al., 2014c). The CDA positively influenced children’s social-emotional development regardless of whether mothers had opened and saved in an OK 529 account (Huang, Sherraden, Kim, et al., 2014).

Although researchers hypothesized that the CDA in SEED OK enhances child development, in part, by improving parent–child interaction, one study provided only weak support for this hypothesis. When children were about 4 years old, mothers in the treatment group reported less parenting stress and better parenting practices than mothers in the control group, but the effects were significant at the .05 level for just 1 of 7 measures: the number of times the mother had screamed at her child in the previous 3 days (Nam, Wikoff, & Sherraden, 2014).

Overall, the impact of the CDA on nonfinancial outcomes followed a pattern similar to the impacts on accounts and assets for college. The CDA improved a number of financial and nonfinancial outcomes, but the positive effects of the CDA were often greater for disadvantaged parents and children than for counterparts in advantaged households. Early evidence has suggested that the CDA has positive effects on two nonfinancial outcomes (i.e., maternal depressive symptoms and children’s social-emotional development), regardless of whether parents had opened or saved in an individual OK 529 account. The pattern is clear: The automatic components of the CDA are essential to the positive effects revealed thus far, and these automatic components are especially critical for children from disadvantaged households.

Themes from Extended Interviews with SEED OK Mothers

As previously noted, qualitative interviews with 60 mothers were conducted when SEED OK children were between 2 and 3 years old. Most of these mothers lived in households with low or moderate incomes (i.e., 72% had household income at or below $50,000). Poverty rates in Oklahoma are higher than national averages, and the rates among Oklahoma children younger than 5 years tend to be higher than the rates among Oklahoma children in all other age groups (U.S. Census Bureau, 2012, Table 709, p. 463).

The interviews demonstrated that mothers of young children recognized the value of saving. The mothers reported their most common motives for saving were short-term goals such as covering bills, responding to emergencies, meeting children’s needs, and saving for a “rainy day” (Gray et al., 2012, p. 30). Nearly all mothers expressed the hope that their children will attend college but also noted multiple barriers to educational attainment. Most identified barriers were social in nature (e.g., adverse influence from peers, falling in love, and having children). Mothers recognized that a college education is expensive. Almost half said they planned to finance their children’s education partly with savings, but few had started to save for educational expenses. Some treatment mothers identified challenges that deterred them from opening their own OK 529 accounts. These included lack of surplus income, lack of information on the accounts and incentives, and misunderstanding of the features of the OK 529 account. Mothers also noted that their children’s college years were far in the future (Gray et al., 2012).

Treatment mothers understood that the funds in the automatic, state-owned SEED OK account could be used only for their children’s postsecondary education. A number of mothers seemed to be more hopeful about their children’s future because of the SEED OK account and the $1,000 deposit. Some mothers said that the initial deposit indicated that someone outside the family cares about their children’s future, and several expressed a belief that the account offers “a sense of security—a little bit of relief that something has begun” (Gray et al., 2012, p. 55). For some mothers, receiving regular account statements and program materials helped them to “see their children as college bound” (Gray et al., 2012, p. 76). Even though the primary components were automatic, the CDA in SEED OK seems to have changed the attitudes of some parents in ways that might improve children’s educational outcomes (Gray et al., 2012).