Abstract
When a romantic relationship becomes serious, partners often confront a foundational decision about how to organize their personal finances: pool money together or keep things separate? In a six-wave longitudinal experiment, we investigated whether randomly assigning engaged or newlywed couples to merge their money in a joint bank account increases relationship quality over time. Whereas couples assigned to keep their money in separate accounts or to a no-intervention condition exhibited the normative decline in relationship quality across the first 2 years of marriage, couples assigned to merge money in a joint account sustained strong relationship quality throughout. The effect of bank account structure on relationship quality is multiply determined. We examine—and find support for—three potential mechanisms using both experimental and correlational methods: merging finances (1) improves how partners feel about how they handle money, (2) promotes financial goal alignment, and (3) sustains communal norm adherence (e.g., responding to each other’s needs without expectations of reciprocity). While prior research has documented a correlation between financial interdependence and relationship quality, our research offers the first experimental evidence that increasing financial interdependence helps newlyweds preserve stronger relationship quality throughout the newlywed period and potentially beyond.
Keep a joint bank account … It’s no longer “his and her money.” The officiant said, “Two become one…” Don’t keep separate accounts. Put all of your money together and begin to look at it as a whole.
—Rachel Cruze (2021), RamseySolutions.com
You need to maintain your own financial identity. This wisdom has been applicable since men were rolling around stones in caves. Your stone, my stone. Your account, my account.
—Kevin O’Leary (Montag 2018), Shark Tank
Marriage involves merging aspects of your life with another person, including, perhaps, your finances. Whether to do so is not obvious, nor is the relevant advice consistent. Money is already a leading cause of arguments between partners (Albrecht 1979; Amato and Rogers 1997; Dew, Britt, and Huston 2012)—couples certainly do not want to make matters worse by choosing the wrong financial account structure. Thus, couples are left wondering what is best for their relationship: to merge or not to merge?
Some academic research suggests that joint accounts might be beneficial, on average. Namely, a small body of nonexperimental research demonstrates a positive correlation between having a joint bank account and relationship quality (Addo and Sassler 2010; Kenney 2006), an interesting association with ambiguous causality. Gladstone, Garbinsky, and Mogilner (2022) identify a particularly intriguing pattern: in a large, longitudinal cohort study, couples who initially reported that they “pool all money” were significantly more likely to still be together 12–14 years later than were couples who initially reported that they “keep all money separate.” These findings offer the strongest suggestive evidence to date that bank account structure exerts a causal effect on relationship quality.
In this article, we seek greater clarity into whether and why joint bank accounts are good for marriage. Specifically, we provide the first experimental test of how couples’ bank account structure influences relationship quality over time (study 1). Using a longitudinal field experiment, we randomly assigned engaged or newlywed couples to merge their money in a joint bank account, to keep money in separate bank accounts, or to a condition where they received no instructions about how to structure their finances. We followed these couples for 2 years, periodically measuring their relationship quality and the extent to which they were satisfied with how they handled and discussed money. In study 2, we surveyed a separate sample of married adults to better understand how bank account structure might influence relationship dynamics.
THEORETICAL GROUNDING
Given the uncertainty around which bank account structure is best for marriage, it is no surprise that different couples make different decisions about how to manage household funds. Across a range of samples over the past 20 years, an estimated 52–65% of married and cohabiting male–female couples in Western nations report only using joint bank accounts (Addo and Sassler 2010; Gladstone et al. 2022; van Raaij, Antonides, and De Groot 2020; Vogler, Brockmann, and Wiggins 2006), though there is variation in when they begin to do so (e.g., some couples open a joint account as soon as possible, others wait until later in the union). About 10–15% of married and cohabiting male–female couples report that they maintain completely separate bank accounts. (It is difficult to determine how many of these couples believe that separate accounts are best and actively chose such an arrangement, and how many of these couples intend to open a joint account at a later date.) Some recent reports suggest that separate accounts are becoming more popular among younger Millennial couples (Bank of America 2018; Kitchener 2018). Other couples use some combination of joint and separate accounts (held by one or both partners).
There are at least three reasons why joint accounts might improve romantic relationships. First, relative to separate accounts, joint accounts are better at prompting people to consider how they might justify their purchases to their partner (Garbinsky and Gladstone 2019). As a result, couples who adopt joint accounts might begin to make more conservative, easily justifiable purchases. Over time, this could reduce difficult-to-justify purchases (as well as the arguments that those purchases can cause once revealed) and improve couples’ financial well-being.
Second, adopting joint accounts might change the way romantic partners talk about money. The process of opening and using a joint account forces partners to be more transparent about how they spend money—there are simply fewer opportunities to conceal troubling purchases or spending habits when using a joint account. The openness required when using a joint account may lead to more productive conversations between partners. For instance, once partners “get everything out in the open,” they may better understand each other’s priorities (e.g., whether it is more important to save as much money as possible or to enjoy life as much as possible without worrying so much about money). Ultimately, these conversations might help partners align their financial goals and “get on the same page” regarding how they plan to spend and save money. As Fitzsimons, Finkel, and vanDellen (2015, 656) note, “learning about each other’s goals makes it possible for partners to develop joint goals.”
Third, merging money in a joint bank account might help preserve the communal nature of the relationship. Most marriages and intimate unions are characterized by communal relationship norms whereby partners respond to each other’s needs as they arise, without expectations of reciprocity (Clark and Mills 1993). A communally-oriented spouse supports their partner because their partner needs support, not because they are pre-paying for later favors. For example, one partner might volunteer to pick up the kids from school so the other (who typically chauffeurs the children) can make an important work deadline. Communal relationship norms are contrasted against exchange relationship norms whereby partners act with the expectation of reciprocity. An exchange-oriented spouse might cook a homemade meal for the family, expecting a “gold star” for effort and for their partner to do the same tomorrow night. Exchange relationships are more common in business and professional environments. For example, being nominated for an award by one’s colleague might compel the nomination recipient to return the favor next year. Partners in an exchange relationship keep a running tally of each partner’s inputs and outputs; communal partners go out of their way to avoid scorekeeping (Clark 1984).
One might hope that scorekeeping tendencies become less prevalent as partners grow closer and more communal over time. Such a trend is the common intuition (Clark and Chrisman 1994, 82). However, there is good reason to suspect the opposite pattern, such that adhering to communal norms might be necessary to establish the relationship in the first place. Once partners are not worried about the relationship dissolving, they may let their guard down, becoming less vigilant about communal norms. Rather than asking your partner to clean the dishes tonight because you are exhausted, you might suggest that it is their turn because you cleaned them last night. Indeed, research that follows couples over time finds that adherence to communal norms decays as partners progress from engagement to early marriage (Clark et al. 2010). This trend is unfortunate, as greater communal norm adherence and lower exchange norm adherence predict greater relationship satisfaction during these early years (Clark et al. 2010).
These patterns are consistent with cultural assumptions about a brief “honeymoon period” in marriage. On average, relationship quality peaks around the wedding day and then begins a long-term decline that is especially steep in the early years (Karney and Bradbury 1995). The extent of this early decline is consequential; indeed, researchers have characterized the first 2 years of marriage as the connubial crucible, a period during which couples’ relational dynamics “foreshadow their long-term marital fate” (Huston et al. 2001, 237). Early interventions are crucial for changing the trajectory of a relationship (Joel and Eastwick 2018). To the extent that merging finances into a joint account helps minimize the natural decline in communal norm adherence over time (Clark et al. 2010), we propose that a joint account may help minimize the natural decline in relationship quality over time.
EMPIRICAL OVERVIEW AND CONTRIBUTIONS
The present research uses experimental and correlational methods to examine whether and why couples’ financial arrangements—an important but understudied aspect of marital life—have implications for relationship quality. Study 1 tests whether couples’ financial arrangements can mitigate the corrosive effect of time on relationship quality. Using a longitudinal field experiment, we randomly assigned couples to merge their money in a joint bank account, to keep their money in separate bank accounts, or to a condition where they received no instructions about how to structure their finances. We recruited only engaged or newlywed couples for this experiment, since those couples’ relationship dynamics are not yet set in stone and are presumably responsive to intervention (Huston et al. 2001; Joel and Eastwick 2018). We followed these couples for 2 years, conducting six waves of data collection across the connubial crucible. We measured changes in relationship quality and financial harmony (i.e., the extent to which partners are satisfied with how they handle and discuss money). We find that couples who transition to a joint bank account are buffered against the normative decline in relationship quality observed within longitudinal research on marriage, an effect due in part to positive changes in financial harmony.
To better understand some of the many ways in which bank account structure might influence relationship quality (in addition to changes in financial harmony; study 1), we surveyed a separate sample of married individuals to identify correlates of bank account structure (study 2). In particular, we examined the relationships among bank account structure, financial harmony, financial goal alignment, and communal norm adherence. For study 2, we recruited individuals from marriages of various lengths (i.e., not just newlyweds) because we wanted to get a glimpse into the day-to-day experience of a broader range of marital couples with different banking arrangements. If the effects of bank account structure take time to develop, examining only newlyweds (as we did in our intervention-focused study 1) could paint a misleading picture. We find that married individuals with completely merged bank accounts (vs. partially merged accounts or completely separate accounts) are more communal, more aligned in their financial goals, more transparent about money, and report greater financial harmony.
Our research has both theoretical and practical significance. Our novel methodology provides the first experimental test of whether merging finances is good for marriage. By following couples over time, we were able to observe whether merging finances has sustained effects on relationship well-being across the connubial crucible (Huston et al. 2001). Importantly, we identify merging finances as one “early intervention” (Joel and Eastwick 2018) that can slow the rate of typical relationship quality decline and place couples on a better path into the future. Second, our research is the first to document a relationship between bank account structure and communal norm adherence. While prior work has theorized that joint accounts move couples away from a market-like exchange of resources toward something more “collectivized” (Treas 1993), this mechanism was not explicitly tested. Third, our research responds to calls for further research on consumer behavior within close relationships (Cavanaugh 2016; Gorlin and Dhar 2012; Liu, Dallas, and Fitzsimons 2019; Olson and Rick 2022; Simpson, Griskevicius, and Rothman 2012). In particular, there has been growing interest in studying how couple members jointly navigate saving, spending, and borrowing decisions (e.g., buying hedonic or utilitarian products, Garbinsky and Gladstone 2019; choosing an automotive loan, Ward and Lynch 2019), as well as how spouses conceal financial behavior from each other (Garbinsky et al. 2020). We examine the relationship implications of a foundational financial decision that couples often confront when the relationship becomes serious.
The current findings have substantive implications for three audiences. First, our research speaks to consumers themselves, many of whom must, at some point, decide whether to merge their money with relationship partners. This is a decision most young adults will eventually face—in the United States, over 80% of adults marry by age 40 (Goodwin, McGill, and Chandra 2009). Second, the demand for “financial therapy” has never been higher (Sullivan 2020), and financial therapists and other professionals who serve couples in an advisory role (e.g., financial planners, marriage counselors) will likely find value in our results. For example, prior work demonstrates that arguments over money are a leading predictor of marital distress and divorce (Amato and Rogers 1997; Dew et al. 2012; Madden and Janoff-Bulman 1981). Our results raise the possibility that financial planners who recommend a joint account to newlywed clients may ease financial tension. Third, our research may offer some guidance to religious organizations (e.g., those who offer premarital counseling services) and government agencies (e.g., the Office of Family Assistance) in designing programs that promote strong marriages.
All materials (e.g., bank account instructions, complete measures from the intake survey and all follow-up surveys, study 1’s pre-registered data analytic plan), model specification, example syntax, and ancillary analyses are posted on Open Science Framework (OSF; https://osf.io/2m7sv/). In addition to providing complete measures on OSF, we present a summary of all measures collected in study 1 at each timepoint in web appendix C. Research protocols for studies 1 and 2 were approved by the Institutional Review Board at the University of Michigan and Indiana University, respectively.
STUDY 1: ARE JOINT OR SEPARATE BANK ACCOUNTS BETTER FOR RELATIONSHIP QUALITY?
Participants
We advertised a “Transition to Marriage Study” via social media, the volunteer section of Craigslist, word-of-mouth at bridal shows, wedding blogs, and mailing letters to customers on a jeweler’s mailing list. We described the study in broad terms and indicated that couples could earn up to $250 at the end of 2 years (those who completed a subset of the waves earned a prorated amount). We selected $250 because we thought it would be appealing enough to draw interest among potential couples, but small enough not to unduly sway couples’ willingness to participate, particularly given the degree of involvement required. In the advertisement, we asked interested adults to individually complete a brief online survey to assess their eligibility. The eligibility screener asked respondents to report their relationship status and duration, how they and their romantic partner manage their money, some demographic information, and their e-mail address. We indicated that the study might involve being asked to change the account structure they and their partner use to manage their money, and we asked whether respondents would be open to such a study.
In terms of eligibility criteria, both partners had to be entering their first marriage, currently living in the United States, and currently be either (1) engaged to be married or (2) married for less than 1 year. The couple had to consist of one male and one female. They were also required to have entirely separate bank accounts at study entry (participants were unaware of this eligibility requirement when completing the screener) and provide a valid e-mail address. The most common reason potential couples were screened out from further participation was not providing a valid e-mail address, which was followed closely by not having completely separate bank accounts (see web appendix B for the percentage of potential participants failing to meet pre-specified eligibility criteria). We note that having separate accounts was still the norm among couples at this stage in their relationships—among adults reporting a bank account structure, 43.8% reported having completely separate accounts, 32.5% reported having a blend of separate and joint accounts, and 23.6% reported having completely joint accounts.
We then e-mailed eligible respondents and asked them to discuss the prospect of participating with their partner. In that e-mail, we invited eligible respondents and their partner to read through and (electronically) sign the informed consent document. After providing their informed consent, both partners completed an intake survey. We ultimately enrolled 230 male–female couples (460 individuals; see web appendix A for the flow of participants over the course of the experiment). On average, partners had known each other for 5 years and were romantically involved for 3 years. Ten percent of couples had children. Partners’ median age was 28 (range: 18–58), and the median household income was $50,000 (range: $0–$450,000); 52% had earned at least a bachelor’s degree, and 74% identified as White or Caucasian. Complete descriptive statistics for our final sample are posted on OSF.
Procedure
Once both partners completed the intake survey (month 0), we randomly assigned couples to one of three bank account structure conditions. We informed couples via e-mail and provided condition-specific instructions. In the Separate Condition, we instructed couples to continue using separate checking and savings accounts and to not open joint checking and savings accounts, for the duration of the 2-year experiment. In the Joint Condition, we instructed couples to open joint checking and/or savings account(s) (within 1 month of them receiving these instructions), use those joint account(s), and discontinue using their separate account(s) for the duration of the 2-year experiment. In the No-Intervention Condition, we told couples that they could manage their money however they liked for the duration of the 2-year experiment. The stages of the experiment and incentive structure are outlined in figure 1.
EXPERIMENT TIMELINE AND INCENTIVE STRUCTURE (STUDY 1)
Note.— Amounts listed above are for each couple. Both partners had to complete the intake survey and all follow-up surveys to obtain the full amounts listed here. Because partners were paid individually, however, each partner of a couple who completed all aspects of the experiment received three checks totaling $125.
Across all three conditions, both partners were asked to independently complete surveys six times over the 2 years of the study—at months 0 (intake), 3, 6, 9, 12, and 24 (as clocked from study entry). At each wave, they completed various measures, many of which were included for the purposes of conducting exploratory analyses. Before analyzing the data, we pre-registered that we would focus our analyses on two key dependent measures—relationship quality and financial harmony—and we pre-registered how these measures would be constructed. We selected items for each composite that (1) were measured at all six timepoints (i.e., some potentially relevant scales were not included because they were only measured at three timepoints) and (2) we believed, a priori, reflected aspects of the same underlying construct. We also note that creating indices of different scales is a common practice—relationship researchers sometimes average related scales to capture a high-level overview of relationship well-being (Bohns et al. 2013; Hui, Molden, and Finkel 2013).
Relationship Quality
Relationship quality is a person’s subjective perception that their relationship is relatively good versus bad (i.e., a person’s evaluation of their relationship; Joel et al. 2020). We operationalized relationship quality as a composite measure (an average of scale z-scores) of three previously validated scales measured at each wave; the composite was then standardized across the entire person-period dataset prior to analysis. Each component scale was highly reliable at each timepoint (all α ≥ 0.87), and the reliability of the composite scale was α = 0.80 across the entire person-period dataset. Partners’ relationship quality scores correlated significantly with one another at each timepoint (all r > 0.60, p < .001).
The first component of the composite measure was a 32-item Couples Satisfaction Index (Funk and Rogge 2007), which included items such as “My relationship with my partner makes me happy” and “I cannot imagine another person making me as happy as my partner does” (0 = not at all true, 5 = completely true). We included this measure because relationship satisfaction is the most commonly used dependent measure in relationship science to assess relationship quality (Joel et al. 2020). The second component was a 10-item Conflict Tactics scale (Straus et al. 1996), which included items such as “(Within the last three months) I shouted or yelled at my partner” (reverse-scored; 0 = this never happened, 7 = happened more than 20 times). The third component was a 4-item High-Maintenance Interactions scale (Finkel et al. 2006), which included items such as “(Over the past month) Interactions with my partner generally went smoothly” (1 = strongly disagree, 7 = strongly agree). We included these latter two scales because they assess relationship quality on a more specific level via perceived daily functioning. Both measures have been previously used as indicators of the relationship quality construct (Finkenauer et al. 2010; Kelmer et al. 2013).
Financial Harmony
We operationalized financial harmony as a composite measure of one scale and five items we developed; these six components were measured at each wave of data collection. The first component was a 10-item Financial Harmony scale (Rick, Small, and Finkel 2011), which was designed to measure the extent to which money is a source of conflict in marriage. The scale includes items such as “When it comes to our finances, my partner and I see eye to eye,” “My partner is satisfied with my attitudes toward money,” and “Money is a constant source of conflict with my partner” (reverse-scored; 1 = strongly disagree, 7 = strongly agree). This scale was highly reliable at each timepoint (all α ≥ 0.88). The five items we developed focused on participants’ satisfaction with how they and their partner manage money (e.g., “Are you happy with the amount of money that you and your partner together are routinely spending?” where 1 = very unhappy and 7 = very happy). The reliability of the composite scale was α = 0.80 across the entire person-period dataset; the composite was standardized across the entire person-period dataset prior to analysis. Partners’ financial harmony scores correlated significantly with one another at each timepoint (all r > 0.45, p < .001).
Methodological Challenges
There are different ways of examining the effect of bank account structure on relationship quality, none of them perfect. We randomly assigned romantic couples with separate bank accounts to different banking conditions and assessed change in relationship quality over time. The “over time” element is key—we could have conducted a lab experiment with couples, but we do not perceive a strong theoretical reason to predict that temporarily activating the experience or representation of a joint account structure would benefit the relationship. Bank account structure is likely to influence day-to-day relationship dynamics—a gradual, cumulative process with downstream implications for relationship quality. Thus, the definitive causal test of the influence of bank account structure calls for random assignment and a longitudinal design, as couples begin building (and continue to build) their marriage together. We considered several possible experimental approaches, such as recruiting couples who already have joint bank accounts and randomly assigning some to separate their money into individual accounts. However, given the complexities involved in de-coupling and deciding how to split shared money, we anticipated that such a condition would pose a particularly extreme risk of attrition.
We ultimately enrolled a sample of 230 eligible engaged and newlywed couples, expecting some attrition. Indeed, attrition is endemic to this sort of research (Bradbury and Karney 2010). Across 83 longitudinal studies on marriage, an average of 31% (median: 29%) of initial samples were lost to attrition (Karney and Bradbury 1995, table 2); this is consistent with the 27% attrition rate observed in a more recent 2-year experiment by Cordova et al. (2014). Collapsing across our three conditions, we observed an attrition rate of 20% (45/230 couples). We anticipated the greatest attrition in the Joint Condition because those couples were required to exert the most effort to comply with instructions (i.e., making a big change is naturally harder than maintaining the status quo; Levitt 2021); all couples began the study with separate accounts, so couples in the Joint Condition were the only ones required to deviate from the status quo. Asking a couple to go to the bank and completely change how they structure their household finances is no small request.
GROWTH CURVE MODEL PREDICTING RELATIONSHIP QUALITY (STUDY 1)
| Model | df | Unstandardized parameter (SE) | t | p |
|---|---|---|---|---|
| Intercept | 107 | 0.288 (0.193) | 1.49 | .140 |
| Separate Condition (vs. Joint) | 108 | –0.539 (0.262) | –2.06 | .042 |
| No-Intervention Condition (vs. Joint) | 106 | –0.624 (0.231) | –2.71 | .008 |
| Time | 115 | 0.010 (0.007) | 1.42 | .159 |
| Separate Condition × Time | 113 | –0.023 (0.009) | –2.44 | .016 |
| No-Intervention Condition × Time | 112 | –0.031 (0.008) | –3.80 | <.001 |
| Model | df | Unstandardized parameter (SE) | t | p |
|---|---|---|---|---|
| Intercept | 107 | 0.288 (0.193) | 1.49 | .140 |
| Separate Condition (vs. Joint) | 108 | –0.539 (0.262) | –2.06 | .042 |
| No-Intervention Condition (vs. Joint) | 106 | –0.624 (0.231) | –2.71 | .008 |
| Time | 115 | 0.010 (0.007) | 1.42 | .159 |
| Separate Condition × Time | 113 | –0.023 (0.009) | –2.44 | .016 |
| No-Intervention Condition × Time | 112 | –0.031 (0.008) | –3.80 | <.001 |
Note.— Relationship quality scores were standardized prior to analysis.
GROWTH CURVE MODEL PREDICTING RELATIONSHIP QUALITY (STUDY 1)
| Model | df | Unstandardized parameter (SE) | t | p |
|---|---|---|---|---|
| Intercept | 107 | 0.288 (0.193) | 1.49 | .140 |
| Separate Condition (vs. Joint) | 108 | –0.539 (0.262) | –2.06 | .042 |
| No-Intervention Condition (vs. Joint) | 106 | –0.624 (0.231) | –2.71 | .008 |
| Time | 115 | 0.010 (0.007) | 1.42 | .159 |
| Separate Condition × Time | 113 | –0.023 (0.009) | –2.44 | .016 |
| No-Intervention Condition × Time | 112 | –0.031 (0.008) | –3.80 | <.001 |
| Model | df | Unstandardized parameter (SE) | t | p |
|---|---|---|---|---|
| Intercept | 107 | 0.288 (0.193) | 1.49 | .140 |
| Separate Condition (vs. Joint) | 108 | –0.539 (0.262) | –2.06 | .042 |
| No-Intervention Condition (vs. Joint) | 106 | –0.624 (0.231) | –2.71 | .008 |
| Time | 115 | 0.010 (0.007) | 1.42 | .159 |
| Separate Condition × Time | 113 | –0.023 (0.009) | –2.44 | .016 |
| No-Intervention Condition × Time | 112 | –0.031 (0.008) | –3.80 | <.001 |
Note.— Relationship quality scores were standardized prior to analysis.
To address this potential asymmetry while still producing approximately equal numbers of participants across the three conditions, we randomized more participants to the Joint Condition. Specifically, for every seven couples, we randomly assigned three to the Joint Condition (total N = 96 couples), two to the Separate Condition (total N = 66 couples), and two to the No-Intervention Condition (total N = 68 couples). In accord with expectations, 34% of couples in the Joint Condition, 12% in the Separate Condition, and 6% in the No-Intervention Condition failed to complete any follow-up surveys after the intake survey.1 Such selective attrition presents an interpretational challenge. We designed our intake survey to help us understand how the results are potentially biased by attrition (e.g., whether couples who drop out differ in observable ways from couples who stick with the study). For example, we included our primary outcome variables—relationship quality and financial harmony—at intake and found that they did not differ among couples who did versus did not comply with experimental instructions. We revisit this important issue later, when discussing robustness checks.
Results
Data Analytic Plan
We pre-registered key elements of our data-analytic plan (e.g., how we would compute composite measures and code compliance with experimental instructions) on OSF (https://osf.io/egp9m), where we also share complete model specification, example syntax, and ancillary analyses (https://osf.io/2m7sv).2 The observations in this experiment are nested, with two partners per couple and up to six timepoints per partner. Thus, we examined the effect of bank account structure on relationship quality with dyadic growth curve modeling (i.e., via distinguishable dyads by partner gender; Bolger and Laurenceau 2013; Kenny, Kashy, and Cook 2006); we used the Satterthwaite (1946) approximation to compute degrees of freedom for tests of the fixed effects. We then examined the potential mediating role of financial harmony with multilevel conditional process modeling (Bauer, Preacher, and Gil 2006; Hayes and Rockwood 2020); we used Kenward and Roger’s (1997) method of computing degrees of freedom for tests of the fixed effects. Again, complete model specification details are shared on our OSF page.
At each timepoint, couple members reported whether they currently maintained completely separate accounts, partially separate/partially merged (“mixed”) accounts, or completely merged accounts. (To encourage honest responding, we told participants in the Joint and Separate Conditions that they would be paid regardless of whether they had been able to comply with our account usage instructions.) We assessed compliance using two sets of pre-registered compliance criteria: strict criteria and liberal criteria (see table 1 for example patterns of participation and their associated compliance codes). The results below come from couples who strictly complied with experimental instructions. Strict compliance in the Separate Condition meant that both partners indicated maintaining completely separate accounts throughout all completed follow-up surveys; strict compliance in the Joint Condition meant that both partners indicated completely merged accounts (at least by their final completed follow-up survey). In addition to following these strict compliance criteria, we replicated key results following liberal compliance criteria (i.e., including couples who partially or temporarily followed instructions, like Joint Condition couples who reported moving to mixed accounts), which yielded virtually identical results (table 3).3 Note that the liberal analyses reported in the Robustness Checks section included couples who at least complied liberally; these analyses included both strictly and liberally compliant couples. In the No-Intervention Condition, there were no bank account instructions with which couples needed to comply, so if they simply completed some or all follow-up surveys, they were coded as strictly compliant (and included in both sets of liberal and strict analyses). Complete compliance details are provided in our pre-registration document.
SOME HYPOTHETICAL EXAMPLES OF PARTICIPATION PATTERNS AND THEIR ASSOCIATED COMPLIANCE CODES (STUDY 1)
| Assigned condition | Follow-up surveys completed by at least one partner | Self-reported account structure in their final follow-up survey | Compliance code |
|---|---|---|---|
| Joint | Some | Completely merged | Strictly complied |
| Joint | All | Completely merged | Strictly complied |
| Joint | Some | Partially merged | Liberally complied |
| Joint | All | Partially merged | Liberally complied |
| Joint | Some | Completely separate | Non-compliant |
| Joint | All | Completely separate | Non-compliant |
| Joint | None | N/A | Non-compliant |
| Separate | Some | Completely separate | Strictly complied |
| Separate | Some | Completely merged | Non-compliant |
| No-Intervention | Some | Completely separate | Strictly complied |
| No-Intervention | Some | Completely merged | Strictly complied |
| Assigned condition | Follow-up surveys completed by at least one partner | Self-reported account structure in their final follow-up survey | Compliance code |
|---|---|---|---|
| Joint | Some | Completely merged | Strictly complied |
| Joint | All | Completely merged | Strictly complied |
| Joint | Some | Partially merged | Liberally complied |
| Joint | All | Partially merged | Liberally complied |
| Joint | Some | Completely separate | Non-compliant |
| Joint | All | Completely separate | Non-compliant |
| Joint | None | N/A | Non-compliant |
| Separate | Some | Completely separate | Strictly complied |
| Separate | Some | Completely merged | Non-compliant |
| No-Intervention | Some | Completely separate | Strictly complied |
| No-Intervention | Some | Completely merged | Strictly complied |
SOME HYPOTHETICAL EXAMPLES OF PARTICIPATION PATTERNS AND THEIR ASSOCIATED COMPLIANCE CODES (STUDY 1)
| Assigned condition | Follow-up surveys completed by at least one partner | Self-reported account structure in their final follow-up survey | Compliance code |
|---|---|---|---|
| Joint | Some | Completely merged | Strictly complied |
| Joint | All | Completely merged | Strictly complied |
| Joint | Some | Partially merged | Liberally complied |
| Joint | All | Partially merged | Liberally complied |
| Joint | Some | Completely separate | Non-compliant |
| Joint | All | Completely separate | Non-compliant |
| Joint | None | N/A | Non-compliant |
| Separate | Some | Completely separate | Strictly complied |
| Separate | Some | Completely merged | Non-compliant |
| No-Intervention | Some | Completely separate | Strictly complied |
| No-Intervention | Some | Completely merged | Strictly complied |
| Assigned condition | Follow-up surveys completed by at least one partner | Self-reported account structure in their final follow-up survey | Compliance code |
|---|---|---|---|
| Joint | Some | Completely merged | Strictly complied |
| Joint | All | Completely merged | Strictly complied |
| Joint | Some | Partially merged | Liberally complied |
| Joint | All | Partially merged | Liberally complied |
| Joint | Some | Completely separate | Non-compliant |
| Joint | All | Completely separate | Non-compliant |
| Joint | None | N/A | Non-compliant |
| Separate | Some | Completely separate | Strictly complied |
| Separate | Some | Completely merged | Non-compliant |
| No-Intervention | Some | Completely separate | Strictly complied |
| No-Intervention | Some | Completely merged | Strictly complied |
ROBUSTNESS CHECKS ADDRESSED IN THE WEB APPENDIX (STUDY 1)
| Question about robustness | Response |
|---|---|
| Do we observe substantively similar results when using different (non-pre-registered) measures of relationship quality (e.g., individual components of our relationship quality composite)? | Yes, though the pattern is stronger when using the pre-registered relationship quality composite (web appendix table E6). |
| Do we observe substantively similar results when using different (non-pre-registered) measures of financial harmony (e.g., individual components of our financial harmony composite)? | Yes, though the pattern is stronger when using the pre-registered financial harmony composite (web appendix table E11). In web appendix D, we examine the factor structure of the financial harmony composite. |
| Do we observe substantively similar results when using relatively liberal versus strict compliance criteria (i.e., if we include couples who partially complied)? | Yes, the results for the growth curve model predicting relationship quality (web appendix table E5) and the multilevel conditional process analysis (web appendix table E10) were similar when using liberal compliance criteria. |
| Is the effect of bank account structure on relationship quality over time nonlinear? | Despite some modest evidence of nonlinearity when following strict compliance criteria (but not liberal compliance criteria), fit indices demonstrate that our linear model presented in table 2 fits the data better than nonlinear models (web appendix table E7). |
| Were the growth curve model results of comparable magnitude for male and female partners? | Yes (web appendix table E5). |
| Were the growth curve model results moderated by household income or differences in partners’ individual incomes? | No (web appendix table E8). |
| Do the growth curve model results hold when controlling for household income, age, race, and education—variables associated with both relationship quality (Bramlett and Mosher 2002; Broman 2005; Karney and Bradbury 1995) and bank account structure (Eickmeyer, Manning, and Brown 2019; Treas 1993; Vogler and Pahl 1993)? | Yes (web appendix table E9). |
| Question about robustness | Response |
|---|---|
| Do we observe substantively similar results when using different (non-pre-registered) measures of relationship quality (e.g., individual components of our relationship quality composite)? | Yes, though the pattern is stronger when using the pre-registered relationship quality composite (web appendix table E6). |
| Do we observe substantively similar results when using different (non-pre-registered) measures of financial harmony (e.g., individual components of our financial harmony composite)? | Yes, though the pattern is stronger when using the pre-registered financial harmony composite (web appendix table E11). In web appendix D, we examine the factor structure of the financial harmony composite. |
| Do we observe substantively similar results when using relatively liberal versus strict compliance criteria (i.e., if we include couples who partially complied)? | Yes, the results for the growth curve model predicting relationship quality (web appendix table E5) and the multilevel conditional process analysis (web appendix table E10) were similar when using liberal compliance criteria. |
| Is the effect of bank account structure on relationship quality over time nonlinear? | Despite some modest evidence of nonlinearity when following strict compliance criteria (but not liberal compliance criteria), fit indices demonstrate that our linear model presented in table 2 fits the data better than nonlinear models (web appendix table E7). |
| Were the growth curve model results of comparable magnitude for male and female partners? | Yes (web appendix table E5). |
| Were the growth curve model results moderated by household income or differences in partners’ individual incomes? | No (web appendix table E8). |
| Do the growth curve model results hold when controlling for household income, age, race, and education—variables associated with both relationship quality (Bramlett and Mosher 2002; Broman 2005; Karney and Bradbury 1995) and bank account structure (Eickmeyer, Manning, and Brown 2019; Treas 1993; Vogler and Pahl 1993)? | Yes (web appendix table E9). |
ROBUSTNESS CHECKS ADDRESSED IN THE WEB APPENDIX (STUDY 1)
| Question about robustness | Response |
|---|---|
| Do we observe substantively similar results when using different (non-pre-registered) measures of relationship quality (e.g., individual components of our relationship quality composite)? | Yes, though the pattern is stronger when using the pre-registered relationship quality composite (web appendix table E6). |
| Do we observe substantively similar results when using different (non-pre-registered) measures of financial harmony (e.g., individual components of our financial harmony composite)? | Yes, though the pattern is stronger when using the pre-registered financial harmony composite (web appendix table E11). In web appendix D, we examine the factor structure of the financial harmony composite. |
| Do we observe substantively similar results when using relatively liberal versus strict compliance criteria (i.e., if we include couples who partially complied)? | Yes, the results for the growth curve model predicting relationship quality (web appendix table E5) and the multilevel conditional process analysis (web appendix table E10) were similar when using liberal compliance criteria. |
| Is the effect of bank account structure on relationship quality over time nonlinear? | Despite some modest evidence of nonlinearity when following strict compliance criteria (but not liberal compliance criteria), fit indices demonstrate that our linear model presented in table 2 fits the data better than nonlinear models (web appendix table E7). |
| Were the growth curve model results of comparable magnitude for male and female partners? | Yes (web appendix table E5). |
| Were the growth curve model results moderated by household income or differences in partners’ individual incomes? | No (web appendix table E8). |
| Do the growth curve model results hold when controlling for household income, age, race, and education—variables associated with both relationship quality (Bramlett and Mosher 2002; Broman 2005; Karney and Bradbury 1995) and bank account structure (Eickmeyer, Manning, and Brown 2019; Treas 1993; Vogler and Pahl 1993)? | Yes (web appendix table E9). |
| Question about robustness | Response |
|---|---|
| Do we observe substantively similar results when using different (non-pre-registered) measures of relationship quality (e.g., individual components of our relationship quality composite)? | Yes, though the pattern is stronger when using the pre-registered relationship quality composite (web appendix table E6). |
| Do we observe substantively similar results when using different (non-pre-registered) measures of financial harmony (e.g., individual components of our financial harmony composite)? | Yes, though the pattern is stronger when using the pre-registered financial harmony composite (web appendix table E11). In web appendix D, we examine the factor structure of the financial harmony composite. |
| Do we observe substantively similar results when using relatively liberal versus strict compliance criteria (i.e., if we include couples who partially complied)? | Yes, the results for the growth curve model predicting relationship quality (web appendix table E5) and the multilevel conditional process analysis (web appendix table E10) were similar when using liberal compliance criteria. |
| Is the effect of bank account structure on relationship quality over time nonlinear? | Despite some modest evidence of nonlinearity when following strict compliance criteria (but not liberal compliance criteria), fit indices demonstrate that our linear model presented in table 2 fits the data better than nonlinear models (web appendix table E7). |
| Were the growth curve model results of comparable magnitude for male and female partners? | Yes (web appendix table E5). |
| Were the growth curve model results moderated by household income or differences in partners’ individual incomes? | No (web appendix table E8). |
| Do the growth curve model results hold when controlling for household income, age, race, and education—variables associated with both relationship quality (Bramlett and Mosher 2002; Broman 2005; Karney and Bradbury 1995) and bank account structure (Eickmeyer, Manning, and Brown 2019; Treas 1993; Vogler and Pahl 1993)? | Yes (web appendix table E9). |
Relationship Quality
Replicating previous research on common marital trajectories (Huston et al. 2001; Johnson et al. 2005; Karney and Bradbury 1995, 1997; VanLaningham, Johnson, and Amato 2001), participants in the No-Intervention Condition exhibited significant reductions in relationship quality over time (B = –0.022, t(60.4) = –5.41, p < 0.001). Note that –0.022 represents the effect of each additional month; in other words, after 24 months, standardized relationship quality scores are expected to decline by 0.53 scale points for these couples. To examine whether merging bank accounts buffers participants against this general decline in relationship quality over time, we created two dummy variables to capture our three-level intervention variable, with the Joint Condition as the reference category. Separate was coded as 1 for couples assigned to the Separate Condition and 0 otherwise; No-Intervention was coded as 1 for couples assigned to the No-Intervention Condition and 0 otherwise. Time was coded in months and centered on month 24 (the end of the experiment). We estimated a dual-intercept model with two levels of nesting (i.e., observations were nested within couples)4 where fixed effects were modeled separately for male and female partners. We allowed for random intercepts and random slopes of Time for each partner. Our results do not vary meaningfully by partners’ gender, and so we report results collapsed across gender (web appendix table E5).
Specifically, the statistical model predicted relationship quality from five predictor terms: (1) Separate Condition, (2) No-Intervention Condition, (3) Time, (4) Separate Condition ×Time, and (5) No-Intervention Condition ×Time. Given our coding scheme, the model’s intercept represents the predicted value of relationship quality at month 24 for couples assigned to the Joint Condition. The Separate Condition “main effect” represents the difference in predicted relationship quality at month 24 between couples in the Joint Condition and the Separate Condition. The No-Intervention Condition “main effect” represents the difference in predicted relationship quality at month 24 between couples in the Joint Condition and the No-Intervention Condition. The Time parameter is the simple slope for couples in the Joint Condition or, more precisely, the effect of the passage of 1-month time on relationship quality for couples in the Joint Condition. The key parameters in the model are the two interaction terms, which represent the difference in simple slopes between couples in the Joint Condition and the Separate Condition (Separate Condition ×Time) and the Joint Condition and the No-Intervention Condition (No-Intervention Condition ×Time). The results of the growth curve model are presented in table 2.
Both condition dummy variables interacted with Time to predict relationship quality (ps ≤ .016), demonstrating that the trajectory for Joint couples was significantly more positive than the trajectory for either Separate or No-Intervention couples (figure 2). Simple slopes analyses revealed that while couples in the Separate Condition (B = –0.013, t(111) = –2.07, p = .041) and the No-Intervention Condition (B = –0.021, t(107) = –4.81, p < .001) exhibited significant declines in relationship quality over time, couples in the Joint Condition (B = 0.010, t(115) = 1.42, p = .159) did not. Again, these beta weights represent the effect of each additional month. After 2 years, couples in the Joint Condition exhibited significantly greater relationship quality than couples in either the Separate Condition (B = –0.539, t(108) = –2.06, p = .042) or the No-Intervention Condition (B = –0.624, t(106) = –2.71, p = .008). There was no significant difference between the Separate and No-Intervention Conditions at the end of the experiment (B = –0.085, t(108) = –0.39, p = .696), or in their declining trajectories (B = –0.008, t(109) = –1.06, p = .293).
THE TRAJECTORY OF RELATIONSHIP QUALITY OVER TIME AS A FUNCTION OF INTERVENTION CONDITION (STUDY 1)
Note.— Relationship quality scores were standardized prior to analysis.
Multilevel Mediation Analyses
Next, we examined whether changes in relationship quality over time were mediated by changes in financial harmony and whether this mediational path depended upon experimental condition. We estimated a “2 × (1–1–1)” mediation model whereby the strength of the indirect effect of time (X) on relationship quality (Y) via financial harmony (M) depends on the bank account dummy variables (W1 and W2). We used the simultaneous estimation strategy whereby M and Y equations are estimated in a single model via selection (or “indicator”) variables (Bauer et al. 2006). Because partner gender did not moderate effects in our primary growth curve model (and to enhance the likelihood of model convergence; Zee and Kumashiro 2019), we collapsed across partner gender when conducting multilevel mediation analyses. Thus, the following model estimates a separate mediation effect for each couple but imposes a common mediation effect for the male partner and female partner within each couple. Complete model results are presented in web appendix table E10.
The results revealed that both dummy variables moderated the a path (ps ≤ .015), providing evidence of moderated multilevel mediation (figure 3). Simple slopes analyses revealed that financial harmony increased substantially among couples in the Joint Condition (p < .001) but remained relatively stable over time for couples assigned to the Separate or No-Intervention Conditions (ps ≥ .278). The average indirect effect (time → financial harmony → relationship quality) was significant for couples in the Joint Condition (B = 0.006, p = .003, Monte Carlo 95% CI: 0.002, 0.011), but not for couples in the Separate Condition (B = 0.001, p = .606, 95% CI: –0.003, 0.005) or in the No-Intervention Condition (B = –0.001, p = .722, 95% CI: –0.003, 0.002). In other words, the Joint Condition prevented the normative decline in relationship quality, in large part, by boosting financial harmony. Indeed, the financial harmony trajectory accounted for about 75% of the relationship quality trajectory in the Joint Condition.5 By contrast, financial harmony remained stable in the Separate and No-Intervention Conditions, and the normative decline in relationship quality was observed in those conditions.
THE EFFECT OF TIME ON RELATIONSHIP QUALITY VIA FINANCIAL HARMONY DEPENDS ON INTERVENTION CONDITION (STUDY 1)
Note.— Residuals and predictions of M and Y by W1 and W2 are not shown in the diagram but were included in the model. Complete model results are presented in web appendix table E10. †p ≤ .10, *p ≤ .05, **p ≤ .01, ***p ≤ .001.
Lastly, we want to draw attention to moderation of the c′ path in the model (figure 3). Relative to couples in the Joint Condition, couples in the Separate Condition are marginally declining in relationship quality over time (B = –0.017, t(111) = –1.87, p = .064) and couples in the No-Intervention Condition are significantly declining in relationship quality over time (B = –0.022, t(110) = –2.85, p = .005). These two slope parameters are residual effects of the intervention after accounting for potential changes in financial harmony over time. In other words, the effect of our intervention on the slope of relationship quality emerges independent of the effect of the intervention on the slope of financial harmony. Thus, the remaining between-group differences in slopes are unexplained by changes in financial harmony, likely due to the omission of one or more other mediators (Zhao, Lynch, and Chen 2010). As noted earlier, the effect of bank account structure on relationship quality is (presumably) multiply determined. We explore other potential differences between couples with joint versus separate bank accounts in study 2, as well as return to this point in the General Discussion.
Robustness Checks
We next conducted a series of robustness checks to examine how sensitive our results are to different data-analytic decisions (see web appendix E for complete details). As mentioned earlier, and consistent with expectations, attrition rates were highest in the Joint Condition. This attrition precludes the use of standard intention-to-treat analyses. However, because we have full data at study entry, we can gain some insight into whether our results are driven by differences between couples who complied with experimental instructions and couples who did not. Thus, we assessed whether couples who complied with our instructions in the Joint Condition differed from (1) Joint non-compliers (i.e., those who did not follow instructions or were lost to attrition), (2) Separate compliers, (3) Separate non-compliers, and (4) No-Intervention compliers. Specifically, we compared relationship quality and financial harmony at study entry across these groups. The results revealed no significant differences between Joint compliers and Joint non-compliers (ps ≥ .657), Joint compliers and Separate compliers (ps ≥ .617), Joint compliers and Separate non-compliers (ps ≥ .262), or Joint compliers and No-Intervention compliers (ps ≥ .907). These results, following both strict and liberal compliance criteria, are presented in web appendix table E1. Note that there were only four couples lost to attrition in the No-Intervention Condition (i.e., they completed no follow-up surveys). We did not include this sixth group of “No-Intervention non-compliers” in these analyses due to concerns about unequal sample sizes and unequal variances increasing the type I error rate across comparisons (Tabachnick and Fidell 2013).
To gain a deeper understanding of the nature of attrition (e.g., whether Joint compliers differed in other, observable ways from the other groups), we pre-registered a plan to expand web appendix table E1 (https://osf.io/98fsk). Specifically, we conducted scale-level analyses of all 73 variables that were measured at intake (e.g., personality scales, household financial practices, demographic variables). Following strict compliance criteria, only 4/73 (5.5%) total tests featured a significant omnibus test across the five groups coupled with a significant contrast between Joint compliers and Joint non-compliers. A binomial test revealed that 4/73 was no different than the probability of finding at least the same number of significant effects by chance alone (i.e., 5.5% vs. 5%; p = .198). These supplementary tables, following both strict and liberal compliance criteria, are presented in web appendix table E2. We conducted a similar set of analyses on partner-level differences at intake, which yielded similar conclusions (i.e., the extent to which partners differed in income, power, personality, etc., did not systematically vary across compliance groups; web appendix table E3).
To complement our scale-level analyses of predictors of non-compliance, we provide another robustness check assessing potential differences in relationship quality at intake. Specifically, we reran the growth curve model featured in table 2 with Time centered on month 0 (instead of month 24). This alternative coding allows us to examine whether relationship quality differed at intake between (1) couples in the Joint Condition and Separate Condition and (2) couples in the Joint Condition and No-Intervention Condition, while controlling for potential differences in trajectories. Consistent with earlier results, we find no differences in relationship quality at study entry when following strict compliance criteria (ps ≥ .555). Complete results are presented in web appendix table E4.
MEANS, STANDARD DEVIATIONS, AND OMNIBUS TEST RESULTS (STUDY 2)
| Dependent variable | Bank account structure | Omnibus test | ||||
|---|---|---|---|---|---|---|
| Completely separate (N = 82) | Partially merged (N = 162) | Completely merged (N = 263) | F | P | ηp2 | |
| Financial harmony | 4.89 (1.22) | 5.01 (1.21) | 5.31 (1.15) | 5.41 | .005 | 0.021 |
| Communal norm adherencea | 1.48 (1.43) | 1.62 (1.43) | 1.91 (1.15) | 4.69 | .010 | 0.018 |
| Exchange norm adherencea | –0.89 (1.84) | –1.17 (1.85) | –1.35 (1.83) | 2.06 | .129 | 0.008 |
| Financial communality | 2.66 (3.43) | 4.62 (3.67) | 8.05 (2.86) | 111.44 | <.001 | 0.307 |
| Financial goal alignment (lower mean = more aligned) | 1.39 (1.29) | 1.23 (1.31) | 0.79 (0.94) | 12.36 | <.001 | 0.047 |
| Financial transparency | 5.76 (1.27) | 5.93 (1.35) | 6.33 (1.04) | 10.18 | <.001 | 0.039 |
| Financial conversation frequency | 4.13 (1.40) | 4.26 (1.19) | 4.37 (1.24) | 1.24 | .290 | 0.005 |
| Dependent variable | Bank account structure | Omnibus test | ||||
|---|---|---|---|---|---|---|
| Completely separate (N = 82) | Partially merged (N = 162) | Completely merged (N = 263) | F | P | ηp2 | |
| Financial harmony | 4.89 (1.22) | 5.01 (1.21) | 5.31 (1.15) | 5.41 | .005 | 0.021 |
| Communal norm adherencea | 1.48 (1.43) | 1.62 (1.43) | 1.91 (1.15) | 4.69 | .010 | 0.018 |
| Exchange norm adherencea | –0.89 (1.84) | –1.17 (1.85) | –1.35 (1.83) | 2.06 | .129 | 0.008 |
| Financial communality | 2.66 (3.43) | 4.62 (3.67) | 8.05 (2.86) | 111.44 | <.001 | 0.307 |
| Financial goal alignment (lower mean = more aligned) | 1.39 (1.29) | 1.23 (1.31) | 0.79 (0.94) | 12.36 | <.001 | 0.047 |
| Financial transparency | 5.76 (1.27) | 5.93 (1.35) | 6.33 (1.04) | 10.18 | <.001 | 0.039 |
| Financial conversation frequency | 4.13 (1.40) | 4.26 (1.19) | 4.37 (1.24) | 1.24 | .290 | 0.005 |
Note.— The results from a series of one-way ANOVA models on all dependent variables.
To examine whether patterns of communal and exchange norm adherence varied across bank account structures, our primary analysis was a mixed ANOVA. Those results are presented in the main text. Here, we present results of a one-way ANOVA on each norm separately.
MEANS, STANDARD DEVIATIONS, AND OMNIBUS TEST RESULTS (STUDY 2)
| Dependent variable | Bank account structure | Omnibus test | ||||
|---|---|---|---|---|---|---|
| Completely separate (N = 82) | Partially merged (N = 162) | Completely merged (N = 263) | F | P | ηp2 | |
| Financial harmony | 4.89 (1.22) | 5.01 (1.21) | 5.31 (1.15) | 5.41 | .005 | 0.021 |
| Communal norm adherencea | 1.48 (1.43) | 1.62 (1.43) | 1.91 (1.15) | 4.69 | .010 | 0.018 |
| Exchange norm adherencea | –0.89 (1.84) | –1.17 (1.85) | –1.35 (1.83) | 2.06 | .129 | 0.008 |
| Financial communality | 2.66 (3.43) | 4.62 (3.67) | 8.05 (2.86) | 111.44 | <.001 | 0.307 |
| Financial goal alignment (lower mean = more aligned) | 1.39 (1.29) | 1.23 (1.31) | 0.79 (0.94) | 12.36 | <.001 | 0.047 |
| Financial transparency | 5.76 (1.27) | 5.93 (1.35) | 6.33 (1.04) | 10.18 | <.001 | 0.039 |
| Financial conversation frequency | 4.13 (1.40) | 4.26 (1.19) | 4.37 (1.24) | 1.24 | .290 | 0.005 |
| Dependent variable | Bank account structure | Omnibus test | ||||
|---|---|---|---|---|---|---|
| Completely separate (N = 82) | Partially merged (N = 162) | Completely merged (N = 263) | F | P | ηp2 | |
| Financial harmony | 4.89 (1.22) | 5.01 (1.21) | 5.31 (1.15) | 5.41 | .005 | 0.021 |
| Communal norm adherencea | 1.48 (1.43) | 1.62 (1.43) | 1.91 (1.15) | 4.69 | .010 | 0.018 |
| Exchange norm adherencea | –0.89 (1.84) | –1.17 (1.85) | –1.35 (1.83) | 2.06 | .129 | 0.008 |
| Financial communality | 2.66 (3.43) | 4.62 (3.67) | 8.05 (2.86) | 111.44 | <.001 | 0.307 |
| Financial goal alignment (lower mean = more aligned) | 1.39 (1.29) | 1.23 (1.31) | 0.79 (0.94) | 12.36 | <.001 | 0.047 |
| Financial transparency | 5.76 (1.27) | 5.93 (1.35) | 6.33 (1.04) | 10.18 | <.001 | 0.039 |
| Financial conversation frequency | 4.13 (1.40) | 4.26 (1.19) | 4.37 (1.24) | 1.24 | .290 | 0.005 |
Note.— The results from a series of one-way ANOVA models on all dependent variables.
To examine whether patterns of communal and exchange norm adherence varied across bank account structures, our primary analysis was a mixed ANOVA. Those results are presented in the main text. Here, we present results of a one-way ANOVA on each norm separately.
A reader might raise a number of other questions about the robustness of our results. Table 3 summarizes some potential questions, offers brief responses, and indicates which sections of the web appendix offer more comprehensive answers.
Couples Randomly Assigned to the No-Intervention Condition
To explore how the current results align with existing nonexperimental literature, we conducted an auxiliary set of analyses examining the association between self-selected banking structure and relationship quality among the 64 couples in the No-Intervention Condition. Namely, we assessed whether couples who self-selected into a joint account at any point in the 2-year study (coded as 1; 28% of couples) experienced greater relationship quality at the end of the study than couples who self-selected into maintaining separate accounts (coded as 0). The pattern appears to support this possibility (B = 0.491, t(53.7) = 1.72, p = .092). However, this marginal difference was of comparable magnitude at the beginning of the study and did not vary over time. Specifically, simple slopes analyses revealed that couples who self-selected into a joint account (B = –0.019, t(58.1) = –2.42, p = .019) or separate accounts (B = –0.023, t(60.2) = –4.67, p < .001) exhibited similar declines in relationship quality over time (interaction: B = 0.004, t(58.7) = 0.46, p = .647).
At first glance, these correlational results may seem contradictory to our experimental results. However, it is important to note the timing differences between couples in the No-Intervention Condition who self-selected into a joint account and couples in the Joint Condition who were assigned to open a joint account. Namely, in addition to any effects of selection, couples in the No-Intervention Condition could have adopted a joint account(s) at any point between months 0 and 24. Indeed, No-Intervention couples who transitioned from having separate accounts to joint accounts did so throughout the 2-year window, with the largest increase in joint account usage observed between months 12 and 24. In contrast, couples in the Joint Condition were instructed to adopt a joint account(s) shortly after completing the intake survey at month 0. Thus, we cannot directly compare couples who eventually chose to open a joint account with those who were immediately assigned to open a joint account.
Discussion
Study 1 provides causal evidence that the adoption of a joint account helps preserve relationship quality across the connubial crucible. While couples assigned to maintain separate accounts or follow a banking arrangement of their choosing experienced the typical decline in relationship quality over time, couples assigned to merge their money in a joint account were buffered against decline. These results do not appear to be driven by differences in compliance or attrition across conditions.
Some readers may be surprised that couples in the Separate Condition fared as well as they did (i.e., their relationship quality declined at about the same rate as couples in the No-Intervention Condition). After all, if separate accounts have several features that are bad for relationships (e.g., making it easier to conceal difficult-to-justify purchases, discouraging communal norm adherence, discouraging financial goal alignment), would we not expect couples in the Separate Condition to clearly be worse off at the end of the experiment? We did not find this pattern particularly surprising, for two reasons. First, note that 72% of No-Intervention couples maintained separate accounts throughout the entire experiment, and most of the No-Intervention couples who self-selected into a joint account did so in the second year of the experiment. Thus, separate account usage was quite widespread in the No-Intervention Condition. Second, Separate Condition couples may have benefited from the experiment taking a potentially difficult decision (i.e., whether to change their existing account structure) off the table. No-Intervention couples always had that decision looming in the background (at least until an account structure change was made). The relatively small number of No-Intervention couples who switched to joint accounts may have experienced some difficulty in getting to that point (e.g., someone bringing up a difficult topic that their partner did not want to get into).
Another key finding from study 1 is that the positive effect of merging finances was driven, at least in part, by positive changes in financial harmony over time (i.e., greater satisfaction with how partners handle and discuss money). Financial harmony is only one potential mechanism underlying the effect of bank account structure on relationship quality. Thus, in addition to financial harmony, we examined other possible mechanisms in study 2 (e.g., communal norm adherence, financial goal alignment). Specifically, we surveyed a separate sample of married individuals to identify correlates of bank account structure within longer-term unions. Study 2 also builds upon study 1 by exploring the potential effects of partially merged accounts (i.e., mixed accounts) relative to completely joint or completely separate accounts. Although this study does not afford clear causal conclusions, any insights we observe provide greater depth in understanding other potential mechanisms.
STUDY 2: A DEEPER LOOK AT WHY JOINT BANK ACCOUNTS ARE GOOD FOR COUPLES
Participants and Procedure
We recruited 507 married individuals (48% female) from Prolific who currently lived in the United States and identified English as their first and primary language. Participants had been married an average of 15 years (range: less than 1 month to 55 years) and 70% had children. Participants’ median age was 40 (range: 21–78), and the median household income was $84,000 (range: $2,000–$1,300,000); 68% had earned at least a bachelor’s degree, and 85% identified as White or Caucasian. Complete descriptive statistics are posted on OSF.
Participants completed four blocks of questions in a randomized order: (1) banking questions, (2) financial harmony, (3) communal and exchange norm adherence (counterbalanced order), and (4) financial goal alignment. Means, standard deviations, and correlations among key variables within these four blocks are presented in web appendix F. The study concluded with demographic measures (e.g., number of children, age, race, education). Complete measures are provided on OSF. No data exclusions were made; we analyzed all data that were available.
Banking Questions
Within the banking block, participants reported their bank account structure via the same item used to code compliance in study 1. Namely, they selected one of three options that best described how they and their spouse currently organize their money (1 = We have not merged our cash finances; we have no joint checking or savings accounts, 2 = We have partially merged our cash finances; we use at least one joint checking or savings account, but at least one of us uses our own personal checking and/or savings account, 3 = We have completely merged our cash finances; any checking and savings accounts we use are joint accounts). Participants who reported having partially merged or completely merged finances were asked to recall when they opened a joint account with their spouse (e.g., before they were married, within the first 6 months of being married, etc.).
Financial Harmony
In another block, we administered our 15-item index of financial harmony from study 1 (α = 0.93). This index included the 10-item Financial Harmony scale (1 = strongly disagree, 7 = strongly agree; Rick et al. 2011) and the five items we developed (e.g., “Are you happy with the amount of money that you and your partner together are routinely spending?” where 1 = very unhappy and 7 = very happy).
Communal and Exchange Norm Adherence
In another block, we asked two sets of questions to capture communality within the marriage: (1) communal and exchange norm adherence and (2) financial communality. First, we asked participants to consider two “types” of marriage that vary in how partners respond to each other’s needs and desires. These two types reflected communal and exchange prototypes (descriptions were adapted from Clark et al. 2010). The communal prompt included statements like “when one person does something for the other, the other should not owe the giver anything.” The exchange prompt included statements like “members of the relationship ought to keep track of benefits given and received in order to keep them in balance.” After reading each prompt, participants indicated their level of agreement with two statements (–3 = strongly disagree, 3 = strongly agree): “Over the past two months, this is the way that I have been operating in my relationship with my spouse” and “Over the past two months, this is the way that my spouse has been operating in their relationship with me.” Responses to these self/partner items were significantly correlated for both communal norm adherence (r(505) = 0.72, p < .001) and exchange norm adherence (r(505) = 0.84, p < .001). Thus, we averaged self/partner responses for each norm; results and conclusions do not change if we analyze the self/partner items separately.
At the end of this same block, we asked participants to imagine that their spouse had received a $1,000 bonus at work and to indicate whose money it was. Responses were on an 11-point scale where 0 = I consider that money to belong entirely to my spouse (“my spouse’s money”) and 10 = I consider that money to be as much mine as it is my spouse’s (“our money”). Higher scores reflected stronger financial communality within the marriage.
Financial Goal Alignment, Transparency, and Conversation Frequency
In another block, we asked participants to think about three different financial goals, which we defined as “desirable outcomes to strive for when managing your money.” Specifically, participants rated how important (1) saving for retirement, (2) saving for special circumstances (e.g., vacation, emergencies), and (3) paying down debt was for themselves and (in a separate item) their spouse (1 = not at all important, 7 = very important). Thus, for each goal, participants made two judgments. We selected these specific goals because a survey among 1,000 Americans identified these as the top 3 financial priorities (Brown 2021). We calculated financial goal alignment by taking the absolute value of the difference between the three self-ratings and their corresponding spouse ratings. We then averaged the three difference scores to create an index (α = 0.75). While a value of 0 reflects perfect alignment between spouses, a value of 6 reflects perfect misalignment.
Within this same block, we also included two exploratory measures about financial conversations, which we defined as “discussions, debates, and/or planned sessions about money”: (1) financial transparency and (2) financial conversation frequency. Financial transparency was measured with a single item whereby participants evaluated how open and honest they and their spouse are with each other about money (1 = not at all transparent, 7 = very transparent). Financial conversation frequency was measured with a single item whereby participants estimated how frequently they have financial conversations with their spouse (1 = almost never, 7 = every day).
Results
Bank Account Structure
Overall, 16% of married participants reported completely separate accounts, 32% reported partially merged accounts, and 52% reported completely merged accounts. These percentages are consistent with prior research on bank account structure within marriages in Western nations (in prior studies, 10–15% of couples report completely separate accounts and 52–65% report completely joint accounts; Addo and Sassler 2010; Gladstone et al. 2022; van Raaij et al. 2020; Vogler et al. 2006). Among participants who reported having partially or completely merged accounts, the majority recalled opening a joint account by the end of their first year of marriage.
Financial Harmony
We conducted a one-way analysis of variance (ANOVA) of Bank Account Structure (Completely Separate, Partially Merged, Completely Merged) on financial harmony. [For this analysis and all subsequent analyses, we focus on key contrasts between couples with completely merged accounts and those with (1) partially merged accounts or (2) completely separate accounts.] The results revealed a significant omnibus effect (see table 4 for all means and omnibus test results). We find that participants with completely merged accounts reported significantly greater financial harmony than participants with partially merged accounts (t(504) = 2.52, p = .012, d = 0.25) or completely separate accounts (t(504) = 2.78, p = .006, d = 0.36). The latter two groups did not significantly differ (t(504) = 0.74, p = .462). These correlational results are consistent with the experimental results presented in study 1—joint account usage (relative to separate account usage) predicts greater financial harmony.6
Communal and Exchange Norm Adherence
As noted previously, we measured communality within the marriage with two sets of questions: (1) communal and exchange norm adherence and (2) financial communality. We examined each set of questions separately. Again, table 4 reports all means and omnibus test results.
First, to examine whether different banking arrangements are associated with different patterns of norm adherence, we conducted a 3 (Bank Account Structure: Completely Separate, Partially Merged, Completely Merged) × 2 (Norm: Communal, Exchange) mixed ANOVA with Norm being within-subjects. The results revealed a significant main effect of Norm (F(1, 504) = 485.22, p < .001, ηp2 = 0.49). Consistent with prior research (Clark et al. 2010), this sample of married participants expressed significantly stronger adherence to communal versus exchange norms (M = 1.67, SE = 0.06 vs. M = –1.14, SE = 0.09). Importantly, this main effect was qualified by a significant interaction (F(2, 504) = 4.38, p = .013, ηp2 = 0.02). Supporting our theorizing, communal norm adherence was significantly stronger among participants with completely merged accounts than those with partially merged accounts (t(504) = 2.27, p = .023, d = 0.23) or completely separate accounts (t(504) = 2.64, p = .009, d = 0.35). The latter two groups did not significantly differ (t(504) = 0.79, p = .430). We also note that exchange norm adherence was significantly weaker among participants with completely merged versus completely separate accounts (t(504) = 1.98, p = .048, d = 0.25). The other two differences were not significant (ps ≥ .268).
Second, we conducted a one-way ANOVA of Bank Account Structure on the degree of financial communality (i.e., the work bonus). The results revealed a significant omnibus effect. Participants with completely merged accounts reported that their spouse’s work bonus was “our money” to a significantly stronger extent than participants with partially merged accounts (t(504) = 10.64, p < .001, d = 1.08) who, in turn, were significantly more communal than participants with completely separate accounts (t(504) = 4.47, p < .001, d = 0.55). This pattern is certainly to be expected, but we felt that it was useful to confirm that new, incoming money (such as one partner’s work bonus) is viewed in a way that is consistent with the couple’s account structure.
Financial Goal Alignment, Transparency, and Conversation Frequency
A one-way ANOVA of Bank Account Structure revealed significant differences in financial goal alignment (table 4). Namely, participants with completely merged accounts reported significantly greater alignment—values closer to 0—than participants with partially merged accounts (t(503) = 3.86, p < .001, d = 0.40) or completely separate accounts (t(503) = 4.15, p < .001, d = 0.57). The latter two groups did not significantly differ (t(503) = 1.02, p = .307).
Lastly, we conducted one-way ANOVAs on financial transparency and financial conversation frequency. We found significant differences across different banking arrangements in financial transparency, but not the frequency of conversation (table 4). Specifically, participants with completely merged accounts report significantly greater openness and honesty than those with partially merged accounts (t(503) = 3.42, p < .001, d = 0.34) or completely separate accounts (t(503) = 3.84, p < .001, d = 0.52). The latter two groups did not significantly differ (t(503) = 1.06, p = .291).
Ancillary Analyses
We also conducted two sets of analyses involving demographic variables. Like study 1, we explored whether two income variables (i.e., household income and differences in partners’ individual income) moderate the relationships between bank account structure and our dependent variables (web appendix H). Neither income variable significantly moderated the relationship between bank account structure and financial harmony. However, there was some movement for communal norm adherence and financial communality such that differences across couples with different banking structures became larger as household income or income discrepancies increased. Importantly, our central results remain robust even after including these additional income parameters in the models.
In addition to household income, we also examined whether the results hold when controlling for age, race, and education (as in study 1). Namely, we regressed each dependent variable on a “Partially Merged” (vs. Completely Merged) dummy, a “Completely Separate” (vs. Completely Merged) dummy, household income, age, race, and education. Overall, the results are robust and remain consistent—couples with completely merged (vs. partially merged or completely separate) accounts report greater financial harmony (ps ≤ .054), communality (ps ≤ .030), financial goal alignment (ps ≤ .001), and transparency (ps ≤ .001; web appendix I).
Discussion
Study 2 examined the association between bank account structure and important relationship dynamics. Consistent with study 1, we find that financial harmony is highest among couples with completely merged accounts (vs. partially merged or completely separate accounts). Moreover, participants with completely merged accounts versus other arrangements report greater communal norm adherence, financial communality, financial goal alignment, and financial transparency. Critically, the results suggest that partial merging (i.e., using a mix of joint and separate accounts) is not associated with the same benefits as complete merging.
These results provide correlational evidence only among married individuals who self-selected into different arrangements and do not speak to causality (e.g., it is unclear whether joint bank accounts encourage greater communality over time or whether greater communality increases the likelihood of opening and using joint bank accounts). That said, the results do offer new clues into how couples’ bank account structure might influence important relationship dynamics. In particular, the results are consistent with our theoretical reasoning that opening and using a joint bank account should (1) promote financial well-being, (2) help partners align their financial goals and “get on the same page” regarding how they plan to spend and save money, and (3) help to sustain communal norm adherence over time.
GENERAL DISCUSSION
Engaged and newlywed couples face numerous joint consumption decisions, spanning from the relatively mundane (e.g., which show to binge on Netflix, where to have dinner) to the consequential (e.g., where to live, whether one partner should leave the workforce). The current research tackles an important question facing romantic couples as they transition to marriage: which banking arrangement is best and why? Study 1 suggests that couples who are randomly assigned to open and maintain a joint bank account are happier 2 years later than couples who are randomly assigned to continue maintaining separate accounts or to manage their money however they see fit. In particular, couples assigned to the Joint Condition are buffered against the normative decline in relationship quality during the connubial crucible (Huston et al. 2001). Those couples benefited (in part) from improvements in financial harmony over time. Our joint bank account intervention may have also helped establish a stronger communal view of the marriage or, at the very least, minimize deviations from the communal ideal (Clark et al. 2010). Halting declines in communality during the early years of marriage is critical, as (1) communality is linked to greater relationship quality (Clark et al. 2010) and (2) the dynamics established early on are likely to persist for years to come (Huston et al. 2001). Indeed, early interventions that can slow the rate of typical decline should place couples on a better path into the future (Joel and Eastwick 2018). Couples in the Joint Condition—all of whom began the experiment with separate accounts—may have established a stronger “we” perspective of their marriage and money that endures over time. Indeed, the correlational evidence presented in study 2 demonstrates that joint account holders (vs. partially joint or separate) in longer-term marriages are especially likely to adhere to communal norms.
As discussed earlier, we expected and observed the highest rate of attrition in the Joint Condition (study 1), likely because these couples were the only ones required to actively deviate from the status quo of separate accounts. This is an important, albeit probably unavoidable, limitation of this type of research. However, note that couples who did versus did not comply with experimental instructions were qualitatively similar at intake. We observe no systematic differences at intake across complying versus non-complying couples in terms of (1) our primary outcome variables (i.e., relationship quality and financial harmony), (2) the 73 variables measured at intake (beyond what would be expected by chance), or (3) various partner-level differences at intake (e.g., differences in partners’ individual incomes). Of course, there may be important differences between compliers and non-compliers at intake that we did not measure. However, note that our focal effects are interaction terms between bank account structure and time. A confound at intake would not provide a natural explanation for the interaction between bank account structure and time (Lynch 1998). In addition, we find that financial harmony improves over time among Joint compliers (vs. Separate or No-Intervention compliers) and that this change over time contributes to their relationship quality. It is unlikely that these moderated, multilevel indirect effects can be explained by differential attrition.
Regarding the generalizability of our findings, it is important to note that merging money is not risk free or ideal for all couples. Having a joint account reduces each partner’s independence and may well place them at elevated risk for controlling or abusive behavior from their spouse. In this sense, merging money speaks to a broader tension in intimate relationships between self-protection and relationship enhancement. Individuals cannot simultaneously maximize both (1) closeness and intimacy in their relationship and (2) protection from the risk that their partner might hurt or exploit them (Murray, Holmes, and Collins 2006). Deep intimacy requires that partners allow themselves to be vulnerable to each other. Note that among 962 couples deemed eligible in our screening survey (web appendix A), 230 couples (all with separate accounts) consented to participate once they understood that participating might require them to merge money. Agreeing to participate reflects some degree of openness to alternative arrangements. If one or both members of the couple were apprehensive about merging, it may not be the case that pooling brings the benefits we see here. Thus, although merging money appears to be a good strategy for sustaining relationship quality on average, it is unlikely to be the best option for all. That said, our results are bolstered by Gladstone et al.’s (2022) findings, who also conclude that joint accounts are positively associated with relationship quality in large-scale correlational studies that use representative samples rather than samples pre-selected for their openness to alternative arrangements.
Future research could extend beyond the bank account structure itself and examine how couples use joint accounts (e.g., the processes by which money goes into and comes out of a joint account). Are all or most of those decisions jointly discussed? Prior work suggests that, early on in relationships, one partner tends to assume greater financial responsibility than the other partner (Ward and Lynch 2019). This initial division of labor is consequential such that partners who take on more financial responsibility develop greater financial literacy over the course of their relationships, while those who offload this responsibility do not. It is possible that merging money into a joint account might prevent (or at least delay) one partner from fully “taking the lead” on money matters. Taking a step back, whether a joint account structure encourages shared financial responsibility remains an open question.
Study 1 measured relationship quality via self-report, which is the gold standard for assessing relationship quality (Funk and Rogge 2007). Of course, self-reports in this context raise concerns about measurement error due to socially desirable responding, imperfect memory, or motivated reasoning (Arias and Beach 1987). However, self-reported measures such as the Couples Satisfaction Index (used in study 1) do predict behavioral indicators of relationship quality (e.g., observers’ ratings of partners’ negativity toward each other during a discussion; Papp and Witt 2010). Future research could complement our self-reported relationship outcomes by measuring objective relationship outcomes (e.g., visits to couples’ therapy, separation, divorce). Another interesting possibility is exploring whether couples with joint (vs. separate) accounts ultimately save more money and/or have better credit scores.
Although we considered household income and differences in partners’ individual income as potential moderators, future researchers could examine other possibilities by adopting different recruitment criteria. For example, as noted, we intentionally recruited couples with separate bank accounts to keep things manageable (i.e., three conditions, as opposed to a factorial design crossing the intervention with couples’ initial account structure). We also focused on the United States and couples entering their first marriage; examining relationship dynamics among couples from different cultures or entering second marriages (those with potentially more complex financial lives and established routines; Burgoyne and Morison 1997) could provide important new insights.
Engaged and newlywed partners are often on the lookout for guidance on how to navigate the difficult decisions that must be faced when starting a life together. As we saw with the quotes that opened this article, the advice around how to handle household finances is often contradictory. We believe that our results provide the strongest evidence to date that merging money in a joint account helps couples preserve relationship well-being as they traverse the connubial crucible, a dynamic that may endure for years to come.
DATA COLLECTION INFORMATION
The research protocol for study 1 was approved by The University of Michigan’s Institutional Review Board (# HUM00058917); study 2 was approved by Indiana University’s Institutional Review Board (# 1707369558). The first author managed data collection for study 1 under the supervision of the second author. Initial recruitment began in 2013 via social media, the volunteer section of Craigslist, Amazon Mechanical Turk, paid advertising on wedding blogs, word-of-mouth at bridal shows, and mailing letters to customers on a jeweler’s mailing list. The final wave of data collection was completed in 2017. The first author managed data collection for study 2, which was conducted via Prolific in August 2022. The first author conducted all data analyses; the fourth author assisted with model specification and confirmed results’ interpretation in study 1. All materials (e.g., bank account instructions, complete measures from the intake survey and all follow-up surveys, study 1’s pre-registered data analytic plan), model specification, example syntax, and ancillary analyses are stored in a project folder on the Open Science Framework: https://osf.io/2m7sv/.
Footnotes
These percentages differ slightly from the percentages provided on the last page of our pre-registration document. In the pre-registration table, “completed” meant both partners completed the survey. There were five couples that were originally coded as “exited immediately” because only one partner completed only one follow-up survey (month 3). However, as we began learning more about the mechanics of multilevel modeling, we realized that growth curve modeling does not require complete data from each partner at each wave. Thus, these five couples provided usable data at month 3. The flow chart presented in web appendix A is an accurate record of participation.
The data are available from the first author upon request. In the era of open science, we realize that this is not an ideal arrangement. However, we chose not to make the data publicly available due to heightened privacy concerns inherent in nonindependent data (e.g., individuals may access the data to see their partner’s responses, violating confidentiality) and the collection of data involving sensitive topics (e.g., personal finances, relationship satisfaction; Joel, Eastwick, and Finkel 2018). The JCR review team had full access to the data.
We feature results following strict compliance criteria (vs. liberal compliance criteria) because these analyses provide the cleanest test of the effect of bank account structure on relationship quality. We note that statistical power is enhanced by having multiple observations per partner per couple (i.e., our dataset includes up to 12 independent observations per couple). Specifically, our dataset includes 230 × 12 = 2,760 rows (2,049 of which are unique, based on the number of follow-up surveys our participants completed—if all participants had completed all follow-up surveys, then all 2,760 rows would be unique).
Although observations (“level 1”) are nested within partners (“level 2”) who are, in turn, nested within couples (“level 3”), we only have two levels of nesting. Dual-intercept models estimate separate effects for male and female partners. After accounting for one male and one female per each male–female couple, there is no more variability to explain at the partner level (i.e., all couples have one male partner and one female partner; Kenny et al. 2006). Thus, there are only two levels of nesting.
In a multilevel mediation model, the total effect = (aj × bj) + cj′ + covariance (aj, bj). For our model, the total effect is estimated to be about 0.008, or (0.025 × 0.231 ≈ 0.006) + 0.002 + 0.0006. So, 0.006/0.008 = 75%. Thus, most of the total effect of X on Y in the Joint Condition was accounted for by the indirect effect via financial harmony.
Like study 1, we draw substantively similar conclusions when using alternative measures of financial harmony in our analyses (e.g., different components of our financial harmony composite). These ancillary results are presented in web appendix table G.
Author notes
Jenny G. Olson ([email protected]) is an assistant professor of marketing at the Kelley School of Business, Indiana University, 1309 East 10th Street, Bloomington, IN 47405, USA.
Scott I. Rick ([email protected]) is an associate professor of marketing at the Ross School of Business, University of Michigan, 701 Tappan Avenue, Ann Arbor, MI 48109, USA.
Deborah A. Small ([email protected]) is the Adrian C. Israel Professor of Marketing at the Yale School of Management, Yale University, 165 Whitney Avenue, New Haven, CT 06511, USA.
Eli J. Finkel ([email protected]) is a professor of psychology at the Weinberg College of Arts and Sciences, Northwestern University, 1918 Sheridan Road, Evanston, IL 60208, USA and a professor of management and organizations at the Kellogg School of Management, Northwestern University, 2029 Sheridan Road, Evanston, IL 60208, USA.
We thank the review team for their constructive feedback throughout the review process, as well as Deborah Kashy and Katherine Zee for statistical guidance. Financial assistance for this research was provided by the Wharton School of the University of Pennsylvania, University of Michigan, Northwestern University, and the Russell Sage Foundation (# 98.13.01). Supplementary materials are included in the web appendix accompanying the online version of the article.
