Abstract
How does financial education lead to improved financial behavior and higher financial well-being? An influential Consumer Financial Protection Bureau model introduced in 2015 proposes that the goal of financial education is to improve financial well-being and that financial education does so by increasing financial knowledge, which improves financial behavior, which improves financial well-being. In this study, the authors test links in the Consumer Financial Protection Bureau model, examining the differential roles of objective and subjective knowledge. They also test whether an analogous model might capture effects of physical health education on physical health knowledge, behavior, and well-being. They report a quasi-experiment comparing changes in financial and physical health knowledge, behavior, and well-being at two time points in a semester for students enrolled in a personal finance class, a personal health class, or neither. This study reports the first causal estimates of flow from financial education to financial knowledge to financial behaviors to a validated measure of subjective financial well-being. Financial education caused large changes in both objective and subjective knowledge. Yet only subjective knowledge mediated the large effects of financial education on changes in downstream behaviors. The authors find weaker but similar results for physical health. The findings suggest that financial education efforts should be refocused to foster subjective knowledge and improved behavior.
Keywords
In a survey by the National Endowment for Financial Education (NEFE), 88% of U.S. adults thought that high school students should be required to take a semester- or year-long course in financial education, covering topics such as spending and budgeting, managing credit, saving, earning income, investing, and managing risk (NEFE 2022). Implicit in this is the assumption that financial education will yield long-term benefits. How and why does financial education cause improved financial behavior measured at a delay? Can financial education improve subjective financial well-being? If so, through what mechanisms?
In consumer research, an emerging literature examines the links of financial knowledge to financial behavior and financial well-being (Dholakia et al. 2016; Hadar, Sood, and Fox 2013; Ward and Lynch 2019). Many studies examine momentary effects of nudges and disclosures on an immediate decision via changing relative preferences among options (e.g., Bolton, Bloom, and Cohen 2011; Hadar, Sood, and Fox 2013; Long, Fernbach, and De Langhe 2018; Mrkva et al. 2021). Consumer researchers have been less involved in research on large-dose financial education designed to produce lasting changes in “good” behaviors and financial well-being (Atlas, Porto, and Xiao 2018; Chin and Williams 2020; Fernandes, Lynch, and Netemeyer 2014). Most work studying the effects of financial education has been by economists (Brown et al. 2016; Kaiser et al. 2022), with open questions about mechanisms that may produce longer-term effects (Horn et al. 2023).
Financial education is regarded by many as an antidote to consumers’ financial struggles. As of the first quarter of 2023, the average credit card debt per borrower was $5,733 (TransUnion 2023). As of June 2023, 43.6 million U.S. citizens owed federal student loan debt averaging $37,717 (Hanson 2023). Only 44% of U.S. adults say they have enough to pay an emergency expense of $1,000.00 or more from their savings (Gillespie 2024), and only 68% can cover a $400 unexpected expense (U.S. Federal Reserve 2022). Between three and four in ten employees cash out their 401(k) when changing jobs in the United States, losing progress on savings and paying taxes and penalties (Wang, Zhai, and Lynch 2023). The median household retirement savings estimate of baby boomers is only $289,000 (Collinson and Cho 2023).
Many have argued that low financial knowledge causes these struggles in an increasingly complex financial world. It is often cited that only about one-third of adults can answer all three of the following questions correctly (Elkins 2017; Lusardi and Mitchell 2014):
Suppose you have $100 in a savings account and the interest rate was 2% per year. After five years, how much do you think you would have in the account if you left the money to grow? A. More than $102; B. Exactly $102; C. Less than $102; D. I don’t know (Correct answer: A) Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After one year, how much would you be able to buy with the money in this account? A. More than today; B. Exactly the same as today; C. Less than today; D. I don’t know (Correct answer: C) Do you think the following statement is true or false? Buying a single company stock usually provides a safer return than a stock mutual fund. (Correct answer: false)
Our research asks how much it matters to know the correct answers to questions like these if we are trying to understand whether financial education causes consumers to perform later positive financial behaviors and enjoy a sense of financial well-being. Surprisingly, results from our quasi-experiment show that when financial education produces large improvements on downstream financial behavior, these effects are mediated by increases in subjective knowledge and not objective knowledge. The same is true of small effects on financial well-being.
Consumer Financial Protection Bureau Model of How Financial Education Produces Increased Financial Well-Being
The working hypothesis among academics, policy makers, and practitioners is that financial education improves objective financial knowledge, thereby improving consumer financial behavior and thereby financial well-being (Consumer Financial Protection Bureau [CFPB] 2015, 2020). The Dodd-Frank Act (2010) specifically establishes an Office of Financial Education “which shall be responsible for developing and implementing initiatives intended to educate and empower consumers to make better informed financial decisions.” The statute includes 29 mentions of “financial literacy” and directs the bureau to work with various agencies “to implement the strategy to improve financial literacy of consumers.”
Figure 1 comes from an influential framework by the CFPB (2015, p. 46). The framework hypothesizes that the social and economic environment—including financial education—affects knowledge and skills that influence financial behavior. Financial behaviors—along with available opportunities—cause financial well-being. Models by Serido, Shim, and Tang (2013) and the FINRA National Financial Capability Study (Lin et al. 2022) show constructs and a causal flow like the CFPB model.
Financial Well-Being: The Goal of Financial Education from CFPB (2015).
Prior studies have tested relationships of only subsets of variables in the CFPB framework, leaving open the possibility of omitted variable bias. Further, though many studies have examined the effects of subjective and objective knowledge on either behaviors or perceived financial well-being, most of these studies used measures of key constructs that lack evidence of reliability and validity (NEFE 2020). This leaves open the possibility that subjective and objective knowledge appear distinct because each has a partial effect on downstream variables, but their indicators are imperfect reflections of the same underlying construct (Birnbaum and Mellers 1979). In this article, we drill down on the “knowledge and skill” elements of the CFPB model by presenting a quasi-experiment, looking at change in behaviors and well-being after an educational intervention. Our study assesses a proposed causal mediational chain of financial educational intervention → knowledge → positive behaviors → well-being consistent with prevailing frameworks posited in the literature (CFPB 2015; Lin et al. 2022; Serido, Shim, and Tang 2013), which we extend to the domain of physical health. Specifically:
We distinguish the roles of objective knowledge from subjective knowledge and skill, focusing on the less appreciated effects of subjective knowledge. We show that financial education produced large effects on objective knowledge and behavior relative to effects found in the literature, but the mechanism by which financial education improved financial behavior was by its effects via improving subjective knowledge. We find no evidence for an indirect mediational flow of financial education → objective financial knowledge → subjective financial knowledge → positive financial behaviors. We test the implied flows in Figure 1 with validated measures of all constructs, providing the first rigorous causal test of how interventions intended to improve financial knowledge affect financial well-being. As an exercise in conceptual replication, we compare these causal flows in the domain of personal finance with causal flows in the domain of personal physical health.
Our article proceeds as follows. We outline our conceptualizations of financial and physical health literacy as having components of both objective and subjective knowledge, note the parallels among these two domains, and explain why it is useful to study educational interventions in both domains simultaneously. We present the results of two pilot studies used to derive valid measures of key constructs. These pilot studies informed our preregistered predictions for our quasi-experiment that yielded our main findings. We end by discussing implications for policy efforts geared at improving educational interventions in both domains.
Financial Literacy and Objective and Subjective Financial Knowledge
In consumer research, there is a long-standing recognition of the distinction between subjective and objective knowledge (Alba and Hutchinson 1987; Carlson et al. 2009; Moorman et al. 2004; Wood and Lynch 2002), as well as distinctions between knowledge about specific choice options versus knowledge of a general domain (Punj and Staelin 1983). Our interest is in how education affects subjective knowledge of a general domain. 1
Financial literacy has been defined as “people's ability to process economic information and make informed decisions about financial planning, wealth accumulation, debt, and pensions” (Lusardi and Mitchell 2014). Much research treats financial literacy as objective knowledge, reflecting accurate information stored in memory (Alba and Hutchinson 2000), most often measured as the percentage of correct answers on a personal finance test or quiz (Allgood and Walstad 2016; Houts and Knoll 2020; Lusardi and Mitchell 2014).
Subjective knowledge—what consumers think they know and the confidence they have in what they think they know (Bearden, Hardesty, and Rose 2001; Carlson et al. 2009)—has garnered increased attention among academics and public policy makers examining personal finance outcomes (Bruggen et al. 2017; CFPB 2015; Chin and Williams 2020; Mende and Van Doorn 2015; Netemeyer et al. 2018). Consistent with the view of financial literacy as domain-specific expertise with subjective and objective components, we conceptualize and measure financial knowledge as two separate constructs: objective financial knowledge and subjective financial knowledge. As noted by Remund (2010, p. 284), “Financial literacy is a measure of the degree to which one understands key financial concepts and possesses the ability and confidence to manage personal finances through appropriate short-term decision-making and sound, long-range financial planning, while mindful of life events and changing economic conditions.” Fernandes, Lynch, and Netemeyer (2014) note the disconnect between the conceptualization of financial literacy as a skill and existing measures focusing on objective knowledge. Their research examines financial behaviors like saving, planning, minimizing debt, cash flow management, investments, managing one's retirement savings, and paying bills on time. In predicting these behaviors, Fernandes, Lynch, and Netemeyer find—while controlling for objective financial knowledge—that there was a substantial partial effect of “confidence in information search” with scale items corresponding to those we use for our “subjective financial knowledge” measure. Thus, the present study connects changes in subjective financial knowledge to a battery of “positive financial behaviors” that require action rather than inaction or choice deferral.
In the economics and finance literatures, though, financial education is assumed to improve financial behavior via improvements in objective financial knowledge. Subjective knowledge, if mentioned at all, is at best an afterthought and at worst unhelpful self-delusion. The most highly cited review article on financial literacy (Lusardi and Mitchell 2014) has a very short section on “objective versus subjective measures of financial literacy,” concluding: “Another interesting finding on financial literacy is that there is often a substantial mismatch between peoples’ self-assessed knowledge versus their actual knowledge, where the latter is measured by correct answers to the financial literacy questions posed. … Even though actual financial literacy levels are low, respondents are generally rather confident of their financial knowledge and, overall, they tend to overestimate how much they know” (p. 15). A major meta-analysis by Kaiser et al. (2022) reports detailed statistics about how financial education affects objective financial knowledge and behavior as well, but the authors state: “We do not include self-assessments of changes in financial knowledge as an outcome, as they could be less reliable than test scores” (p. 260). Of their 76 randomized control group studies, 50 include an objective knowledge measure. A few include a measure of subjective knowledge but do not treat it as a distinct construct. Only 3 of the 50 studies mention conclusions about subjective knowledge (Brugiavini et al. 2020; Shephard, Kaneza, and Moclair 2017; Skimmyhorn et al. 2016).
Similarly, one of the most important recurring surveys documenting trends in financial literacy is the FINRA National Financial Capability Study. That survey has used a single-item assessment of subjective knowledge: “How would you rate your overall understanding of personal finance?” or “How would you assess your overall financial knowledge?” reflecting that subjective knowledge has been a footnote in understanding financial behavior and well-being. The most recent FINRA summary of the National Financial Capability Study makes no mention of any finding relating to that measure, with all discussion focusing on objective financial knowledge and measures of financial behavior and well-being (Lin et al. 2022).
Still, several correlational studies report that measures of subjective knowledge are better predictors of self-reported “good” financial behaviors than are measures of objective knowledge (Allgood and Walstad 2016; Anderson, Baker, and Robinson 2017; Fernandes, Lynch, and Netemeyer 2014; Xiao, Chen, and Chen 2014). The dominant view, though, is that financial education causes downstream effects via the proximate outcome of objective financial knowledge (not subjective knowledge), yet to our knowledge little causal evidence supports this view.
We present a quasi-experiment causally assessing the mediational chain of a financial educational intervention → objective and subjective knowledge → positive behaviors → well-being, which we extend to the domain of physical health. Table 1 compares the merits of our research relative to prior literature with respect to a financial educational intervention (if any) on downstream measures of knowledge and well-being, showing these unique advantages:
Conceptual Replication: Educational Effects in the Domains of Physical Health
We previously noted that our study also examines potential parallel educational intervention effects in the domain of physical health. This raises a question: why study the effects of educational interventions in the personal finance and physical health domains simultaneously? There are several reasons. First and foremost, we believe strongly in the value of “conceptual” replication, as it is always an inductive empirical question whether findings replicate across domains. According to Cook and Campbell's (1979) concept of deliberate sampling for heterogeneity, more inductive support is shown for relationships in a study if the findings prove robust across variations on multiple dimensions. Researchers rarely have a good sense of generality of findings; nor do they have a strong sense of what background factors might moderate them, producing different treatment effects if those factors were allowed to vary. Only accumulation of evidence from “imperfect” replications addresses that indeterminacy (Farley, Lehmann, and Ryan 1981). Our tests of physical health education on downstream health outcomes demonstrate the robustness of a model via deliberate sampling for heterogeneity.
Second, education efforts in both domains have similar motivations, claims, and policy recommendations. Just as personal finance knowledge is low, consumer knowledge in the domain of physical health is also low (Ratzan and Parker 2000, p. vi), and the current state of physical health for many Americans is poor. For example, the Centers for Disease Control and Prevention reported 41.9% prevalence of obesity among U.S. adults age 20 or over in the period leading up to March 2020 (Stierman et al. 2021). Only 24.2% of U.S. adults met the physical activity guidelines for aerobic and muscle-strength training in 2020 (Elgaddal, Kramarow, and Reuben 2022). Thirty-four million adults still smoked cigarettes as of 2020 (U.S. Department of Health and Human Services 2020). Like calls for more financial education (Dodd-Frank 2010), there have been similar arguments that more health education is needed to improve physical health knowledge and well-being (Centers for Disease Control and Prevention 2023). Simonds (1974, p. 9) states: “Minimum standards for ‘health literacy’ should be established for all grade levels K through 12. Those school districts that fall below standard should be provided with federal aid to develop programs with teachers qualified to teach health education.” Third, in both domains, we see small or mixed effects of education interventions on positive behaviors. Fernandes, Lynch, and Netemeyer (2014) find very small effects of financial education on financial behaviors, and Kaiser et al. (2022), when comparing a financial education group to a randomized control group, find a mean financial behavior difference of only .10 standard deviation units in 76 studies. Studies examining the effects of physical health education also show small effect sizes (Lustria et al. 2013; Snyder et al. 2004). 2
Related Studies.
Notes: OK = objective knowledge, SK = subjective knowledge. Dash indicates not reported.
Fourth, there are striking similarities among the constructs in the Figure 1 CFPB (2015) model and constructs in frameworks hypothesizing the effects of health literacy. Core to the CFPB model are the “knowledge and skills” that lead to performing financial behaviors that ultimately are hypothesized to affect financial well-being. Several health literacy models have these same “knowledge and skills,” operationalized as objective and subjective knowledge, as key constructs affecting health behaviors, which ultimately affect physical health well-being (Nutbeam 2008). Thus, we suggest that the knowledge and skills components of the CFPB model are potentially relevant to the domain of physical health.
Health Literacy and Objective and Subjective Health Knowledge
Definitions and measures of health literacy vary (Nutbeam 2008; Waters et al. 2018). As with financial literacy, these definitions and measures typically include subjective and objective knowledge components. Several conceptualizations favor a subjective “feeling of knowing” that leads to making healthy decisions. For example, Bishop et al. (2016, p. 889) define health literacy as the perceived “ability/capacity to obtain, process, and understand basic health information and services needed to make health decisions.” They assess it with a three-item measure that reflects this perception. Others also tap the perception that one can identify reliable information to improve one's health (Osborne et al. 2013; Waters et al. 2018).
As with personal finance, consumer research has distinguished objective and subjective physical health knowledge and highlighted the role of subjective knowledge in willingness to engage in external search and healthy behaviors (Moorman et al. 2004). Subjective physical knowledge has led individuals to feel empowered, take greater control over their behaviors, and achieve a greater sense of health security by feeling they understand health issues (Anderson et al. 2016). Other studies note relative effects of subjective versus objective health knowledge. For example, operationalized as a subjective knowledge construct, health literacy has been found to positively affect physical health (Zhang, Or, and Chung 2020), whereas objective physical health knowledge is not related to engaging in healthy behaviors (Nagy-Pénzes et al. 2020).
Preregistered Predictions
Figure 1 implies that changes in one's environment (e.g., exposure to financial education) change financial knowledge, which in turn changes financial behaviors, which in turn change financial well-being. Based on prior evidence (Fernandes, Lynch, and Netemeyer 2014; Netemeyer et al. 2018) and our own pilot and survey studies (described subsequently), our first preregistered prediction is that subjective financial knowledge will dominate objective financial knowledge in mediating effects of financial education on changes in positive financial behaviors.
The 2015 CFPB report argued that “a growing consensus is emerging that the ultimate measure of success for financial literacy efforts should be improvement in individual financial well-being” (p. 4). Thus, we predict that we will find the flow-through of financial education to financial knowledge to financial behaviors to perceived financial well-being, as implied by the CFPB framework. If this flow-through holds true, one still might expect smaller financial education effects on well-being than on behavior. So, beyond the test of financial education effects on valid measures of subjective financial well-being, a key contribution of our study is to test the total mediational chain implied by the CFPB model, and to distinguish financial education’s indirect effects on positive behaviors and financial well-being via changes in subjective knowledge versus such effects via changes in objective knowledge.
Finally, we predict an indirect effect of financial education to financial knowledge to financial well-being that does not go through changes in financial behavior. The CFPB model does not anticipate such a path, but we consistently found such a direct effect in correlational studies we conducted prior to our quasi-experimental study.
Given our conceptualizations of objective and subjective physical health knowledge, the parallels we noted between the personal finance and physical health domains, and results of pilot studies described subsequently, we tested whether these predictions would hold for physical health. Consistent with statistical reporting recommendations of an American Marketing Association task force (McShane et al. 2024), we report point and interval parameter estimates rather than null hypothesis tests.
Pilot Studies
Our measures and quasi-experiment predictions were refined by pilot studies. Fernandes, Lynch, and Netemeyer (2014) have developed and validated a 12-item measure of objective personal finance knowledge. We knew of no validated measure of objective physical health knowledge. Web Appendix A summarizes Pilot Studies A (N = 541) and B (N = 922), in which we used Amazon Mechanical Turk workers to validate a 15-question measure of objective physical health knowledge, assessing four domains of physical health knowledge recommended by academics and health policy advocates: (1) nutrition/diet, (2) disease risks/outcomes, (3) obesity risks/outcomes, and (4) exercise/fitness. Prior to the quasi-experiment, we consulted with professors teaching personal finance and personal physical health courses to refine measures relevant to college-age students, yielding measures of objective financial and physical health knowledge of 15 and 20 items, respectively. We converted objective knowledge to percentages in the analyses reported herein. Web Appendix B shows the measures used for the main quasi-experiment.
Pilot Studies A and B also helped refine our preregistered predictions for the main quasi-experiment. The key findings are as follows. First, correlations are quite small between objective and subjective financial knowledge (r = .09 in both pilot studies) and between objective and subjective physical health knowledge (r = .03 and r = −.02 in Pilot Studies A and B), showing that objective and subjective knowledge are different constructs (Mende and Van Doorn 2015). We also observed much larger correlations between objective financial knowledge and objective physical health knowledge (r = .42 and r = .50 in Pilot Studies A and B), and between subjective financial knowledge and subjective physical health knowledge (r = .47 and .50 in Pilot Studies A and B).
Second, the correlations of subjective knowledge with behaviors and well-being were stronger than corresponding correlations with objective knowledge. In Pilot Study A, the subjective financial knowledge correlations with positive financial behaviors and current financial well-being were .51 and .34, respectively; the corresponding correlations with objective financial knowledge were .17 and .17. The subjective physical health knowledge correlations with positive physical health behaviors and current physical health well-being were .33 and .23; the corresponding correlations with objective physical health knowledge were −.04 and −.03. In Pilot Study B, the subjective financial knowledge correlations with positive financial behaviors, current financial well-being, and perceived future financial well-being were .51, .25 and .63, respectively; the corresponding correlations with objective financial knowledge were .16, .26, and .00. The subjective physical health knowledge correlations with positive physical health behaviors and current and perceived future physical health well-being were .41, .40, and .43; the corresponding correlations with objective physical health knowledge were −.21, −.02, and −.18.
These patterns of correlation suggested that subjective knowledge would be the stronger driver of behaviors and perceived well-being within each domain, thus helping us refine our preregistered predictions for the main quasi-experiment study. 3
A Causal Test Via an Education Field Experiment
Our main study tests whether financial education produces downstream effects consistent with Figure 1, and whether downstream effects are driven by financial education effects on subjective rather than objective financial knowledge. The study was a difference-in-difference quasi-experiment with three primary goals and contributions:
We test whether financial education causally affects financial behavior, via mediators of objective and subjective financial knowledge. Our preregistered prediction was that subjective financial knowledge would dominate objective financial knowledge in mediating effects of financial education on changes in positive financial behaviors. We test whether financial education effects on financial knowledge and behavior flow through to affect financial well-being, as implied by the CFPB Figure 1 model, via the following mediational chain: financial education → changes in objective and subjective financial knowledge → changes in positive financial behaviors → changes in financial well-being. We test whether the causal flows for the effects of financial education on financial knowledge, behavior, and well-being are like those that we find for the effects of physical health education on health knowledge, behavior, and health well-being.
We compared three groups of first- and second-year college students enrolling in either a 14-week semester-long personal finance course, a 14-week semester-long personal physical health course, or a control group of same-aged students enrolled in neither course. Exposure to different types of education creates variation in the “social and economic environment” in the leftmost panel of Figure 1. For all students, we measure constructs at the beginning and end of the semester. This allows us to measure within-person changes from Time 1 to Time 2 (hereafter designated “T2-T1”) rather than analyzing these measures at a single point in time.
Requirements to Credibly Test Relative Mediation by Subjective Versus Objective Knowledge
The CFPB model in Figure 1 implies a mediational chain and the relative roles of objective and subjective knowledge in explaining financial education effects on financial behavior. Prior research has shown that financial education effects on behavior are small unless the educational intervention has many hours of instruction and unless the measure of delayed financial behavior is close in time to the end of the instructional delivery (Fernandes, Lynch, and Netemeyer 2014). Our study, shown in the top row of Table 1, has three key advantages over any prior study:
Our financial education intervention is a semester-long course with many hours of instruction. Our key constructs are measured for treatment and control groups before the class began and again after the course was complete so that the results are relevant to policy debates about delayed effects of large-dose classroom financial education. Compared with any prior causal study of effects of financial education on financial knowledge and behavior, our study has stronger measurement: validated measures of positive financial behaviors, subjective knowledge, objective knowledge, and financial well-being. We find very large treatment effects of financial education on T2-T1 changes in objective financial knowledge, subjective financial knowledge, and financial behavior using these reliable and validated measures, providing conditions conducive to assessing mediation.
Because of these three points, when we show that the financial education treatment effects on a validated measure of positive financial behaviors are mediated by subjective knowledge, but not by objective knowledge, one cannot apply the standard counterarguments that a negligible mediating effect of objective knowledge was due to weak treatment effects on objective knowledge or low reliability and validity of the objective knowledge measure.
These three points also make us uniquely able to test for the mediational paths connecting financial education to changes in a validated measure of perceived financial well-being. The same points apply to our conceptual replication of these tests in the domain of physical health, that is, health education → T2-T1 changes in health knowledge → T2-T1 changes in health behavior → T2-T1 changes in health well-being, where again we can test the differential mediating roles of objective and subjective physical health knowledge on downstream outcomes.
This study conceptually replicates a prior quasi-experiment we conducted in the fall of 2019. That study found large effects of financial education on objective and subjective financial knowledge, as well as large effects of financial education on financial behavior that were mediated by changes in subjective financial knowledge, but not by changes in objective financial knowledge. This prior study also found effects of physical health education on objective and subjective health knowledge, but at half the size of the aforementioned financial education effects, with negligible effects of health education on health behavior. In both domains, indirect treatment effects of education on within-domain downstream behaviors were mediated by changes in subjective rather than objective knowledge. We expected the current study to replicate these results. We preregistered our predicted replication and extension of that study on the Open Science Framework (https://osf.io/pkjr5). Data are provided on our ResearchBox page (https://researchbox.org/588).
Method
Subjects
College students served as experimental subjects as they represent a particularly appropriate sample for our study. Research shows that college students have little knowledge pertaining to personal finance (Lusardi, Mitchell, and Curto 2010; Serido, Shim, and Tang 2013). A 2009 FINRA National Financial Capability Study concludes that 18- to 25-year-olds lack the financial capability to make good financial decisions on their own (FINRA 2009); a 2013 FINRA study finds that college-age students are much more likely to engage in costly credit card behaviors (e.g., paying late fees; paying only the monthly minimum) than their older counterparts (Xiao et al. 2014), and while most college students (86%) know that taking a student loan means repaying more than they borrow, only about one out of three understand the financial terms associated with the loan (College Ave Student Loans 2019).
Studies also show that young adult knowledge of physical health issues is quite low (Jackson and Waters 2004). Sexton-Radek (2016) find that most college students did not possess adequate knowledge with respect to the risks of smoking, high cholesterol, and excessive alcohol consumption. Werner and Betz (2020) report that only 26.8% and 23.9% of college students could accurately report recommended physical activity and nutrition guidelines advocated by the U.S. Department of Health and Human Services. In their review article, Kühn et al. (2022) conclude that the general level of health literacy and knowledge among university students is insufficient and needs improvement for students to live physically healthy lifestyles.
Design
We employed a quasi-experimental design in which undergraduate students self-selected into one of three conditions during spring semester 2021. To test causality more validly, we use a difference-in-difference approach, comparing the nonequivalent groups in their changes in knowledge and behavior between the start of the semester (T1) and the end of the semester (T2). Our design can be diagrammed as in Table 2, with “O” indicating an observation and “X” indicating a treatment.
Study Design.
Notes: “O” indicates an observation, and “X” indicates a treatment.
Students receiving the financial education intervention enrolled in an elective college course on personal finance (n = 51). Those receiving the physical health education intervention enrolled in an elective college course on personal physical health (n = 146). A control group (n = 228) was drawn from a student subject pool enrolled in neither course. There was no overlap among these three groups. The sample was 62% female and 38% male and included 35% persons of color (see the Web Appendix for full racial and ethnic demographics).
After the first week of the semester, six students dropped the personal finance course, leaving 45 who completed both the T1 and T2 surveys. Eleven students dropped the physical health course, and another six did not complete the T2 survey, leaving 129 students who completed both surveys. Seventy-four control group students chose not to participate in the T2 survey, leaving a total of 154 students who completed both surveys. This left 328 students for all analyses that follow. We made no exclusions. Because we compare the financial education group with the control and the health education group with the control, the precision of our estimates of treatment effects benefits from the large control group.
Focal T2-T1 Gain Score Measures
In the first week of spring semester 2021, prior to any course content (T1), students across all conditions completed an online survey containing focal measures and covariates shown in Web Appendix B. We gathered measures of objective and subjective financial knowledge, current and future financial well-being, objective and subjective physical health knowledge, and current and future physical well-being.
Based on prior research (Fernandes, Lynch, and Netemeyer 2014) and in consultation with faculty teaching the personal finance course, we gathered responses to nine positive personal finance behaviors that best fit the college student population. To assess positive physical health behaviors, we used the 16-item Health Behavior Checklist of Hampson, Edmonds, and Goldberg (2018), as well as 15 items suggested by the professor teaching the physical health course. All behavioral items requested responses in a yes/no format: thus, the total number of personal finance behaviors performed could range from 0 to 9, and the total number of personal physical health behaviors performed could range from 0 to 31. We converted both the finance and physical health behavioral measures to a percentage of behaviors performed. Students responded to these exact same measures at the end of the semester (T2), prior to final exams. We computed T2-T1 gain scores for each of our focal measures. We converted changes in objective knowledge and positive behaviors to changes in percentages.
Covariates
We also measured 21 covariates that were later used in robustness checks of our results. Several of these covariates were specifically assessed to examine the potential effect of selection bias into one of the experimental conditions. One common concern in difference-in-difference designs with self-selection into conditions is that people with greater interest or ability might tend to self-select into the treatment group and gain more between the pretest and posttest for that reason rather than because of a treatment. We therefore measured prior interests in finance and health as well as T1 objective and subjective knowledge about money and health, T1 positive money and health behaviors, T1 current and expected financial and health well-being, and the percentage of college expenses being paid for by parents. The professors teaching the personal finance and personal health courses suggested other possible correlates of selecting into their courses, including interest in social activities, satisfaction with personal relationships, satisfaction with university experience, currently having a job (part-time or full-time; 1 = yes, 0 = no), male gender (1 = yes, 0 = no), and race (self-identified person of color; 1 = yes, 0 = no).
We also measured more general constructs that may have shared variance with our specific measures of positive behaviors and objective and subjective knowledge. We measured social desirability bias using a five-item short form (Hays, Hayashi, and Stewart 1989) because of a concern that self-reports of positive financial and health behaviors may be prone to social desirability bias. We included an eight-item numeracy scale (Weller et al. 2012) that might relate to objective financial and health knowledge, and a six-item short form on general self-efficacy (Chen, Gully, and Eden 2001) that might relate to our subjective knowledge measures.
Our primary preregistered analyses regress our focal T2-T1 gain score dependent variables on two dummy variables: finance (took personal finance class) and health (took personal health class). The reference condition was the control group. By analyzing T2-T1 gain scores, we control for selection due to main effects of individual difference variables. Subsequently, we describe two preregistered robustness analyses to evaluate whether selection of participants into classes might plausibly explain our key findings. These analyses made use of a battery of T1 measures as covariates in an analysis of covariance (ANCOVA) framework.
Threats to Internal Validity in Difference-in-Difference Quasi-Experiments
No method is generally necessary or sufficient to draw valid conclusions from a study. Researchers must think in terms of specific critiques and whether a given critique is (1) plausible given their priors, and (2) consistent with the specific pattern of data found. This is true of all designs, including true randomized control group designs (Hutchinson, Kamakura, and Lynch 2000; Shadish, Cook, and Campbell 2002).
Difference-in-difference quasi-experiments rule out many threats to internal validity associated with nonequivalent control group designs, rendering such threats impossible as explanations of group differences in T2-T1 changes on our dependent measures. These include selection threats of “main effects” of observed or unobserved individual differences in subjects and treatment groups that affect both the pretest and the posttest equally. The main threats to valid causal inference are selection × maturation effects whereby the three groups would have changed differentially over time in the absence of any treatments (Shadish, Cook, and Campbell 2002, pp. 139–40).
Specifically, in the present context, might there be some unobserved difference between participants in the control group and either the financial education group or the physical health education group? Critically, might some such difference plausibly cause illusory differences in the trajectory of growth over time on our key dependent variables? Because our main aim is to illuminate the relative roles of objective and subjective knowledge, we check for these plausible unobserved factors that might explain the key mediation findings we report subsequently.
To test whether people with greater interest or ability might tend to self-select into the treatment group and gain more between the pretest and posttest for that reason rather than because of a treatment, we included a test suggested by Shadish, Cook, and Campbell (2002, p. 240). A “rich get richer” account should operate both between and within groups, and so one should observe higher variance on key measures at the posttest than at the pretest. We tested for such alternative accounts, finding no support (see Web Appendix B).
In our study, we did not conduct multiple pretests to test the “parallel trends” assumption that the groups were changing at the same rates for two or more time periods before the treatments (Angrist and Pischke 2008). Our study used a single pretest that also precluded use of synthetic controls (Li 2020). Gill, Sridhar, and Grewal (2017) present an alternative way to address unobserved selection by maturation confounds in difference-in-difference quasi-experiments in which, as in our case, individuals self-select into treatment groups and no parallel trends tests are plausible. We rely on a “selection on observables” approach by using an ANCOVA, comparing the three groups on their T2-T1 gain scores for a given measure by the full battery of T1 measures excluding that measure. In that analysis, we can evaluate whether some characteristic of individuals at T1 predicts T2-T1 gain within and across groups.
To affect inference about the magnitude of an average treatment effect parameter, imbalance on a covariate must “explain” the observed treatment effect. Suppose that the treatment and control groups (coded 1 or 0) differed in the proportion of male participants (r = −.1), and the within-group correlation of the male gender dummy with T2-T1 gains in positive financial behavior is r = .15. That would imply that the selection effect on its own would cause the financial education group to have an advantage over the control group of −.1 × .15 = −.015, a minuscule effect. If the advantage of financial education over control was r = .24 (corresponding to d = .5), the magnitude of potential bias from a gender confound would not be material, but in any case, ANCOVA would correct the treatment estimate for that initial imbalance.
Preliminary Results: Correlations Among T2-T1 Change Scores for Our Focal Constructs
We first examined the pattern of zero-order correlations among our T2-T1 gain score focal independent and dependent variables with the total sample (N = 328). By correlating gain scores (rather than scores at T2 only), we subsequently partial out any variable relation due to shared “main effects” of individual characteristics affecting treatment effects. For example, T2 objective and subjective financial knowledge correlate (r = .34, 95% CI = [.24, .43]), but T2-T1 gain scores for these two variables have a lower correlation (r = .21, 95% CI = [.10, .31]). Similarly, T2 positive financial behaviors correlate with T2 subjective financial knowledge (r = .53, 95% CI = [.45, .60]) and objective financial knowledge (r = .17, 95% CI = [.06, .27]), but corresponding T2-T1 gains in positive financial behaviors correlate less with T2-T1 gains in subjective financial knowledge (r = .25, 95% CI = [.15, .35]) and objective financial knowledge (r = .05, 95% CI = [−.06, .14]).
We found parallel effects in T2-T1 gains in the domain of physical health. Gains in objective and subjective health knowledge correlate (r = .20, 95% CI = [.09, .30]). T2-T1 gains in physical health behavior correlate with gains in subjective health knowledge (r = .19, 95% CI = [.08, .29]), but not with gains in objective health knowledge (r = .03, 95% CI = [−.08, .14]).
Likewise, T2-T1 gains in future financial well-being correlate with gains in subjective financial knowledge (r = .36, 95% CI = [.26, .45]), but not with gains in objective financial knowledge (r = .021, 95% CI = [−.09, .13]). Gains in future health well-being correlate with gains in subjective health knowledge (r = .14, 95% CI = [.03, .24]) but not objective health knowledge (r = .00, 95% CI = [−.11, .11]). In both domains, gains in subjective knowledge, but not objective knowledge, predict gains in behavior; gains in expected future well-being in a domain are predicted by gains in subjective knowledge, but not objective knowledge. These patterns agree with our prior correlational studies summarized in Web Appendix C when analyzing these constructs at a single point in time rather than analyzing correlations among T2-T1 change scores for these measures.
Main Results
Does Financial Education Affect Knowledge, Behaviors, and Well-Being?
A main interest is in the mediational mechanisms by which financial education produces changes over time in downstream positive financial behaviors and perceived financial well-being. We compared the three groups’ T2-T1 gains over the course of the semester on our key dependent variables, thereby controlling for unobserved selection effects that arise as “main effects” of any individual difference characteristic that might differ among the groups.
In Table 3, Panel A, the key coefficients are in the bolded row for the finance dummy, comparing gains in the financial education group to the control. We see that the financial education group gained 8.39% more than the control group in objective financial knowledge percentage, .89 scale points more in subjective financial knowledge, and 8.51% more in positive financial behavior percentage. One can characterize each of these gains by a Hedges’ g (like d, the mean difference divided by the pooled within-group standard deviation). The effect of financial education on changes in objective financial knowledge percentage, subjective financial knowledge, and positive financial behavior percentage had g = .541, 1.392, and .416, respectively. In Table 3, Panel B, we see similar results for the health dummy on the health response variables, though not as pronounced as the effects found for the financial education group.
Effects of Finance and Physical Health Education on T2-T1 Changes in Objective Knowledge, Subjective Knowledge, and Positive Behaviors.
As a comparison, Kaiser and Menkhoff (2020) report 37 quasi-experiments testing school-based financial education interventions. They find that education treatments have, on average, sizeable impacts on (objective) financial knowledge (g = .33 SD) and smaller effects on financial behaviors among students (g = .07 SD). Thus, treatment effects in Table 3 are large relative to other reports in the financial education literature. This implies that we have unusually high power to investigate the mediational mechanisms that are the focus of this article.
Robustness Analysis 1
Our ResearchBox page presents the full models for Robustness Analysis 1. The addition of 20 T1 covariate controls to our two treatment dummies has little effect on model fit compared with our main specification with treatment dummies alone. For the financial measures, Fchange(21, 302) values for T2-T1 objective financial knowledge percentage, subjective financial knowledge, and positive financial behavior percentage were 1.14, 2.17, and 1.54, respectively. For physical health, the corresponding Fchange statistics were .70, 1.30, and 1.61. Across all analyses with 126 covariate coefficient estimates, only 9 (7%) had 95% CIs excluding zero. Our treatment effects on T2-T1 changes in Table 3 are poorly explained by preexisting group differences at Time 1 in those covariates.
Robustness Analysis 2
We conducted a second robustness check, detailed on our ResearchBox page. We regressed the T2 measures of our dependent variables on the same treatment dummies while adding our full battery of T1 covariate measures. Cook and Reichardt (1979) explain that in before-after designs with multiple randomized groups, ANCOVA using the T2 measure as a dependent variable and the T1 measure as a covariate is typically more precise than a within-person analysis of change scores. The two are equivalent only if the slope of the effect of a T1 response on T2 response equals 1. McKenzie (2012) also recommends ANCOVA on posttests as an alternative to difference-in-difference based on better statistical precision.
As expected, we find that when regressing T2 posttest measures of financial and physical health knowledge and behavior, adding the full battery of covariates improves fit substantially compared with analyzing T2 knowledge and behavior measures as a function of treatment dummies alone. For the financial measures, Fchange(21, 302) values for objective financial knowledge percentage, subjective financial knowledge, and positive financial behavior percentage were 14.56, 15.12, and 8.42, respectively. For physical health, the corresponding Fchange statistics were 8.19, 9.80, and 12.78. For each focal T2 measure, the coefficient for the corresponding T1 covariate was large in magnitude: .60, .59, and .53 for objective financial knowledge, subjective financial knowledge, and positive financial behaviors, and .53, .42, and .76 for the corresponding health variables.
Table 4 compares the estimated treatment effects from the base model in Table 3 to the estimates from Robustness Analyses 1 and 2. Comparing our main specification to Robustness Analyses 1 and 2 in Table 4, Panel A, we see very similar coefficients for the finance treatment dummy versus the control. Compared with the control, financial education increased objective financial knowledge percentage by between 8.39% and 10.62%, increased subjective financial knowledge by between .74 and .89 of a scale point, and increased positive financial behavior percentage by between 8.51% and 13.88%.
Coefficients for Financial Education and Physical Health Education Dummy Versus Control for Three Estimation Methods.
Notes: Robustness Analysis 1 refers to the T2-T1 ANCOVA, whereas Robustness Analysis 2 refers to the T2 ANCOVA.
Table 4, Panel B, shows the corresponding coefficients for the health treatment dummy versus the control in the three specifications. Treatment effects of health education are all less than half the size of the corresponding effects of financial education in Table 4, Panel A. We see similar coefficient estimates across rows for the three specifications for all three measures. Compared with the control, health education increased objective health knowledge percentage between 2.57% and 5.18%, increased subjective health knowledge between .35 and .42 of a scale point, and increased positive health behavior percentage between 3.23% and 4.15%. In Robustness Analysis 1, adding the T1 covariates slightly reduced coefficients and increased standard errors, so the 95% CIs include zero for T2-T1 changes in objective health knowledge percentage and positive health behavior percentage.
If our treatment effects were caused by selection into treatments based on covariates, we should find small coefficients compared with our main specification. That is not what we found. Of the six Robustness Analysis 1 coefficients, three are slightly larger and three are slightly smaller. The same is true when we compare the Robustness Analysis 2 coefficients with our main specification.
Are Financial Education Effects on Financial Behavior Mediated by Changes in Objective and Subjective Financial Knowledge?
We used the dummy variable regression approach within PROCESS (Hayes 2018, Model 4) to understand the effects of our finance and physical health treatments (relative to control) on changes in positive financial and physical health behaviors as mediated by within-domain changes in objective and subjective knowledge. Our ResearchBox page provides details, data, and code. Figure 2, Panel A, shows the process by which financial education (vs. control) caused T2-T1 gains in percentage of positive financial behaviors.
Effects of Financial Education and Physical Health Education.
The top path in Figure 2, Panel A, shows that (compared with the control), finance education led to an a11 = 8.39% gain in objective financial knowledge. However, there was a weak relation between gains in objective financial knowledge and gains in positive financial behaviors: a 1% gain in objective financial knowledge was associated with no change in financial behaviors (b11 = −.00%). The bottom path of Figure 2, Panel A, shows that finance education lifted subjective financial knowledge by a12 = .89 of a scale point, and a one-point gain in subjective knowledge was associated with a b12 = 7.31% gain in positive financial behaviors.
The “total effect” of finance education (vs. control) was to lift positive financial behaviors by c1 = 8.51% (95% CI = [1.69, 15.33]). The total effect is the sum of bootstrapped estimates of three effects in Figure 2, Panel A: the indirect effect of finance education on financial behaviors via gains in objective financial knowledge (a11 × b11 = −.02%, 95% CI = [−1.39, 1.71]) and subjective financial knowledge (a12 × b12 = 6.50%, 95% CI = [2.57, 11.20]), plus the “direct” (partial) effect of finance education on financial behaviors with those two mediators included in the model,
Are Physical Health Education Effects on Health Behavior Mediated by Changes in Objective and Subjective Physical Health Knowledge?
Figure 2, Panel B, shows how health education (vs. control) caused T2-T1 gains in percentage of positive health behaviors. Similar to our findings with financial education, the top path in Figure 2, Panel B, shows that (compared with the control), health education led to an a21 = 3.02% gain in objective health knowledge. However, there was a weak relation between gains in objective health knowledge and gains in positive health behaviors: a 1% gain in health knowledge was associated with no change in health behaviors (b21 = −.01%). The bottom path of Figure 2, Panel B, shows that health education lifted subjective health knowledge by a22 = .39 of a point, and a one-point gain in subjective knowledge was associated with a b22 = 2.79% gain in positive health behaviors.
The “total effect” of health education (vs. control) was to lift positive health behaviors by c2 = 4.15% (95% CI = [1.35, 6.95]). The total effect is the sum of bootstrapped estimates of three effects in Figure 2, Panel B: the indirect effect of health education on health behaviors via gains in objective health knowledge (a21 × b21 = −.02%, 95% CI = [−.45, .54]) and subjective health knowledge (a22 × b22 = 1.08%, 95% CI = [.29, 2.28]), plus the “direct” (partial) effect of health education on health behaviors with those two mediators included in the model,
Does Financial Education Affect Financial Well-Being?
Next, we examined whether financial education produced T2-T1 changes in perceived financial well-being, as implied by the CFPB (2015) model. Based on T2-T1 gain scores, we found that compared with the control condition, the finance group gained more in current financial well-being (b = .36, 95% CI = [.12, .60], Hedges’ g = .505) and expected future financial well-being (b = .30, 95% CI = [.09, .52], g = .469). These results—in the top row of Table 5, Panel A—are broadly consistent with the CFPB model in Figure 1. We believe this is the first preliminary causal evidence of a positive effect of financial education on a psychometrically validated measure of perceived financial well-being.
Coefficients for Financial Education and Health Education Dummy Versus Control for Three Estimation Methods.
Notes: Robustness Analysis 1 refers to the T2-T1 ANCOVA, whereas Robustness Analysis 2 refers to the T2 ANCOVA.
We again added our study covariates and conducted the same two robustness analyses used with our prior analyses. Table 5, Panel A, compares the estimated finance treatment effects from the main specification in Row 1 with the estimates from Robustness Analysis 1 and Robustness Analysis 2. Comparing our main specification with Robustness Analysis 1 and 2 in Table 5, Panel A, we see very similar coefficients for the finance treatment dummy versus the control for current (b = .36, .27, and .20, respectively) and future (b = .30, .37, and .34) financial well-being. Again, estimates with covariates had slightly wider 95% CIs. For all but current financial well-being in Robustness Analysis 2, 95% CIs excluded zero.
As previously noted, this study replicates an earlier version of this study that included the same measures of current financial well-being and overall life satisfaction but no measures of expected future well-being. That earlier study did not find financial education treatment effects on current financial well-being, with 95% CIs excluding zero. Thus, our findings from the current study should be taken as encouraging for the possibility that financial education affects downstream financial well-being, but much more research is needed on this issue.
Does Physical Health Education Affect Health Well-Being?
Table 5, Panel B, shows results when we followed the same analytic approach to assess effects of health education versus the control on measures of T2-T1 gains in current and expected future health well-being. Table 5, Panel B, shows trivial differences between the health education and control groups in T2-T1 gains in current (b = .07, 95% CI = [−.06, .20], Hedges’ g = .132) and expected future (b = −.03, 95% CI = [−.18, .11], g = .052) health well-being. Table 5, Panel B, shows similar coefficients in Robustness Analyses 1 and 2.
Thus, we found detectable effects of finance education on financial well-being, but not similar effects of health education. Tables 3 and 4 showed that finance treatment effects were more than double the size of health treatment effects in effecting changes in knowledge and behavior. We found a similar result in our earlier version of this study: financial education treatment effects were large relative to those of the health education course.
Are Financial Education Effects on Well-Being Mediated by Changes in Objective or Subjective Financial Knowledge and Changes in Financial Behavior?
The CFPB (2015) model implies the following chain of effects: financial education → financial knowledge → financial behavior → financial well-being. Using PROCESS (Hayes 2018) with a custom b-matrix, we tested whether financial education (vs. control) causes T2-T1 changes in objective knowledge and subjective financial knowledge, expecting that those changes in knowledge predict T2-T1 changes in the percentage of positive financial behaviors performed, which predict T2-T1 changes in perceived current and future financial well-being.
Consistent with the CFPB model, we found a small indirect effect from financial education → T2-T1 subjective financial knowledge → T2-T1 positive financial behavior → T2-T1 perceptions of future financial well-being (indirect effect = .032, 95% CI = [.005, .078]). We found an indirect path from financial education → T2-T1 subjective financial knowledge → T2-T1 future financial well-being (indirect effect = .249, 95% CI = [.132, .362]).
Here, subjective knowledge dominated objective knowledge as a mediator of financial education effects on downstream variables. We found trivial indirect effects of financial education → T2-T1 objective financial knowledge → T2-T1 positive financial behavior → T2-T1 current financial well-being (.000, 95% CI = [−.005, .007]) or T2-T1 future financial well-being (−.018, 95% CI = [−.070, .015]). As noted in Figure 2, Panel A, the objective knowledge → behavior link was weak, and the link from T2-T1 gains in financial behavior to gains in current financial well-being was also weak. The weak relation of T2-T1 changes in financial behavior to changes in current financial well-being also explains our finding of trivial indirect effects of financial education → T2-T1 subjective financial knowledge → T2-T1 positive financial behavior → current financial well-being (−.018, 95% CI = [−.070, .152]). The zero-order treatment effects on current well-being shown in Table 5, Panel A, arise due to a direct/partial effect of financial education (.380, 95% CI = [.112, .641]). Zhao, Lynch, and Chen (2010) argue that “direct” effects may reflect omitted mediators.
Are Health Education Effects on Well-Being Mediated by Changes in Objective or Subjective Health Knowledge and Changes in Health Behavior?
We ran similar mediational tests of the effects of health education on current and future health well-being. Here, we found negligible effects, but did find small indirect effects of health education on T2-T1 future health well-being broadly consistent with what we found in the domain of finance. The indirect link from physical health education → T2-T1 subjective health knowledge → T2-T1 health behavior → T2-T1 current health well-being was near zero (.004, 95% CI = [−.002, .012]), as was the indirect link to T2-T1 future health well-being (.006, 95% CI = [001, .016]). Again, we found a small indirect link of health education → T2-T1 subjective health knowledge → T2-T2 current health well-being (.044, 95% CI = [.003, .113]), and a small indirect link to T2-T1 perceived future health well-being (.049, 95% CI = [.001, .126]). All the preceding findings were unchanged in mediation models that added covariates for T2-T1 changes in relationship status, and current and future well-being in the domain other than the one being predicted. See our ResearchBox page for data and code. 4
Summary
Our study addressed three main issues. First, we tested whether financial education effects on behavior were mediated by changes in financial knowledge implied by the CFPB (2015) model in Figure 1. We found that, indeed, T2-T1 changes in financial behavior were mediated by T2-T1 changes in subjective financial knowledge. We found minuscule mediating effects of changes in objective financial knowledge and large effects of changes in subjective financial knowledge. To date, scholars have ignored improvements in subjective financial knowledge as a key outcome of financial education. That should change.
Second, the CFPB model implies that the end goal of financial education is to improve financial well-being. We provide the first formal causal test of the effect of financial education on a validated measure of financial well-being. The CFPB model implies that effects of financial education on financial knowledge and behavior should flow through to improve financial well-being: financial education → financial knowledge → financial behaviors → financial well-being. We found small effects consistent with that model. In addition, we found an indirect effect of financial education → T2-T1 subjective financial knowledge → T2-T1 future financial well-being, holding constant effects on financial behavior. We hope that our work inspires further research on the mechanisms by which financial education might improve financial well-being.
Third, we tested whether the causal flows for the effects of financial education on financial knowledge, behavior, and well-being are similar for the effects of physical health education. Here we find mostly parallel effects, though for our health education course, effects were notably smaller in magnitude. Health education (compared with control) led to T2-T1 improvements in objective health knowledge, subjective health knowledge, and positive health behaviors. Again, improvements in health behavior were mediated by changes in subjective health knowledge and not by changes in objective health knowledge. We did not find improvements in health behavior flowing through to predict improvements in health well-being, but we did find small indirect effects of health education → T2-T1 subjective health knowledge → T2-T1 current and future health well-being.
General Discussion
The most important general finding of our study is the power of subjective knowledge in transmitting effects of financial education to have positive effects for performing positive behaviors and expectations of well-being, accounting for the effects of objective knowledge. We also found causal evidence for the effects of an educational intervention to increase within-domain knowledge and behaviors. Though our financial education effect on objective knowledge is large relative to what has been reported in prior meta-analyses, subjective knowledge again showed the stronger effects on downstream constructs. We found causal evidence for a financial education → gains in subjective financial knowledge → gains in positive financial behaviors linkage. These mediational chains seemed to extend to gains in financial well-being. Though not as pronounced, we found similar effects for a physical health education → subjective health knowledge → positive health behaviors chain of effects. Within both domains, subjective knowledge had stronger effects on downstream constructs, relative to objective knowledge. 5
Our study has several implications for educational interventions around money and health. First, and as noted previously, subjective knowledge—what individuals think they know and the confidence they have in what they think they know—has stronger motivational effects on behavior in both domains than does objective knowledge. This suggests that educational efforts should focus just as much (if not more) on increasing consumers' subjective knowledge relative to increasing objective knowledge of facts and figures in both domains. Domain-specific factual objective knowledge has the propensity to decay in memory over time (Fernandes, Lynch, and Netemeyer 2014), potentially diminishing its effect on behaviors. This is consistent with the premise that financial numeracy alone (objective financial knowledge) may not be sufficient for consumers to understand and respond to specific financial decisions and behaviors (Bolton, Bloom, and Cohen 2011). Subjective knowledge may have stronger effects because it empowers individuals to act (Anderson et al. 2016). Interestingly, education effects on subjective knowledge may persist longer after an educational intervention than effects on objective knowledge (Atlas, Porto, and Xiao 2018).
What educational approaches might increase subjective knowledge, financial self-efficacy, and the likelihood of performing positive financial behaviors? Table 6 lists several approaches. Some of these approaches are from behavioral domains other than personal finance, but we argue that all approaches share the common thread of conveying information in easy-to-understand terms to reduce complexity (Long, Fernbach, and De Langhe 2018).
Educational Approaches to Increase Personal Finance Outcomes.
Consider the comparisons of two teaching methods by Drexler, Fischer, and Schoar (2014). Microentrepreneurs in the Dominican Republic were randomly assigned to a control group or one of two experimental groups. The groups received largely overlapping curricula over five to six weeks except for one or two sessions, when one group received training in formal accounting principles and the other group learned simple rules of thumb for financial decision making. Rules of thumb focused on separating business and personal accounts and personal versus business expenses. This made it much easier for microentrepreneurs to keep track of whether their businesses were profitable. The authors found that the rule-of-thumb training had a substantially more positive impact than the accounting training. This advantage was particularly striking for those who had reported initial low interest in training and low education. We speculate that rules of thumb may promote more subjective knowledge and confidence in one's ability to enact what one has learned than fact-based accounting education.
Other approaches share this tendency to focus on simple procedural principles rather than on financial facts. For example, the Dave Ramsey show outlines “Baby Steps” to build financial self-efficacy and well-being. These steps revolve around the simple rhetoric of paying off debt, saving, and having a frugal life, and are communicated in a manner void of technical terms that may be off-putting to the average consumer. Some evidence suggests that this type of approach positively affects behaviors and personal finance self-efficacy (Chopra 2021).
“Coproduction” might be another pedagogical approach to increase subjective knowledge and tendency to follow through on what one has learned (Mende and Van Doorn 2015), as when financial counselors and their clients cocreate the counseling. Coproduction is an important form of learning by doing. Learning by doing has been advocated as a route to confident skill development in education in mathematics (Linchevski and Kutscher 1998), entrepreneurship (Rasmussen and Sørheim 2006), and training of emergency room doctors (Vozenilek et al. 2004). Learning via coproduction may also be applicable in the domains of personal finance and physical health. For example, Carpena et al. (2019) show that financial goal setting combined with professional counseling encouraged positive financial behaviors, such as writing a budget and starting savings.
Perhaps just-in-time financial education (Fernandes, Lynch, and Netemeyer 2014) and financial education delivered at a teachable moment (Miller et al. 2014) produce greater downstream effects by enhancing motivation and subjective financial knowledge via enhanced perceived relevance. People learn what they think they need to know when they think they need to know it (Ward and Lynch 2019). These approaches require that a benevolent employer, financial institution, or policy maker be aware of when a critical financial decision is imminent (Lynch 2019; Sussman, Hershfield, and Netzer 2023; Wang, Zhai, and Lynch 2023).
Putting all of this together, consider a hypothetical financial education course aiming to improve student loan decisions. For example, financial education aimed to improve student loan choice might clearly target high school seniors (rather than college sophomores or high school freshmen). Pairs of high school students could work together with one another filling out college loan applications rather than receiving fact-based instruction on what is required to fill out a loan application. This might boost self-efficacy in working outside of class on the FAFSA (Free Application for Federal Student Aid) and other college loan information. The instruction might involve simple rules such as how to search websites that compare colleges on value for money, how to compare loan options, and how to avoid borrowing more than one needs just to be able to afford a richer lifestyle. Similar principles could inform financial education for new employees by exposing them to clear information and procedural guidance for choosing employer-based benefits.
Another implication is that our work shows that there may be value in uniting literatures on financial literacy and health literacy and how they shape behaviors in both domains. We show that behavior and expectations of well-being in each domain are governed by similar mechanisms with a central role for increasing subjective knowledge. This naturally leads to questions about whether educational efforts to increase the subjective knowledge in one domain affect downstream constructs in the other domains. Some scholars suggest that conveying information that increases subjective knowledge in both domains within the same educational intervention may benefit behaviors and well-being across both domains (Ryu and Fan 2023). Ubel's (2019) work, for example, shows how a lack of self-efficacy in navigating the financing of medical care leads to suboptimal health choices. Schickedanz et al. (2023) show that embedded financial coaching during pediatric health visits improved low-income families’ adherence to future health visits and vaccinations. In our study we (unexpectedly) found that financial education led to small T2-T1 gains in positive health behaviors. However, we did not see other evidence of spillover of health education to positive financial outcomes or of financial education to positive health outcomes.
Last, policy makers have recently turned to viewing financial well-being as a key outcome variable for policy interventions—at least as important as effects on financial behavior (CFPB 2015; Netemeyer et al. 2018). Our quasi-experiment showed effect sizes of financial education on changes in objective financial knowledge and financial behavior that are quite large compared with those reported in prior meta-analyses. This allowed us to detect small education effects on financial well-being. Effects on physical health well-being were trivial in magnitude. Much more work is needed on financial and health education effects on well-being.
Supplemental Material
sj-pdf-1-ppo-10.1177_07439156241228197 - Supplemental material for Financial Education Effects on Financial Behavior and Well-Being: The Mediating Roles of Improved Objective and Subjective Financial Knowledge and Parallels in Physical Health
Supplemental material, sj-pdf-1-ppo-10.1177_07439156241228197 for Financial Education Effects on Financial Behavior and Well-Being: The Mediating Roles of Improved Objective and Subjective Financial Knowledge and Parallels in Physical Health by Richard G. Netemeyer, John G. Lynch, Donald R. Lichtenstein and David Dobolyi in Journal of Public Policy & Marketing
Footnotes
Joint Editors in Chief
Jeremy Kees and Beth Vallen
Associate Editor
Sterling A. Bone
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
