Introduction

A primary explanation for disproportionately low rates of marriage among economically disadvantaged groups is the perception of a so-called economic bar to marriage. This economic bar is defined as a set of multiple markers of economic success, including earnings, employment, and asset achievements, and is the litmus test by which economically disadvantaged couples determine whether their relationships are ready for marriage (Edin and Kefalas 2005; Gibson-Davis et al. 2005). The existence of the bar has gained widespread acceptance and is often described in the literature on marriage among low-income couples as a primary barrier to marriage among that group (cf. Carlson et al. 2004; England et al. 2013; Manning et al. 2014).

We argue that the bar’s salience for marriage entry among low-income couples may have been overstated in the prior literature. Very little evidence exists as to whether meeting the bar is actually correlated with marriage entry, and several questions about the economic bar remain unanswered. For example, is attaining multiple types of economic success, as implied by the economic bar, necessary for marriage, or can marriage entry be predicted just as strongly by using only one measure of economic well-being, such as employment? Additionally, does the economic bar predict entry into cohabitation? If the economic bar were indeed related to cohabitation, then the theoretical importance of the bar would be undercut, insofar as it has been conceived as applying only to marriage (Gibson-Davis et al. 2005). Moreover, is meeting the economic bar associated with marriage, even if it is based on the mother’s economic contributions alone? The bar is described as a couple-level measure (Edin and Kefalas 2005; Rackin and Gibson-Davis 2017), yet most of the literature on family formation among low-income parents has stressed the importance of the man’s, rather than the woman’s, economic well-being (Harknett and McLanahan 2004; Smock et al. 2005). Finally, does meeting the economic bar have positive associations with parents’ assessment of their romantic relationship? Associations between meeting the economic bar and relationship quality would be notable insofar as parental relationship quality is positively correlated with healthy parent-child interactions (Goldberg and Carlson 2014).

Using three waves of data from the Building Strong Families (BSF) project, an experimentally evaluated relationship program among low-income parents (n = 4,444), we present the most systematic examination of the economic bar to date. First, we address the bar’s salience and definition by examining how meeting the bar is associated with four union types (marriage, cohabitation, dating, and not romantically involved) and by testing whether the bar predicts marriage over and above any one aspect of economic well-being. Second, we examine the importance of gender by investigating whether a bar defined based on the man’s economic accomplishments has the same associations with marriage as does a bar defined based on the woman’s accomplishments. Third, we examine the spillover effects of the bar by examining whether meeting the bar has positive associations with relationship quality.

In addressing these questions, we combine a well-established literature that evaluates the importance of men’s and women’s income for marriage entry (Oppenheimer et al. 1997; Sweeney 2002; Xie et al. 2003) with an emerging literature documenting the salience of wealth, homeownership, and job quality on marital transitions (Addo 2014; Schneider 2011; Schneider and Reich 2014). Collectively, these studies suggest that multiple economic domains influence marriage decisions. To date, however, studies have largely analyzed economic well-being variables in isolation. Our study, in contrast, incorporates numerous dimensions of financial status and provides the first quantitative analysis of how a set of markers, as measured through the economic bar index, is associated with relationship status and quality.

Background

The Economic Bar to Marriage

A central proposition in understanding marriage patterns among low-income parents is the economic bar to marriage. This economic bar, consisting of a set of employment, earnings, and financial markers, acts as a barrier to marriage because couples believe that only those who have met the bar should marry (Edin and Kefalas 2005; Gibson-Davis et al. 2005; Smock et al. 2005). Meeting the bar prior to marriage is considered important because economic standing is perceived to play a crucial role in marriage, and stress from unmet financial needs can tear relationships apart. Therefore, financial preparedness is an important hedge against an unstable marriage (Edin et al. 2004).

Achieving the bar also plays a normative function, signaling to friends and family that the couple has achieved what is necessary for marriage and is worthy enough to take on the institution (Cherlin 2004). In this regard, the existence of the bar is evidence of the ongoing shift in the cultural framing of marriage (Cherlin 2009). Marriage has become a capstone achievement, signaling that a couple has “arrived,” both financially and emotionally (Cherlin 2004; Edin and Kefalas 2005).

When low-income individuals describe the economic bar to marriage, they describe the importance of many factors, including employment, earnings, and asset accumulation. For employment, individuals indicate that parents should have “good” jobs—those that are stable, reliable, and provide benefits (Edin and Kefalas 2005; Gibson-Davis et al. 2005). The couple should also have sufficient earnings to avoid living paycheck to paycheck and public assistance receipt. Living paycheck to paycheck introduces stress into the relationship, and public assistance receipt indicates that the couple is not financially self-sufficient (Edin et al. 2004). Finally, the couple should have a sufficient level of assets, such that the couple can pay for a wedding or a down payment (Gibson-Davis 2007).

By including numerous markers of economic well-being in the bar’s definition, respondents are implicitly merging two bodies of work on marriage determinants. Numerous studies have focused on how employment and earnings relate to marriage entry (Carlson et al. 2004; Oppenheimer 2003; Sweeney 2002; Wilson 1987). Other work has considered job quality (as measured by occupation prestige, presence of benefits, union membership, or standardized work hours) (Oppenheimer et al. 1997; Piotrowski et al. 2015; Schneider and Reich 2014), the avoidance of material hardship and debt (Addo 2014; Gassman-Pines and Yoshikawa 2006), and asset accumulation and homeownership (Gassman-Pines et al. 2006; Schneider 2011). Measures of job quality, wealth, and material hardship do not appear to diminish the importance of earnings and employment but rather complement their effects (Oppenheimer et al. 1997; Schneider and Reich 2014). The complementarity of these factors is conceptually consistent with the multidimensional nature of the bar and suggests that fitness for marriage is assessed based on a set of markers rather than factors that operate in isolation.

Qualitative research has also demonstrated that low-income couples believe that this bar applies to marriage but not cohabitation (Rackin and Gibson-Davis 2017). Cohabiting is usually considered a less-serious relationship step than marriage (Huang et al. 2011; Sassler 2004). Additionally, most cohabiting unions, particularly those among disadvantaged individuals, do not begin as a precursor to marriage, and very few transition to marriage (Lichter et al. 2016; Sassler and Miller 2017). Cohabitation is conceptually and logistically distinct from marriage (Sassler and Miller 2017) and as such, it is not subject to the same economic litmus test (Manning et al. 2014; Sassler and McNally 2003).

Despite the consensus about the importance of the economic bar, only a few studies have examined whether meeting the economic bar is actually associated with marriage. Gibson-Davis (2007, 2009), using both qualitative and quantitative data from the Fragile Families and Child Wellbeing Study (Fragile Families), found that consistent with expectations, meeting the economic bar was positively correlated with marriage entry among low-income couples. However, these studies used an economic bar measure that included only two items (earnings growth and homeownership) and did not test whether the economic bar was associated with entry into cohabitation.

Notwithstanding the relative paucity of literature, the construct of the economic bar is gaining theoretical traction in the literature on family formation (England et al. 2013; Gassman-Pines et al. 2017; Manning et al. 2014; Schneider and Hastings 2015). We argue that the bar’s importance in the literature may have outpaced our understanding of its function, and that replication and additional analyses are needed to understand how the economic bar is related to relationship status transitions.

Male Versus Female Economic Contributions

Whether men’s or women’s economic circumstances matter more for family formation decisions has long been debated. Men’s employment and earnings have been found to consistently relate to marriage entry (Oppenheimer 2003; Smock et al. 2005; Xie et al. 2003); the evidence on women’s earnings is decidedly mixed (Lichter and Graefe 2007; Sweeney 2002). Theoretically, women’s earnings could have opposing effects on marriage entry. The independence effect suggests that women may be less likely to marry because they are better able to support themselves; the income effect suggests that higher earnings make women more attractive on the marriage market (Becker 1981; Hannan et al. 1978; Oppenheimer et al. 1997). Studies have generally supported the income effect: low-income women holding steady jobs are more likely to marry than those who do not (McLaughlin and Lichter 1997), and earnings are positively associated with marriage for young women in recent cohorts (Kuo and Raley 2016). Men’s labor market attachment increases the likelihood of marriage, over and above women’s, whereas women’s labor force participation may not have any effect on marriage entry, accounting for men’s earnings (Gibson-Davis 2009; Smock et al. 2005).

In contrast, in qualitative research, low-income individuals have described the economic bar as a couple-level indicator that does not necessarily prioritize the man’s financial position over the woman’s (Edin and Kefalas 2005; Gibson-Davis et al. 2005; Rackin and Gibson-Davis 2017). Both the man and the woman needed to contribute in order to meet the bar. Having enough money for a down payment, for example, was simply out of reach based on one partner’s earnings alone. Moreover, women reported needing to be economically independent of their partners so that they would have a fallback position should the relationship dissolve (Edin and Kefalas 2005). The bar was thus described as something that required contributions from both men and women.

To summarize, the qualitative work serving as the basis for the economic bar suggests that the gender of the contributor should not matter and that the important determinant of marriage entry is the couple-level achievement of the bar. In contrast, quantitative studies on marriage entry have generally found that couples appear to be more responsive to men’s than to women’s economic status when deciding on marriage timing.

Relationship Quality

Qualitative research on the economic bar to marriage has found that couples closely link achieving the economic bar and higher relationship quality (Edin and Kefalas 2005). Individuals reported that achieving the economic bar would have positive spillover effects on the tenor of their relationships and would serve as an important hedge against future stress and conflict (Rackin and Gibson-Davis 2017).

These findings echo long-standing research utilizing quantitative methods showing that economic strain decreases marital functioning (Liker and Elder 1983). Day-to-day economic stress depletes the emotional reserves that couples have for each other, leading to increased stress and conflict as well as lower levels of support and affection (Conger and Elder 1994; Conger et al. 1990). Economic pressure and hardship has been associated with relationship dissolution (Halliday Hardie and Lucas 2010; Wu and Pollard 2000) and relationship hostility (Masarik et al. 2016).

Given that financial stability also has implications for power dynamics within a couple (England and Farkas 1986), the associations between economic well-being and relationship quality may be particularly important to women, especially those who are socioeconomically disadvantaged. Women in lower-income relationships may perceive that men who are contributing financially to the couple (and not just to men’s own financial interests) are more dedicated to, and invested in, the relationship. Having a joint bank account, for example, may signify that the man is “all in” in the relationship and is not holding anything back (Addo and Sassler 2010). Low-income men, relative to women, may be less sensitive to the signal of financial stability for relationship quality.

Associations between economic stability and parental relationship quality matter because the tenor of the parents’ relationship has spillover effects on the parent-child relationship (Goldberg and Carlson 2014). Low-income parents with higher-quality relationships can better support and nurture their children (Berger and McLanahan 2015). The potential indirect association meeting the bar and parent-child relationships underscores the bar’s importance, above and beyond any associations it has with marital status.

Study Hypotheses

We hypothesize that meeting the bar will be positively correlated with marriage entry but will not be associated with cohabitation. Meeting the bar should also be a stronger predictor of marriage entry than any one item by itself. Consistent with the quantitative evidence, we expect that men’s achievement of the bar will be a stronger predictor of marriage entry than women’s achievement of the bar. We also hypothesize that meeting the economic bar to marriage will lead to increases in relationship quality, and that these associations may be stronger for woman than for men. Because having both members meet the bar is more difficult than having either parent meet the bar, we expect fewer couples to meet the both-parent bar. We also expect (but have little prior literature to guide our expectation) that couples in which both members individually meet the economic bar will be more likely to marry than couples in which the bar is measured across the couple.

Method

Data Source and Sample

Data come from the Building Strong Families (BSF) evaluation. Funded by the U.S. Department of Health & Human Services Administration for Children & Families as part of the Healthy Marriage Initiative, BSF was offered in eight sites in geographically diverse cities between July 2005 and March 2008 (for more information on the BSF program and evaluation, see Wood et al. 2012). The BSF data are ideally suited to examine the economic bar to marriage because it contains a rich longitudinal set of financial measures on a sample of low-income mothers and fathers as well as information on relationship quality and entry into marriage over time.Footnote 1

The sample comprises couples aged 18 and older who were romantically involved at the initial baseline survey, were expecting or had a newborn child (under 3 months of age), and were unmarried at the time of the child’s conception. An initial sample of 5,102 couples were recruited by approaching parents in prenatal clinics and hospital maternity wards, and through referrals from Women, Infants, and Children (WIC) and other social services agencies serving low-income families (Dion et al. 2010). BSF couples were randomly assigned to the treatment condition (offered relationship-skills training) or to a control group (no services offered). Evaluations of the BSF program found no treatment effects of the program on marriage or relationship quality (Wood et al. 2012). BSF participants were demographically similar to Fragile Families respondents (Dion et al. 2010).

Our sample consisted of couples who participated in the baseline survey and either the 15-month or 36-month follow-up survey, where participation was defined as at least one member of the couple being observed in at least one round. Couples who were married at baseline (n = 320) and did not participate in either follow-up survey (n = 338) were excluded. The final sample size was 4,444 couples. Slightly more than 87 % of couples had at least one member participate at 15 months, and 83 % of couples had at least one member participate at 36 months.

We used multiple imputation to account for missing data across measures, a method that replaces missing data with a probable value based on other available information from the dataset. Whereas levels of missingness on baseline data were less than 1 %, levels of missingness were substantially higher at follow-ups (up to 23 % for select father-reported variables and 11 % for mother-reported variables). Although multiple imputation cannot completely account for bias due to missing data, it improves consistency and efficiency compared with other methods (Young and Johnson 2011).

We imputed missing values using the chained equation method, generating 20 imputed data sets to impute missing data on all the categorical and continuous independent variables included in our data set. Following the recommended practice of including all measures that are used in analytic models (White et al. 2011), the imputation model included the variables used to construct the economic bar, the baseline covariates (detailed later), and treatment status and program site. Variables that were measured at both waves were each included. We conducted multiple checks on the stability of the imputation model and analyses and found the imputed data patterns to be highly consistent with the original data set.

Our analysis sample includes a small number of couples for whom outcomes were not observed at 15 or 36 months. Of the 4,444 couples, 315 couples had neither parent participate at 15 months (at least one member of these couples was observed at 36 months), and 480 couples had neither parent participate at 36 months (at least one member of these couples was observed at 15 months). Because the literature on multiple imputation offers conflicting advice as to whether cases that are observed in one round but not the other should be dropped (Young and Johnson 2015), we elected to keep couples who participated in either wave in order to maximize the sample size. Supplementary analyses found virtually no substantive differences in the 15- and 36-month results when the 315 and 480 couples, respectively, were dropped (although estimates were slightly less precise).

At baseline, nearly three-quarters of the sample reported living together all or most of the time (see Table 1), and the majority of couples (69 % of mothers and 74 % of fathers) believed that they would marry each other. Relationship quality scores (measured on a 1–4 scale, described later) were relatively high (greater than 3) for both mothers and fathers. The sample was relatively disadvantaged, insofar as only 7 % of mothers and 24 % of fathers reported annual earnings greater than $25,000. Approximately 40 % of the sample did not have their high school diploma or GED.Footnote 2 Nearly one-quarter of fathers reported not being employed, and three-quarters of mothers reported receiving Medicaid or State Children’s Health Insurance Program (SCHIP). The majority of parents identified as non-Hispanic black, and most couples already had one child together. Couples were in their mid-20s, with fathers being slightly older than mothers.

Table 1 Baseline characteristics of couples (n = 4,444)
Full size table

Measures

Relationship Status

At both 15 and 36 months, relationship status was defined using four mutually exclusive binary categories: married, cohabiting, dating, and broken up. Respondents who reported being married were treated as currently married. Cohabiting and dating couples reported being romantically involved; cohabiting couples reported living with their partner “all” or “most” of the time, whereas dating couples reported living with their partner “some” or “none” of the time. Broken-up couples reported not being romantically involved with each other. Each measure required affirmative responses from both members of the couple for the outcome to be considered true. In cases of conflicting responses, the less-serious relationship status was used (e.g., we coded a couple as cohabiting if one member said they were married but the other member said they were cohabiting). If only one member of the couple was observed, that person’s report was used.

Relationship Quality

Maternal and paternal relationship quality at 15 and 36 months was measured by the six-item Support and Affection scale (alphas for both mothers and fathers at both waves were .93 or greater). Items included “partner knows and understands me” and “partner respects me,” using a 1 (never) to 4 (often) response scale. Responses were averaged across the six items to form a single score. Relationship quality measures could conceivably differ for the mother and the father, so this measure was not combined into a couple-level measure.Footnote 3

The Economic Bar to Marriage Index

The economic bar to marriage, measured at 15 months, was a seven-item index capturing the set of available economic factors, ranging from 0 (met none of the criteria) to 7 (met all of the criteria). All items have been explicitly identified in qualitative studies as constituting the bar (Edin and Kefalas 2005; Gibson-Davis 2009; Gibson-Davis et al. 2005) or have been shown in quantitative work to be associated with marriage entry (Addo 2014; Schneider and Reich 2014). We combined each of the factors described later into a single index, unit-weighting each factor. Unit-weighting performs as well as, if not better than, other strategies for creating indices from combined multiple components (Graefe 2015).

Health insurance measured the presence of private insurance for either the parent or the child. A bank account and homeownership indicated the presence of those items for the parent. Earnings growth reflected higher reported earnings at the 15-month follow-up survey than at baseline.Footnote 4 Steady employment was defined as having worked at least 30 h per week over the past month. Avoidance of welfare receipt was indicated by parents who reported that they had not received Temporary Aid to Needy Families (TANF) or food stamps (now known as SNAP) in the past month (for both steady employment and welfare receipt, month was the only time frame referenced in the 15-month survey). Lack of material hardship, measured over the past 12 months, was based on responses to three items if (1) respondents had not been able to pay their rent or mortgage, (2) they had their utilities shut off, or (3) they had been evicted. Parents without material hardship were those who answered no to all three questions.

We constructed four definitions of the bar, based on who in the couple contributed the items. The either-parent bar was the summation of items for either the mother or the father (e.g., if either the mother or the father had earnings growth, the couple was scored 1). The mother-only (father-only) bar was the summation of only the mother’s (father’s) items. The both-parents bar was the summation of items that both members of the couple had.

An example clarifies these bar constructions. In the case where both mother and father had earnings growth, that couple scores a 1 for earnings growth in all four bar definitions. If neither member of the couple had earnings growth, the couple is scored 0. If mother has earnings growth but not the father, the couple scores 1 on the either bar, 1 on the mother-only bar, and 0 on the father-only bar and on the both-parents bar. If the father has earnings growth but not the mother, the couple scores 1 on the either bar, 1 on the father-only bar, and 0 on the mother-only bar and on the both-parents bar.

We define meeting the bar as having an economic bar index score of at least 4. Previous literature (Gibson-Davis 2009), beyond indicating that meeting the economic bar implies exceeding a threshold, has not specified where that threshold should fall. Couples may evaluate their fitness for marriage based on a subjective sense of whether they have achieved the bar rather than achieving a specific number of items. Even though 4 is an admittedly arbitrary cut-point, it is a reasonable threshold and signifies that couples had achieved the majority of items on the index. Following others (Watson and McLanahan 2011), we perform robustness checks on our cutoff to evaluate its performance as a meaningful threshold (detailed later).

Descriptively, 67 % of couples met the either-parent bar, relatively to only 14 % of couples who met the both-parent bar (see Table 2). Nearly one-half of couples met the father-only bar, whereas approximately one-third of couples met the mother-only bar. Across indices, the most common items that couples achieved were the absence of material hardship and no public assistance. The least likely item to be achieved was homeownership.Footnote 5 Fathers, relative to mothers, were more likely to be employed (65 % vs. 37 %), have earnings growth (39 % vs. 26 %), have private health insurance (45 % vs. 28 %), and avoid public assistance receipt (63 % vs. 46 %). Gender differences in other items were minimal.

Table 2 Economic bar to marriage index (n = 4,444)
Full size table

Baseline Covariates

Baseline covariates consist of demographic, relationship structure and quality, and mental health characteristics. Demographic characteristics are mother’s race and ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, or other race or ethnicity besides Hispanic); whether the father’s race/ethnicity differed from the mother’s race/ethnicity; mother’s and father’s age; and whether the mother and father have a high school diploma. Economic characteristics include employed (yes/no) as well as whether the mother received any of the following public assistance in the past year: TANF, SNAP, or Medicaid. The relationship structure characteristics are whether the couple was cohabiting full-time or most of the time, the number of children the couple had together, the number of children each partner had with someone else, and whether the mother was pregnant (as opposed to having already had the child). Baseline relationship quality is measured by a seven-item scale (alpha = .71 for the mother and .80 for the father). Parents were asked to rate statements, such as partner “shows love and affection toward you” and “will not cheat on you,” on a 1 (strongly agree) to 4 (strongly disagree) scale. Mental health is measured through a six-item scale used to assess depression (mother alpha = .71, father alpha = .70).Footnote 6

Analytic Strategy

Using multinomial logistic and linear regression, respectively, we model relationship status and quality at 15 and 36 months as a function of the economic bar at 15 months, baseline covariates, and program site.Footnote 7 Treatment status is also included as a covariate, even though evaluations of BSF found no treatment effects (Wood et al. 2012). These models were used to generate predicted values of relationship status and quality for those who did and did not meet the bar (predicted values generated with covariates held constant). Results for the multinomial logistic regressions used in the predicted probabilities of relationship status are presented in the online appendix, Table A2Footnote 8; the linear regression models used to generate predicted relationship quality scores are presented in Table 3. All models were weighted using weights developed by BSF evaluators (Moore et al. 2012). Analyses were run separately for each bar definition.

Table 3 Impact of economic bar on relationship quality (n = 4,444)
Full size table

In select cases, we compared the estimates associated with meeting one bar with the estimates associated with meeting a different bar. A comparison of bar definitions for relationship quality was straightforward given that linear regression coefficients can easily be compared across models. In contrast, the multinomial logistic regression coefficients used to estimate relationship status cannot be directly compared across models (Karlson et al. 2012; Mood 2010). We addressed this problem by using linear probability models to estimate the association between meeting the bar and one relationship category (i.e., using a binary indicator of marriage). Coefficients from linear probability models can be compared using standard techniques.

Results

Relationship Status and the Economic Bar

The first set of results (Fig. 1) presents, for parents who met and did not meet the either-parent bar at 15 months, the predicted probabilities of their relationship status at 15 and 36 months. Apart from associations with the bar, these relationship results are consistent with other work on low-income parents (Carlson et al. 2004; McLanahan 2011), insofar as relatively few couples were married at either time point, with relatively high fractions of couples dissolving their relationships.

Fig. 1
figure 1

Predicted probabilities of relationship status at 15 and 36 months, meeting the economic bar, either-parent definition (n = 4,444). Data labels that are underlined differ from each other by the cutoff score (p < .05). Results control for demographics, economic factors, and relationship characteristics of parents

Full size image

At both time points, using the either-parent definition of the bar, couples who met the bar were significantly more likely to marry than couples who did not meet the bar.Footnote 9 At 15 months, meeting the bar was associated with a 47 % increase in the likelihood of marrying (13.4 % vs. 9.1 %); at 36 months, meeting the bar was associated with a 34 % increase (17.2 % vs. 12.8 %). At both time points, meeting the bar was also associated with a decrease in the probability of cohabiting. Given that the relationship categories were mutually exclusive, the decrease in cohabitation, when coupled with the increase in marriage, suggests that that the bar increased the likelihood that cohabiting couples moved into marriage. The bar had null or minimal associations with the other two relationship categories; regardless of whether parents met the bar, approximately 40 % of couples at 15 months and 55 % of couples at 36 months were broken up or dating.

Analyses using linear probability models to facilitate cross-model comparisons indicated that meeting the both-parent bar was associated with larger increases in marriage than meeting the either-parent bar (results not presented but available upon request). At both 15 and 36 months, the share of couples who met the both-parent bar and were married was higher than the share of couples who met the either-parent bar and were married; the difference between the either- and both-parent bars association with marriage was statistically significant (p < .05) for both time points.

Male Versus Female Contributions to the Economic Bar

Figure 2, panels a and b show the association between meeting the bar and relationship status by the gender of the parent who met the bar (e.g., the mother-only vs. father-only bar). The mother-only and father-only bar had similar positive associations with marriage entry. At 15 months, in the mother-only bar definition, the predicted probability of marriage for couples who met the bar was 15.4 %, a statistically significantly higher estimate than the 10.6 % of couples who did not meet the bar and married. For the father-only bar, the predicted probability of marriage was 15 % for those who met the bar and 8.9 % for those who did not (p < .05 for the difference between the two). Linear probability models indicated that association between meeting the mother-only bar and marriage was not statistically different from the association between meeting the father-only bar and marriage.

Fig. 2
figure 2

Predicted probabilities of relationship status at 15 months (panel a) and 36 months (panel b), meeting the bar and contributor’s gender (n = 4,444). Data labels that are underlined differ from each other by the cutoff score (p < .05). Results control for demographics, economic factors, and relationship characteristics of parents

Full size image

The father-only and mother-only bars differed somewhat in their associations with cohabitation at 15 months. Meeting the father-only bar, but not meeting the mother-only bar, was associated with a significant decrease in cohabitation. In linear probability models, the difference between father-only and mother-only bars’ association with cohabitation was marginally significant (p < .10). Keeping in mind the increase in marriage, the decrease in cohabitation for the father-only bar suggests that meeting the father-only bar encouraged cohabiting couples to marry.

At 36 months, the pattern of results for marriage entry mirrored those at 15 months, insofar as the share of couples who were married were similar between the mother-only (18.5 %) and father-only (19 %) bars. For each bar, couples who met the bar were significantly more likely to marry than couples who did not (p < .05). In a linear probability model with marriage as the binary dependent variable, the magnitude of meeting the father-only bar on marriage entry was larger than meeting the mother-only bar (p < .05).

Meeting either the mother- or father-only bars had small and generally nonsignificant associations with the nonmarital relationship categories at 36 months. Meeting the mother-only bar was associated with a small but negative effect on cohabitation, whereas meeting the father-only bar was associated with a small but positive effect on cohabitation. Neither of these effects reached statistical significance at conventional levels, and linear probability models indicated no statistically significant difference between father-only and mother-only bars’ association with cohabitation. Both the mother- and father-only bars had very modest negative effects on dating or being broken up (the only association that was statistically significant at conventional levels was for the father-only bar and dating). When considered with the positive associations on marriage, results indicate that movement into marriage for those meeting the mother-only bar came from couples who were broken up or dating, whereas movement into marriage for those meeting the father-only bar came from couples who were cohabiting, broken up, or dating.

Relationship Quality and the Economic Bar

Thus far, results suggest that all four versions of the economic bar were associated with increased odds of marriage, with minimal differences in the odds of marriage between the father-only and mother-only bar. We next evaluated how meeting the economic bar was related to self-reported relationship quality at both 15 and 36 months (Table 3).

Across bar definitions, the both-parent bar was consistently and positively associated with relationship quality. Meeting the both-parent bar was associated with statistically significant increases in maternal and paternal relationship quality at 15 months (standard deviation (SD) = .13 for mothers and .07 for fathers.) as well as increases in maternal relationship quality at 36 months (SD = .14). In contrast, the either-parent bar was not statistically significantly related to either gender’s relationship quality at either 15 or 36 months, with quite small (e.g., .01) effect sizes.Footnote 10

When considered by the gender of the parent, results indicated that when mothers contributed to the economic bar, they reported significantly higher relationship quality. In the mother-only bar, at 15 months, mothers in couples who met the bar reported relationship quality that was .05 SD higher than the relationship quality scores of mothers who had not met the bar. At 36 months, that difference was .07 SD.Footnote 11

In contrast, meeting the bar had only weak associations with father’s relationship quality. Meeting the mother-only, father-only, or either-parent bar was not associated with paternal relationship quality. The results for father’s relationship quality did not differ at conventional levels of statistical significance from those of mother’s. Nevertheless, results are suggestive that mother’s, but not father’s, relationship quality was more likely to improve in response to the bar, and that may be particularly the case if the mother herself contributed to the bar.

Robustness Checks

In supplementary analyses, we tested how each of the seven individual components of the index related to the transition to marriage. We used a logistic regression model to predict marriage entry, controlling for the same set of covariates used earlier and including each index item into seven successive models.Footnote 12 Results are presented in Fig. 3, which shows the odds ratios of marriage for meeting the either-parent bar (the bar labeled “all items”) and then odds ratios for each item. Bars that are filled in indicate that the item was significantly associated with marriage at p < .05; odds ratios that are underlined indicate a coefficient that is significantly different at p < .05 from that of the either-parent bar.

Fig. 3
figure 3

Odds ratios predicting the likelihood of marriage, comparison of all bar items with individual bar items, 15 and 36 months (n = 4,444). Solid bars indicate statistically significant relative risk ratios (p < .05). Data labels that are underlined differ from the either-parent economic bar cutoff (p < .05). Models control for demographics, economic factors, and relationship characteristics of parents

Full size image

Relatively few of the individual index items were by themselves statistically significantly related to marriage entry. At 15 months, having a bank account, avoiding public assistance receipt, having health insurance, and homeownership were statistically significantly related to marriage; at 36 months, only bank account and avoiding public assistance were. Notably, these items (with the exception of bank account) did not produce odds ratios that were statistically larger than that of meeting the either-parent bar. Having a bank account may be strongly predictive of marriage because of its close correlation with the other bar items (for example, having a bank account is necessary for buying a house, facilitates receiving earnings, and so on).

Importantly, consistent with expectations, couples in which either parent had experienced earnings growth were no more likely to marry than those without earnings growth. The association between the economic bar and marriage was significantly larger than the association between earnings growth and marriage (p < .05).

We also constructed the economic bar index seven times, with each successive iteration dropping one of the seven items; that is, we excluded health insurance the first time we constructed the bar, included health but excluded earnings growth the second time, and so on. At 15 and 36 months, each of these seven alternative bars was significantly related to marriage. Thus, the economic bar construct does not appear to be driven by any one item.

Finally, we tested the robustness of having four items as the cutoff point for bar achievement. If the appropriate cut point was four items, then we would expect to see a jump in marriage probabilities from having three to having four items. Additionally, when compared with the marriage probabilities associated with having four items, marriage probabilities associated with having zero, one, two, or three items should be smaller and statistically significantly different.Footnote 13

We began our test by computing the predicted probability of marriage at 15 months for couples at each value of the bar (see Table 4). Couples who reported either zero or one item were collapsed because very few couples reported zero items. We then compared whether the predicted probability of marriage when the couple reported zero/one, two, three, five, six, or seven items was statistically different than the predicted probability when the couple reported four items. This analysis was repeated for each version of the bar. We also redid this analysis using three or five as the reference category (results available upon request).

Table 4 Comparison of association between marriage and individual items at 15 months (n = 4,444)
Full size table

Results for three of the four bars were consistent with expectations. In the either-parent bar, the predicted probabilities of marriage for those with the zero/one, two, and three items were similar (8.6 %, 8.8 %, and 9.2 %, respectively) but increased by 3.6 percentage points for those with four items (12.8 %). Additionally, the predicted probability of marriage for those with the zero/one, two, and three items was statistically different from the predicted probability for those with four items; in supplementary tests, the zero/one, two, and three categories did not differ from each other.

The father-only and both-parent bar exhibited a similar pattern of results, although the difference between the three- and four-item category in the both-parents bar was not statistically significant at conventional levels. For the mother-only bar definition, results were less convincing. Although the probabilities for zero/one and two items were smaller and statistically significantly different from that of four items, the increase in predicted probabilities between three and four items was small (only 1.9 percentage points), and the two categories did not differ significantly from each other. Supplementary analyses for this bar, however, in which three or five items served as the omitted category did not suggest that those two cutoffs were statistically more robust than the four-item threshold, insofar as the predicted probabilities of all three item categories (three, four, or five) were statistically indistinguishable from each other. Results were thus less supportive of a clear cutoff point for the mother-only bar but suggest that a four-item cutoff is a reasonable threshold for the other definitions of the bar.

Discussion

In this study, we examined the economic bar to marriage, a commonly referenced yet arguably understudied concept in the literature on marriage among low-income couples. Our results suggest that the enthusiasm for the marriage bar is warranted, insofar as meeting the bar was positively associated with entry into marriage. Importantly, the bar’s associations with marriage could not be accounted for by changes in earnings or employment, underscoring the bar’s legitimacy as distinct from individual key economic indicators. Additionally, meeting the bar did not make couples more likely to cohabit—and in some definitions, the bar was associated with decreases in cohabitation—which is consistent with the idea that the economic bar pertains only to marriage entry (Gibson-Davis et al. 2005). Positive associations with marriage were robust to different definitions of the bar and different index items used in bar composition. In short, the economic bar passed our theoretical and empirical tests and performed as an important predictor of marriage entry among low-income couples.

One of our key contributions was to examine the gender of the parent who achieved the economic bar, and our findings add nuance in this area. We found evidence of the importance of both mothers’ and fathers’ economic contributions in marriage transitions. Low-income couples have described the economic bar as a couple-level indicator that is not unique to either the male or female (Edin and Kefalas 2005; Rackin and Gibson-Davis 2017). Consistent with those findings, at 15 months, relative to couples who did not meet the bar, couples who met either the mother-only or the father-only bar were significantly more likely to be married rather than cohabiting. Yet the magnitude of association with marriage entry was significantly larger for the father-only than the mother-only bar, and at 36 months, only the father-only bar remained significantly related to marriage. Thus, like others (Smock et al. 2005), our findings suggest that father’s economic contributions may be more important than mother’s in determining marriage entry. Nevertheless, consistent with the bar’s theoretical underpinnings, the bar’s association with marriage was not driven solely by the father’s contributions.

Achieving the economic bar was positively associated with mothers’ reports of relationship quality. Meeting the bar was associated with small increases in relationship quality in the two versions of the bar in which, by definition, the mother had to have contributed. For the two versions of the bar in which the mother did not necessarily need to contribute, meeting the bar was not associated with relationship quality. Thus, in contexts where the mother reported making positive economic gains, she also reported feeling more support and affection in her relationship; when those gains were not present, a mother reported no difference in the tenor of the relationship. The associations with maternal relationship quality were small, and when compared with father’s, were not statistically significantly different at conventional levels. Nevertheless, the findings highlight the importance of achieving the economic bar for maternal relationship quality.

Our study has limitations. Results are descriptive, not causal, and may be biased by unobserved factors. Data limitations meant that we could not include other important components of economic well-being, such as a couple’s asset levels, and our three-item measure of material hardship may be insufficient to measure economic hardship more generally. We could evaluate only one dimension of relationship quality (support and affection), leaving other important aspects (e.g., conflict or violence) unobserved. Finally, the sample was composed of voluntary participants in a relationships-skills intervention program. Results cannot be generalized to all low-income couples with a nonmarital conception.

Our findings raise additional issues for future research. Our seven-item scale may reflect the presence of some underlying, unobserved propensity. This unobserved factor may be economic in nature, but it also could be something unrelated to economic circumstances, such as self-efficacy. It is also possible that the economic bar, like indices of adverse childhood experiences (e.g., Felitti et al. 1998) or cumulative risk (e.g., Evans et al. 2013), represents the accumulation of related but distinct factors rather than representing an underlying construct. Our data do not permit us to disentangle these issues, but we encourage future research in this area.

Although not our study’s focus, we also found that very few couples moved into marriage. Even maintaining a romantic tie was difficult: by 36 months, the modal relationship category was being broken up. Our results are consistent with Fragile Families relationship trajectories (McLanahan 2011) but differ from that sample insofar as all BSF couples were romantically involved at baseline. Thus, even in a sample in which everyone has had a child in the context of a romantic relationship, nearly 25 % will be broken up within 15 months, and more than 40 % will dissolve within three years. The fragility of parental relationships highlights the challenges policy-makers face in encouraging union formation among low-income couples.

The encouraging news for those wishing to promote union formation among low-income parents is that meeting the economic bar was associated with marriage entry. We also provide encouraging results for scholars whose enthusiasm for the bar may have outpaced the evidence for its existence. Our quantitative data confirm the existence of this qualitatively derived concept, suggesting that couple-level economic progress may play a role in marriage formation.