The Effects of Financial Knowledge, Skill, and Self-Assessed Knowledge on Financial Well-Being, Behavior, and Objective Situation

Abstract

The effects of certain abilities on financial outcomes have been debated for several years. Some argue that financial knowledge is key to financial success, while others have found financial skill and self-assessed knowledge are more important. This study contributes to this debate by providing a descriptive analysis, whereby regression is used to study the simultaneous effects of financial knowledge, financial skill, and self-assessed knowledge on financial well-being, financial behavior, and objective financial situation. Although our methodology does not allow us to determine if relationships are causal, we show that self-assessed knowledge has little to no relationship with financial well-being, may have contrasting relationships with components of objective financial situation, and is weakly associated with good financial behaviors. Financial skill has the strongest relationship with financial well-being and financial behaviors, as well as some components of objective financial situation. Despite having a relatively weak (compared to financial skill) association with financial well-being and financial behaviors, financial knowledge has the strongest relationship with many components of objective financial situation.

1. Introduction

It is well-established that finances play a big role in our lives and impact our well-being. In 2018, Netemeyer et al. showed that the impact of financial well-being (FWB) on overall well-being is approximately equivalent to the combined impact of job satisfaction, assessment of physical health, and satisfaction with relationship support. The importance of FWB is also documented by the Consumer Financial Protection Bureau (CFPB), who state that ‘enhancing financial well-being is the ultimate goal of financial capability policies, programs, and interventions’ (CFPB, 2017b). To achieve a strong FWB, consumers must navigate a financial environment that has grown in complexity over the last several years (e.g., Célérier & Vallée, 2017). It now includes credit cards, several loan options such as lines of credit, mortgages, and payday loans, and a myriad of investment options such as stocks, bonds, real estate, and cryptocurrency. In addition, defined contribution pension plans have been widely adopted in place of defined benefit plans, placing more responsibility on the individual to plan for their retirement. Numerous studies from countries all over the world have shown that financial knowledge (FK) levels are low (e.g., Nicolini et al., 2013; Klapper & Lusardi, 2020; Bottazzi & Oggero, 2023; Lusardi & Streeter, 2023; Sconti & Fernandez, 2023), even among C-suite executives (Anderson et al., 2017), suggesting that many individuals are not well-prepared for today’s complicated financial environment.

Recognizing the limited FK of consumers worldwide, many governments have prescribed a solution of financial education programs designed to increase FK. For example, the CFPB’s mandate involves improving the FWB of Americans through improving their financial literacy. This decision is backed by several studies that have demonstrated relationships between FK and better financial behaviors (FBs) or outcomes (e.g., Lusardi, 2003; Agnew & Szykman, 2005; Norvilitis et al., 2006; Van Rooij et al., 2011; Bottazzi & Oggero, 2023; Lusardi & Streeter, 2023; Sconti & Fernandez, 2023). Fernandes et al. (2014) acknowledged that programs aimed to increase FK are a natural response but went on to show that FK has a limited or non-existent relationship with several FBs after accounting for abilities and psychological traits that had been omitted from other studies. Since then, more studies have followed suit, finding that the association between FK and either FBs or FWB is weak after controlling for several abilities and traits (e.g., Tang et al., 2015; CFPB, 2018b; Collins & Urban, 2020). Several studies have found that financial skill (FS) and self-assessed knowledge (SAK) are more important for positive financial outcomes (e.g., Allgood & Walstad, 2016; Anderson et al., 2017; CFPB, 2018b; Riitsalu & Murakas, 2019; Collins & Urban, 2020). It remains an open question to determine which attribute(s) financial programs should focus on most to improve financial outcomes.

Despite the considerable interest in understanding the effects of FK, FS, and SAK on financial outcomes and the documented importance of controlling for potentially important confounding variables (e.g., Fernandes et al., 2014), we are not aware of a regression study that has considered FK, FS, and SAK simultaneously. In this study, we perform a descriptive analysis through conducting several regressions to simultaneously consider the effects of FK, FS, and SAK on FWB, FB, and objective financial situation (OFS). Our findings contribute to the ongoing discourse about household finances and may provide some clarity regarding the contradictory results found in the literature.

2. Literature Review

One of the challenges in household finance research is that the definitions of various terms are not unanimously agreed upon. For example, the definitions of FWB from the CFPB (2017b) and Netemeyer et al. (2018) are not identical. In this study, we use data from the CFPB’s 2016 National Financial Well-Being Survey (CFPB, 2017b), so we have opted to use their definitions throughout this paper. The CFPB defines FWB as having control over day-to-day, month-to-month expenses, having the capacity to absorb a financial shock, being on track to meet financial goals, and having the financial freedom to make the choices that allow enjoyment of life. FK describes a capacity to correctly answer objective questions on topics such as investing, interest rates, and housing. FS refers to an ability to gain reliable information to facilitate financial decisions, process information to make sensible financial decisions, and execute financial decisions and adapt as necessary to ensure goal attainment. FB is based on financial management (e.g., paying credit card bills in full), planning (e.g., creating steps to stick to a budget), following through on financial intentions, and savings habits. OFS is composed of factors such as resources (e.g., savings, insurance), material hardships (e.g., inability to afford housing), and credit standing. In discussing the literature, we attempt to be as consistent as possible with these definitions without omitting relevant studies with slightly different definitions (e.g., FK always refers to objective questions but studies on FWB, even if their definitions are not identical to the CFPB’s, will be discussed).

Among FK, FS, and SAK, our impression is that FK is by far the most scrutinized. Extensive research has been devoted to studying the effects of FK on a variety of financial and/or economic outcomes. At a macro level, lower levels of FK are associated with a higher probability of default for a country’s commercial banking system (Klapper & Lusardi, 2020). At a micro level, individuals with lower levels of FK are less likely to own stocks (Van Rooij et al., 2011), more likely to feel overwhelmed when faced with certain pension decisions (Agnew & Szykman, 2005), and more likely to have higher debt loads (Norvilitis et al., 2006; Bottazzi & Oggero, 2023; Lusardi & Streeter, 2023). Furthermore, there appears to be a positive and causal link between participation in financial education programs (the presumed end goal of which is improved FK) and improved FBs (Kaiser et al., 2022). Multiple recent studies have concluded FK also has a significant relationship with FWB (e.g., Philippas & Avdoulas, 2020; Selvia et al., 2021) and financial resilience (Bottazzi & Oggero, 2023; Lusardi & Streeter, 2023). Lusardi and Mitchell (2014) suggest that the benefits of FK are plentiful, including ‘savvier saving and investment decisions, better debt management, more retirement planning, higher participation in the stock market, and greater wealth accumulation’.

In light of these positive results, it is perhaps surprising that there is a body of literature suggesting that FK does not have a meaningful impact on FBs such as cash flow management, credit management, saving, and planning (Tang et al., 2015; CFPB, 2018b; Sehrawat et al., 2021). Importantly, the findings of Fernandes et al. (2014) demonstrate that statistically significant relationships between FK and various FBs can disappear once one controls for certain abilities (e.g., numeracy) and traits (e.g., risk tolerance), suggesting that a link between FK and FB might be explained by personality and cognitive factors. FB is believed to completely mediate the relationships FK has with FWB and OFS (e.g., CFPB, 2018b; Sehrawat et al., 2021) (i.e., the relationship FK has with FWB and OFS is only via FB), so the suggestion that the link between FK and FB is fabricated by confounders also suggests that FK does not impact FWB or OFS. Indeed, Collins and Urban (2020) found that the effect of FK on FWB is very small—not significantly different from zero—after accounting for controls such as demographics and self-assessed financial abilities.

Other attributes (i.e., FS and SAK) have not received as much attention as FK. Although it may seem like these should be highly correlated with FK, studies have found a disconnect between them (e.g., Robb & Woodyard, 2011; Allgood & Walstad, 2016; Phelps & Metzler, 2024). Not only are FS and SAK different from FK, but several studies have found that these factors are more important for financial outcomes. Robb and Woodyard (2011) created a combined measure of FS and SAK, which they called financial confidence, and showed that this is more related to FB than FK. In another four studies, FS was found to be more strongly related to financial outcomes than was FK. Notably, these are the only four studies we are aware of that considered both FK and FS. In the first study, Netemeyer et al. (2018) used perceived financial self-efficacy (which is very similar to FS) and FK, among other things, to build models for current money management stress and expected future financial security, which combine to form their definition of FWB. Although perceived financial self-efficacy was only strongly related to current money management stress in one of their two studies (and not the one involving a representative sample of the American population), it was strongly related to expected future financial security, while FK surprisingly had a slightly negative relationship with expected future financial security. In the second study, the CFPB (2018b) developed and tested a conceptual model whereby the relationship FK and FS have with FWB and OFS is completely mediated by FB. Although the relationship between FK and FB was statistically significant, the CFPB concluded that the effect size was immaterial and opted to remove FK from the conceptual model but retain FS. In the third study, Collins and Urban (2020) found that, unlike FK, FS has a meaningful relationship with FWB. The fourth study (Netemeyer et al., 2024) found evidence of a causal link between FS (although they refer to it as subjective knowledge) and FBs and FWB. Like with FS, some studies have found that SAK is more strongly related to FWB and FB than actual FK (Allgood & Walstad, 2016; Anderson et al., 2017; Riitsalu & Murakas, 2019; Lind et al., 2020). We note that while several studies have suggested higher SAK is related to better FWB and/or FBs, there may be a downside to higher SAK as well. Comerton-Forde et al. (2022) considered both FWB and OFS, finding that while self-assessed understanding of financial products is positively associated with FWB, it is negatively associated with OFS.

In a recent clustering study considering FK, FS, and SAK, Phelps and Metzler (2024) found evidence that all three may be related to financial outcomes, but definitive conclusions could not be made because clustering does not account for confounding variables. To attempt to understand the relationships these attributes have with financial outcomes while accounting for confounding variables, regressions (or extensions thereof) are typically used. However, none of the past regression studies have considered all three of FK, FS, and SAK simultaneously. Xiao and Porto (2022) observed that few studies consider even both FK and FS without grouping them into a single construct. As mentioned previously, we were able to find only four such studies (CFPB, 2018b; Netemeyer et al., 2018; Collins & Urban, 2020; Netemeyer et al., 2024), none of which considered SAK. As evidenced by Fernandes et al. (2014), the omission of an important confounding variable can substantially impact the resulting findings. Thus, we believe that it is important for a regression study to be conducted that analyzes the effects of each of FK, FS, and SAK on financial outcomes while controlling for the effects of the other two. It is this gap in the literature that we address with our study.

3. Data

The CFPB is an influential organization within the household finance academic community, having made a number of contributions (e.g., CFPB, 2017a, 2018b). One of their contributions is the 2016 National Financial Well-Being Survey, which was made publicly available to support additional research on FWB and other related components (CFPB, 2017b). It is a rich source of information, with over 100 survey responses from 6394 respondents in the USA. In addition to questions covering FWB, FK, SAK, FS, FB, and OFS (see

Table A1. Survey questions asked about each of the five constructs of household finance defined by the Consumer Financial Protection Bureau (CFPB). Some of the questions have been paraphrased. This is nearly a reproduction of S1 Table 1 from Phelps and Metzler (2024).

Construct	Question
Financial well-being	I could handle a major unexpected expense.
	I am securing my financial future.
	Because of my money situation, I feel like I will never have the things I want in life.
	I can enjoy life because of the way I’m managing my money.
	I am just getting by financially.
	I am concerned that the money I have or will save won’t last.
	Giving a gift for a wedding, birthday, or other occasion would put a strain on my finances for the month.
	I have money left over at the end of the month.
	I am behind with my finances.
	My finances control my life.
Financial knowledge (Houts and Knoll’s scale)	Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account?
	Considering a long time period (e.g., 10–20 years), which asset normally gives the highest return: savings accounts, bonds, or stocks?
	Normally, which asset displays the highest fluctuations over time: savings accounts, bonds, or stocks?
	When an investor spreads their money among different assets, does the risk of losing a lot of money increase, decrease, or stay the same?
	If you were to invest USD 1000 in a stock mutual fund, it would be possible to have less than USD 1000 when you withdraw your money.
	Whole life insurance has a savings feature while term does not.
	Housing prices in the US can never go down.
	Suppose you owe USD 3000 on your credit card. You pay a minimum payment of USD 30 each month. At an Annual Percentage Rate of 12% (or 1% per month), how many years would it take to eliminate your credit card debt if you made no additional new charges?
	If interest rates rise, what will typically happen to bond prices?
	A 15-year mortgage typically requires higher monthly payments than a 30-year mortgage, but the total interest paid over the life of the loan will be less.
Financial knowledge (Lusardi and Mitchell’s scale)	Suppose you had USD 100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow?
	Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account?
	Buying a single company’s stock usually provides a safter return than a stock mutual fund.
Self-assessed knowledge	How would you assess your overall financial knowledge?
Financial skill	I know how to get myself to follow through on my financial intentions.
	I know where to find the advice I need to make decisions involving money.
	I know how to make complex financial decisions.
	I am able to make good financial decisions that are new to me.
	I am able to recognize a good financial investment.
	I know how to keep myself from spending too much.
	I know how to make myself save.
	I know when I do not have enough information to make a good decision involving money.
	I know when I need advice about my money.
	I struggle to understand financial information.
Financial behavior	I consult my budget to see how much money I have left.
	I actively consider the steps I need to take to stick to my budget.
	I set financial goals for what I want to achieve with my money.
	I prepare a clear plan of action with detailed steps to achieve my financial goals.
	I follow through on my financial commitments to others.
	I follow through on financial goals I set for myself.
	I paid all my bills on time.
	I stayed within my budget or spending plan.
	I paid off my credit card balance in full each month.
	I checked my statements, bills, and receipts to make sure there were no errors.
	Putting money into savings is a habit for me.
Objective financial situation	I worried whether our food would run out before I got money to buy more.
	The food that I bought just didn’t last and I didn’t have money to get more
	I couldn’t afford a place to live.
	I or someone in my household needed to see a doctor or go to the hospital but didn’t because we couldn’t afford it.
	I or someone in my household stopped taking medication or took less than directed due to costs.
	One or more of my utilities was shut off due to non-payment.
	How confident are you that you could come up with USD 2000 in 30 days if an unexpected need arose?
	How much do you have in savings today (in cash, checking, and savings account balances)?
	How many financial products do you own (e.g., checking/savings account, life insurance, pension, etc.)?
	Were you contacted in the past year by a person or company trying to collect past-due debt?
	Were you turned down for credit in the past year?
	Did you choose not to apply for credit in the past year because you thought you’d get turned down?
	In a typical month, how difficult is it for you to cover your expenses and pay all your bills?

in Appendix A), the survey asked several questions related to demographics (e.g., age, income, race/ethnicity, level of education; see

Table A2. A summary of the demographic makeup of the survey respondents in the 2016 National Financial Well-Being Survey. This table is a reproduction of Table 1 from Phelps and Metzler (2024).

Demographic Group	Percentage in Each Demographic Group
Age
18–34	23.93
35–44	12.95
45–54	16.81
55–69	27.04
70+	19.27
Race or ethnicity
White, non-Hispanic	70.35
Black, non-Hispanic	10.71
Other, non-Hispanic	5.25
Hispanic	13.68
Gender
Male	52.42
Female	47.58
Education
Less than high school	6.71
High school or GED	25.37
Some college or Associate’s degree	30.23
Bachelor’s degree	20.52
Graduate or professional degree	17.17
Marital status
Married	59.84
Widowed	5.63
Divorced or separated	10.78
Never married	18.00
Living with partner	5.76
Household income
Less than USD 30,000	19.16
USD 30,000–USD 49,999	16.91
USD 50,000–USD 74,999	18.08
USD 75,000–USD 99,999	14.94
USD 100,000+	30.92

in Appendix B) and traits and abilities related to finances, such as frugality, self-control, and numeracy. Many of the survey’s questions are measured using a Likert scale. The dataset includes scores for FWB (CFPB, 2017a), FK (Lusardi & Mitchell, 2011; Houts & Knoll, 2020), and FS (CFPB, 2018a), which are computed from the raw survey responses. Except for the FK scale developed by Lusardi and Mitchell, all of these scores were created using item response theory, so the scores are based on a weighted aggregation of the questions shown in Table A1. Although scores for FB and OFS (CFPB, 2018b) were also created, they were not included in the dataset. The sample of survey respondents overrepresents some parts of the American population, so the CFPB computed a weight for each respondent based on their age, race/ethnicity, gender, education, household income, census region, whether or not they own a home, whether or not they live in a metropolitan area, and income in relation to federal poverty levels.

Although we have already stated how the CFPB defined FK, FS, and SAK, it is important to understand how the CFPB measured these constructs and the potential relationships between them. As discussed earlier, one might expect FK to be highly correlated with FS and perhaps even more so with SAK. However, after removing 57 respondents without a meaningful response for either FS or SAK, the Pearson correlations that FK has with FS and SAK are only 0.19 and 0.23, respectively. The most correlated constructs are FS and SAK, with a Pearson correlation of 0.63. This might be because both FS and SAK are measured using subjective questions. This is an important additional distinction between FK and FS; not only is one focused on knowledge while the other is focused on day-to-day abilities, but the way that they are measured differs too. As indicated by the low correlation between FK and SAK, the difference between objective and subjective measurements can be substantial. Subjective measurements can also be biased; on this same dataset, Phelps and Metzler (2024) demonstrated that different groups of people have varying biases about their FK. Such biases might limit the effectiveness of metrics obtained via answers to subjective questions.

4. Methodology

In this study, we investigated the effects of FK, FS, and SAK on financial outcomes, which include FWB, FB, and OFS. To do so, we used ordinary least squares and logistic regression, two techniques that have been used in related studies (e.g., Van Rooij et al., 2011; Fernandes et al., 2014; Collins & Urban, 2020). Ordinary least squares regression involves modeling a continuous dependent variable, while logistic regression is used to model a binary dependent variable.

The ordinary least squares regression is well-suited to modeling FWB and the logistic regression is well-suited to modeling binary outcomes, such as whether or not an individual has health insurance. However, many of the survey questions were measured using a Likert scale. For these responses, which have only a small number of possible values, we converted the numeric scales to binary variables to permit the use of logistic regression by defining a breakpoint in the scale for each question (e.g., ≥4 is encoded as 1, <4 is encoded as 0). Details about the threshold for individual questions are available in

Table A3. Initial scale and binary encoding of survey questions. This table details (paraphrased) survey questions, their initial scale, and how they were encoded for the logistic regression models. For all questions, those who refused to answer the question were removed. This is a reproduction of S2 Table 1 from Phelps and Metzler (2024).

Survey Question	Initial Scale and Binary Encoding
I consult my budget to see how much money I have left.	These were measured on a scale from 1 (strongly disagree) to 5 (strongly agree). Responses of 4 or greater were considered a ‘yes’ (encoded as 1) and remaining responses were considered a ‘no’ (encoded as 0).
I actively consider the steps I need to take to stick to my budget.
I set financial goals for what I want to achieve with my money.
I prepare a clear plan of action with detailed steps to achieve my financial goals.
Putting money into savings is a habit for me.
I follow through on my financial commitments to others.	These were measured on a scale from 1 (not at all) to 5 (completely). Responses of 4 or greater were considered a ‘yes’ (encoded as 1) and remaining responses were considered a ‘no’ (encoded as 0).
I follow through on financial goals I set for myself.
I paid all my bills on time.	After the survey, two possible answers (not applicable and never) were combined, so these were measured on a scale from 1 (not applicable/never) to 5 (always). Responses of 4 or greater were considered a ‘yes’ (encoded as 1) and remaining responses were considered a ‘no’ (encoded as 0).
I stayed within my budget or spending plan.
I paid off my credit card balance in full each month.
I checked my statements, bills, and receipts to make sure there were no errors.
I worried whether our food would run out before I got money to buy more.	These were measured on a scale from 1 (never) to 3 (often). Responses of 1 were considered a ‘no’ (encoded as 0) and remaining responses were considered a ‘yes’ (encoded as 1).
The food that I bought didn’t last and I didn’t have money to get more.
I couldn’t afford a place to live.
I or someone in my household needed to see a doctor or go to the hospital but did not go because we couldn’t afford it.
I or someone in my household stopped taking a medication or took less than directed due to the costs.
One or more of my utilities was shut off due to non-payment.
How confident are you that you could come up with USD 2000 in 30 days if an unexpected need arose within the next month?	This was measured on a scale from 1 (certainly could not) to 4 (certainly could) and included an ‘I don’t know’ option. Responses of 4 were considered a ‘yes’ (encoded as 1) and remaining responses were considered a ‘no’ (encoded as 0).
How much money do you have in savings today (in cash, checking, and saving account balances)?	Respondents were provided with several ranges as options. We encoded those that reported savings of USD 5000 or more as 1 and others as 0. Those that didn’t know or preferred not to say were removed.
In the past 12 months, have you been contacted by a person or company trying to collect past-due debt from you?	Respondents had the options of ‘yes’, ‘no’, and ‘not sure’. Responses of ‘yes’ and ‘not sure’ were encoded as 1, while ‘no’ was encoded as 0.
Were you turned down for credit in the past year?	A response of ‘yes’ was encoded as 1 and a response of ‘no’ was encoded as 0.
Did you choose not to apply for credit in the past year because you thought you’d get turned down?
In a typical month, how difficult is it for you to cover your expenses and pay all your bills?	This was measured on a scale from 1 (not at all difficult) to 3 (very difficult). Responses of 1 were considered an ability to easily make ends meet and encoded as 1, while other responses were encoded as 0.

in Appendix C. This process was conducted in exactly the same way as described in previous work on these data (Phelps & Metzler, 2024).

Our model for FWB can be expressed by the following:

$${FWB}_i={\beta }_0+{\beta }_1{FK}_i+{\beta }_2{FS}_i+{\beta }_3{SAK}_i+\mathit{\gamma }{Controls}_i+{ϵ}_i,$$

(1)

where ${FWB}_i$ is the financial well-being for individual $i$, ${FK}_i$ is financial knowledge, ${FS}_i$ is financial skill, ${SAK}_i$ is self-assessed knowledge, $\mathit{\gamma }$ is a vector of coefficients for our control variables (explained in a subsequent paragraph), and ${ϵ}_i$ represents random error.

Our model for each component of OFS and FB took the following form:

$$\mathrm{l}\mathrm{o}\mathrm{g}({\displaystyle \frac{p_i}{1-p_i}})={\beta }_0+{\beta }_1{FK}_i+{\beta }_2{FS}_i+{\beta }_3{SAK}_i+\mathit{\gamma }{Controls}_i,$$

(2)

where $p_i$ is the probability that individual $i$ takes value 1 for the dependent variable (e.g., having health insurance, consulting a budget) and the remaining components are exactly as described for the ordinary least squares regression model.

In all of our models, our controls include gender, age, race/ethnicity, education, household income, marital status, presence of a financially dependent child, retirement status, confidence in ability to achieve financial goals, frugality, perceived economic mobility, self-control, and objective numeracy (using the survey’s multiple choice question on this topic but not the open-ended one), which have been identified in previous studies as potential confounders (e.g., Fernandes et al., 2014; Allgood & Walstad, 2016; CFPB, 2018b) and are directly available in our data. We opted to use the same representation of self-control as the CFPB (2018b), combining the results from three questions into a single score. We also controlled for census region, home ownership status, if the individual lives in a metropolitan area, and income in relation to the federal poverty level, which were involved in the CFPB’s survey weighting process (CFPB, 2017b). As recommended in Solon et al. (2013), we performed regressions both with and without the survey weights. Weighted regressions were conducted using the svyglm function from the survey package (Lumley, 2024) in R. Our reported regression results are from regressions without inclusion of the weights, but the results from regressions with the weights are available in the Supplementary Materials. Robust standard errors (White, 1980) were computed using the sandwich package (Zeileis, 2004, 2006; Zeileis et al., 2020), with the lmtest package (Zeileis & Hothorn, 2002) used to compute the p-values associated with each coefficient. All the controls were encoded as factor variables, which is the most flexible way of modeling them without including interactions. By including an extensive list of potential confounders as control variables in the models and encoding them in this way, we mitigated the risk of omitted variable bias.

Prior to modeling, all the numeric variables were standardized to have a mean of zero and standard deviation of one, permitting comparisons of the magnitude of the effect for the continuous predictors. For each regression, all missing values were removed prior to fitting the model. Since the models have different responses, the datasets used to fit each model are not exactly the same.

5. Results

The coefficients and robust standard errors for the numeric variables from the linear regression of FWB on FK, FS, SAK, and the controls are shown in

Table 1. Regression for financial well-being.

Financial Skill	Financial Knowledge	Self-Assessed Knowledge
0.208 ** (0.017)	0.066 ** (0.013)	0.013 (0.014)

Coefficients and robust standard errors (in parentheses) for scaled numeric variables from the regression on financial well-being. The full set of coefficients is shown in Table S1 in the Supplementary Materials. ** and * indicate the coefficient is statistically significant with $p<0.01$ and $p<0.05$, respectively.

. See Table S1 in the Supplementary Materials for a table with the full set of coefficients (i.e., including controls). Recall that the numeric variables (FWB, FK, FS, and SAK) were all scaled. This means that the coefficient for a numeric variable represents the expected increase in FWB (in standard deviations) for one standard deviation increase in the independent variable. It also means the magnitudes of the coefficients for continuous variables can be directly compared to one another. Both FK and FS are highly significant ($p<0.001$) but SAK is not ($p\approx 0.348$). The coefficient for FS is more than three times as large as the coefficient for FK. Results from the regression when incorporating survey weights are similar (see Table S2 in the Supplementary Materials).

In the next two paragraphs, we will describe key patterns that emerge from

Table 2. Regressions for financial behavior.

Response	Financial Skill	Financial Knowledge	Self-Assessed Knowledge
Consult budget	0.366 ** (0.051)	−0.128 ** (0.040)	0.029 (0.041)
Plan steps to stick to budget	0.444 ** (0.053)	−0.039 (0.040)	0.133 ** (0.041)
Set goals	0.572 ** (0.056)	0.026 (0.042)	0.148 ** (0.044)
Plan steps to reach goals	0.603 ** (0.050)	−0.168 ** (0.041)	0.185 ** (0.044)
Follow through on commitments to self	1.636 ** (0.090)	0.053 (0.051)	0.179 ** (0.059)
Follow through on commitments to others	0.757 ** (0.075)	0.312 ** (0.057)	0.119 * (0.052)
Pay bills on time	0.339 ** (0.077)	0.344 ** (0.066)	0.037 (0.063)
Stay within budget	0.716 ** (0.063)	0.057 (0.044)	−0.054 (0.047)
Pay credit card in full	0.536 ** (0.055)	0.183 ** (0.042)	−0.059 (0.046)
Check statements	0.448 ** (0.065)	0.014 (0.049)	0.152 ** (0.047)
Saving is a habit	0.560 ** (0.067)	−0.010 (0.047)	−0.065 (0.049)

Coefficients and robust standard errors (in parentheses) for scaled numeric variables from the logistic regressions for financial behavior. The full set of coefficients is shown in Tables S3 and S4 in the Supplementary Materials. ** and * indicate the coefficient is statistically significant with $p<0.01$ and $p<0.05$, respectively. The coefficient with the largest value for each response is bolded.

and

Table 3. Regressions for objective financial situation.

Response	Financial Skill	Financial Knowledge	Self−Assessed Knowledge
Worried about running out of food †	−0.384 ** (0.068)	−0.451 ** (0.063)	0.171 ** (0.059)
Ran out of food †	−0.397 ** (0.070)	−0.505 ** (0.064)	0.118 * (0.059)
Could not afford housing †	−0.133 (0.083)	−0.613 ** (0.080)	−0.007 (0.065)
Could not afford health care †	−0.260 ** (0.064)	−0.387 ** (0.058)	0.159 ** (0.055)
Could not afford medication †	−0.178 ** (0.062)	−0.323 ** (0.058)	0.165 ** (0.052)
Utilities shutdown †	−0.199 * (0.093)	−0.723 ** (0.099)	0.074 (0.078)
Has life insurance	−0.017 (0.046)	0.085 * (0.040)	0.084 * (0.042)
Has health insurance	−0.048 (0.051)	0.273 ** (0.045)	0.056 (0.044)
Has retirement account	−0.012 (0.054)	0.381 ** (0.046)	0.152 ** (0.048)
Has non−retirement investments	0.145 ** (0.053)	0.476 ** (0.046)	0.113 * (0.050)
Can raise USD 2000 in 30 days	0.524 ** (0.060)	0.443 ** (0.049)	−0.040 (0.052)
Has liquid savings of USD 5000 or more	0.412 ** (0.061)	0.359 ** (0.050)	−0.124 * (0.053)
Contacted by debt collector †	−0.293 ** (0.062)	−0.116 * (0.055)	0.122 * (0.051)
Rejected when applying for credit †	−0.195 ** (0.071)	−0.275 ** (0.064)	0.111 (0.061)
Did not apply for credit because of anticipated rejection †	−0.332 ** (0.074)	−0.151 * (0.064)	0.096 (0.059)
Can make ends meet	0.460 ** (0.056)	0.221 ** (0.046)	−0.045 (0.048)

Coefficients and robust standard errors (in parentheses) for scaled numeric variables from the logistic regressions for objective financial situation. The full set of coefficients is shown in Tables S7–S9 in the Supplementary Materials. ** and * indicate the coefficient is statistically significant with $p<0.01$ and $p<0.05$, respectively. The coefficient with the largest absolute value for each response is bolded and negative financial outcomes are denoted with ^†.

. These will be synthesized in the next sections. Table 2 shows the coefficients and robust standard errors for the scaled numeric variables from our regressions for FBs. The full suite of coefficients is reported in Tables S3 and S4 in the Supplementary Materials. FS is clearly most related to FB, while FK and SAK seem to have only a limited association with FB. FK is statistically significant in about half of the regressions, but the magnitudes of the coefficients are generally small, and for two of them—consulting a budget and planning steps to reach financial goals—the coefficient is negative. This may be because those with higher levels of FK do not need to consult their budget or explicitly plan steps to reach their goals. All of the statistically significant relationships between SAK and various FBs are positive, but the effect sizes are small relative to FS. The weighted regressions produced similar findings (see Tables S5 and S6 in the Supplementary Materials), but with even more emphasis on the relative importance of FS. The only FBs for which we found FK and FS have statistically significant relationships are planning steps to reach goals (both FK and SAK), following through on commitments to others (just FK), and following through on commitments to self (just SAK). The relationships between FS and all eleven FBs remain significant when using the weights in the regressions.

For the logistic regressions for components of OFS, the coefficients and standard errors for the scaled numeric variables are shown in Table 3. The coefficients for entire models (i.e., with controls) are available in Tables S7–S9 in the Supplementary Materials. The table shows that FK consistently has the strongest negative relationship with material hardships (e.g., running out of food) and the strongest positive relationship with owning financial products (e.g., life insurance, having a retirement account). FS is generally significant at the 0.05 level for material hardships but not for owning financial products, and the magnitude of its coefficient is always smaller than the magnitude of the coefficient for FK for these aspects of OFS. For the six miscellaneous aspects of OFS at the bottom of Table 3 (e.g., has liquid savings of USD 5000 or more, contacted by a debt collector), FS tends to be the most strongly related. FK also has statistically significant relationships with these components of OFS, but its coefficient only has the largest magnitude regarding being rejected applying for credit. These findings are generally consistent with the results from the weighted regressions, although fewer of the relationships are statistically significant (see Tables S10–S12 in the Supplementary Materials). For SAK, we found that it is positively and statistically significantly related to three of the four financial products but also four of the six material hardships. For the miscellaneous components, SAK is positively associated with negative outcomes and negatively associated with positive outcomes, although only two out of six of these relationships are statistically significant. The weighted regressions reveal the same general trends, but none of the relationships are statistically significant.

6. Discussion

Our regression analyses allow us to identify relationships or associations between different components of household finance, but they do not allow us to identify if one component causes changes in another. Consequently, we restrict our commentary to associations between variables and thoughts that we hope will spur discussion and future studies in the household finance academic community.

6.1. When Considered Simultaneously, How Do FK, FS, and SAK Relate to FWB?

Through a linear regression of FWB on FK, FS, SAK, and many controls, we found that FS is most strongly related to FWB, while FK and SAK have little to no relation with FWB. In particular, SAK is not significantly related to FWB at a significance level of 0.05. This counters the findings of Anderson et al. (2017) and Riitsalu and Murakas (2019), but neither of these studies included FS or controlled for psychological characteristics. Indeed, when we remove FS and our controls for confidence in ability to achieve financial goals, frugality, perceived economic mobility, self-control, and objective numeracy, SAK is significantly related to FWB ($p<0.001$) with a coefficient more than three times as big as the one for FK. Although we found a statistically significant relationship between FK and FWB, the effect size is small, so our findings are broadly consistent with those of Collins and Urban (2020), who did not find a significant relationship between FK and FWB in their analysis of the CFPB data. It is not surprising that our results are not identical to theirs. Although we included many of the same control variables in the regression model, the variables are not identical, and, perhaps more importantly, we represented FK in different ways; although we both used questions developed by Houts and Knoll to assess FK, we used Houts and Knoll’s (2020) FK score, while Collins and Urban (2020) used the proportion of correct responses on a nine-item FK test. As an additional robustness check, we also fit a model where we measured FK using the number of correct answers in Lusardi and Mitchell’s (2011) FK test as our measure for FK. We again found a statistically significant relationship between FK and FWB, but the coefficient for FK was even smaller.

6.2. When Considered Simultaneously, How Do FK, FS, and SAK Relate to Components of FB?

In our unweighted regressions for various components of FB, we found that FK has only a limited relationship with FBs. At a significance level of 0.05, FK has a statistically significant effect on only five of eleven FBs, and the coefficients for two of those five FBs are negative. These results are consistent with the CFPB’s findings (2018b) from the same dataset, which led them to conclude that FK should be removed from their conceptual model. SAK has a significant effect in the regressions for six of the FBs, all of which are positive. In several instances, the coefficient for SAK is larger than the coefficient for FK, which is consistent with previous work (e.g., Allgood & Walstad, 2016; Anderson et al., 2017; Riitsalu & Murakas, 2019). Based on the weighted regressions, both FK and SAK only have statistically significant relationships with two of the eleven FBs.

Although FK and SAK have some relationship with FBs, their relationships are dwarfed by the relationship FS has with the FBs. FS has a positive and statistically significant relationship with all eleven FBs, with the largest coefficient in almost all cases. For many of the FBs, the coefficient for FS is substantially larger in magnitude than the coefficients for either FK or SAK. This is consistent with the four studies we found that considered both FK and FS (CFPB, 2018b; Netemeyer et al., 2018; Collins & Urban, 2020; Netemeyer et al., 2024), all of which found that FS is more strongly related to financial outcomes.

6.3. When Considered Simultaneously, How Do FK, FS, and SAK Relate to Components of OFS?

The unweighted regressions for aspects of OFS revealed that SAK is positively related to owning financial products. However, they also revealed that SAK is positively related to negative outcomes like experiencing material hardships and negatively related to positive outcomes such as having liquid savings of USD 5000 or more, although these findings are not statistically significant at the 0.05 level when weights are included in the regression. Even though the effect sizes are relatively small, this association with both positive and negative aspects of OFS is surprising. However, this is not inconsistent with other studies; Allgood and Walstad (2016) and Anderson et al. (2017) found an association between SAK and owning financial products, while Comerton-Forde et al. (2022) found that self-assessed understanding of financial products is negatively related to OFS. In Comerton-Forde et al. (2022), the magnitude of the standardized coefficient for self-assessed understanding of financial products is similar to that of the coefficient for being unemployed. In addition, their study did not explicitly consider FK as a predictor, so we wonder if the coefficient for self-assessed understanding of financial products would be even larger after controlling for FK. To our knowledge, our study is the first to find SAK has both positive and negative relationships with different components of OFS within a single study. The association with both positive and negative components of OFS suggests that more work should be carried out to attempt to understand the nuances of the relationships SAK has with financial outcomes.

Another interesting result from these regressions is that FK has a much stronger association with several aspects of OFS than does FS. Although FS still tends to have a significant relationship with components of OFS, the relationship between FK and material hardships and owning financial products is much stronger. This is quite surprising considering the previous results from other studies that have concluded FK is relatively unimportant when conducting regressions for financial outcomes (e.g., Anderson et al., 2017; CFPB, 2018b; Collins & Urban, 2020), as well as our own findings showing that FK has limited relationships with FWB and FBs. However, the regression results shown in Table 3 provide overwhelming evidence that FK has a meaningful relationship with OFS, specifically material hardships and the ownership of financial products. Given this strong relationship between FK and OFS and the belief that FB completely mediates the relationship between these two (CFPB, 2018b; Sehrawat et al., 2021), we would expect to find cases where the relationship between FK and a FB is stronger than the relationship between FS and that FB, but we have only a single case of that—paying bills on time—and the coefficients associated with FK and FS for this FB are nearly the same. We investigate these seemingly contradictory results in the next section.

7. Post Hoc Analysis: Explaining How Our Results Co-Exist

Although somewhat counterintuitive, our results suggest FK is strongly associated with OFS but weakly associated with FWB. However, the clustering study of these data demonstrated a disconnect between OFS and FWB (Phelps & Metzler, 2024), and previous results have shown that a given construct (i.e., self-assessed understanding of financial products) can be positively associated with FWB and negatively associated with OFS (Comerton-Forde et al., 2022), so this finding is plausible. To further solidify our understanding of how these results can co-exist, we performed some additional regression analyses. We constructed a very rudimentary OFS score and fit these two models:

$${OFS}_i={\beta }_{\mathrm{0,1}}+{\beta }_{\mathrm{1,1}}{FK}_i+{\beta }_{\mathrm{2,1}}{FS}_i+{\beta }_{\mathrm{3,1}}{SAK}_i+{\mathit{\gamma }}_1{Controls}_i+{ϵ}_i$$

(3)

$${FWB}_i={\beta }_{\mathrm{0,2}}+{\beta }_{\mathrm{1,2}}{FK}_i+{\beta }_{\mathrm{2,2}}{FS}_i+{\beta }_{\mathrm{3,2}}{SAK}_i+{\beta }_{\mathrm{4,2}}{OFS}_i+{\mathit{\gamma }}_2{Controls}_i+{ϵ}_i,$$

(4)

where ${OFS}_i$ is the objective financial situation for individual $i$, ${FK}_i$ is financial knowledge, ${FS}_i$ is financial skill, ${SAK}_i$ is self-assessed knowledge, ${FWB}_i$ is financial well-being, ${\mathit{\gamma }}_1$ and ${\mathit{\gamma }}_2$ are vectors of coefficients for our control variables, and ${ϵ}_i$ represents random error. Like in our earlier regressions, all numeric variables were standardized before modeling. It is worth noting that the estimated value of ${\beta }_{\mathrm{4,2}}$ was only 0.401, suggesting that a roughly 2.5 standard deviation increase in OFS is needed for a one standard deviation increase in FWB. This supports the finding of a disconnect between FWB and OFS (Phelps & Metzler, 2024).

Plugging Equation (3) into Equation (4) yields a formula very similar to our initial regression for FWB (see Equation (1)), but with decomposed coefficients, which can help us understand them better. The coefficient for FK is ${\beta }_{\mathrm{1,1}}{\beta }_{\mathrm{4,2}}+{\beta }_{\mathrm{1,2}}$ and the coefficient for FS is ${\beta }_{\mathrm{2,1}}{\beta }_{\mathrm{4,2}}+{\beta }_{\mathrm{2,2}}$. Our coefficient estimates from the regressions are ${\widehat{\beta }}_{\mathrm{1,1}}\approx 0.182$ and ${\widehat{\beta }}_{\mathrm{1,2}}\approx -0.005$, providing ${\widehat{\beta }}_{\mathrm{1,1}}{\widehat{\beta }}_{\mathrm{4,2}}+{\widehat{\beta }}_{\mathrm{1,2}}\approx 0.068$. Using ${\widehat{\beta }}_{\mathrm{2,1}}\approx 0.090$ and ${\widehat{\beta }}_{\mathrm{2,2}}\approx 0.171$, we obtain ${\widehat{\beta }}_{\mathrm{2,1}}{\widehat{\beta }}_{\mathrm{4,2}}+{\widehat{\beta }}_{\mathrm{2,2}}\approx 0.207$. Note that, to two decimal places, we have obtained the same coefficients as shown in Table 1. The fact that ${\widehat{\beta }}_{\mathrm{1,1}}>{\widehat{\beta }}_{\mathrm{2,1}}$ supports our finding that FK is more strongly related to OFS than is FS. However, it appears the relationship between FK and FWB is entirely through OFS, since ${\widehat{\beta }}_{\mathrm{1,2}}\approx 0$. On the contrary, FS is still related to FWB even given OFS, explaining how FS can have a much stronger relationship with FWB than FK, despite having a weaker relationship with OFS than FK. It is possible that, even given the same OFS, FS could lead to more positive feelings towards aspects of FWB like control of short-term finances and capacity to absorb a shock. Notably, the relationship we found here between FS and FWB could be caused in part by the fact that both are subjective constructs; it may be that their relationship is strengthened by some characteristic of individuals that affects how they respond to subjective questions (e.g., optimism). We would be interested in the findings of a similar analysis using an objective measure of FS.

Perhaps even more puzzling than the results discussed in the previous paragraphs are the strong relationships between FK and components of OFS yet limited relationships between FK and FBs. It is common to form a theoretical model whereby the effects of FK on OFS are completely mediated by FB (e.g., CFPB, 2018b), and indeed this seems like a reasonable framework. Thus, given the relationships between FK and components of OFS, it seems that we should observe multiple FBs that are more related to FK than FS, or potentially a single FB with a much stronger relationship with FK than with FS, but neither of these appear to be the case. Recall that our regression analyses do not allow us to determine causality. It is possible that our findings are an artifact of reverse causality. For example, it may be that those with a retirement account and/or non-retirement investments have more FK because they had to make decisions about how to invest their money, so they learned about various investment products. If this is the case, we no longer have a need for FBs to mediate the relationship between FK and OFS in order to explain our findings. This explanation is consistent with the CFPB’s (2018b) decision to remove FK from their conceptual model because of its limited relationship with FB.

It is less clear, however, how reverse causality could cause the relatively strong relationship between FK and material hardships. For example, it does not seem like an ability to afford food would lead someone to gain FK. One possibility is that those with financial products like retirement accounts gain their FK when making decisions about their accounts, but also simply are more comfortable financially so hardships like running out of food are less likely. If this explains much of what we observed, then the importance of FK in our regressions for material hardships should decrease when we consider only those without such assets. In Appendix D, we show the coefficients for FS, FK, and SAK for regressions on material hardships considering only renters who do not have a retirement account, non-retirement investments, or life or health insurance. The coefficients for FK are all still statistically significant ($p<0.01$), with four of the six still having the largest magnitude. For the other two—worrying about running out of food and running out of food—the coefficients are nearly the same as those for FS. This suggests that, in general, reverse causality is unable to explain our findings about the relationship between FK and OFS. We are unable to conduct the same kind of test for owning financial products, so it is possible that reverse causality explains much of those relationships, but reverse causality does not appear to be a strong explanation for the relationship between FK and material hardships.

Omitted variable bias could also play a role in the results we obtained. However, we have no reason to believe that this influences our results more than other similar studies. In fact, our study has provided evidence that other studies (e.g., Anderson et al., 2017; Riitsalu & Murakas, 2019) might have concluded that SAK is related to FWB because of the omission of important controls, including FS. As we have pointed out, our study is the only one that we are aware of that has simultaneously considered the effects of FK, FS, and SAK on financial outcomes. Given the inclusion of all three attributes and the extensive list of controls we included in the models, as well as the very reasonable conceptual model proposed by the CFPB (2018b), we posit that the most likely reason for our seemingly contradictory results is the omission of at least one important FB. Note that this FB would not be included as a control in any of our models (in the same way that other FBs were not included as controls) because FBs have been treated as (potentially intermediate) outcomes, so this missing FB would not cause omitted variable bias. However, this missing FB could be the reason why the CFPB (2018b) concluded that FK was unimportant when studying the same dataset that we have. Their conceptual model required that the effects of FK on OFS pass through FBs, so missing FBs could lead to concluding that FK has very little relationship with OFS even if that is not the case. It may be worthwhile for future studies to consider additional FBs in order to test our hypothesis.

The co-existence of some of our results helps provide some clarity regarding the mixed results in the literature. Based on our results, those that have argued that FK has a relatively limited association with FWB and several FBs (after sufficiently controlling for other factors) are right, but those who have found that FK is important for other financial outcomes are also correct. The details of the outcome variable make a larger difference than may have been expected; results change when the outcome is FWB compared to OFS, and we even see different results for different components of OFS. The lack of consistency in definitions in the literature becomes an even bigger challenge when the details are so important. Creating standardized definitions may help the academic community reach a consensus on the effects of FK, FS, and SAK on financial outcomes.

8. Other Limitations

Although we have already discussed some of the study’s limitations (e.g., potential omitted variable bias, potential reverse causality), there are other potential limitations to our study. First, our study is based on data from 2016; there have been changes in financial markets and tools, as well as economic conditions since then. While we do not believe these changes have led to substantial changes in the core relationships between the financial constructs considered in our study, we cannot say this with certainty. Second, our results might not generalize to all demographic groups. Although we controlled for many demographics (e.g., age, gender, race/ethnicity, income), we did not include interactions between these terms and FK, FS, and SAK in our models. The effects of FK, FS, and SAK could differ between these groups, but this was not captured in our study.

9. Conclusions

We have conducted what we believe to be the first regression study that has simultaneously considered the effects of FK, FS, and SAK on various aspects of household finance. We found that FS is strongly related to FWB and FBs, while FK and SAK have limited or statistically insignificant relationships with those aspects of household finance. SAK has both positive and negative relationships with various aspects of OFS, although these effects also tend to be small and, in the weighted regressions, statistically insignificant. FK demonstrated a strong relationship with OFS, which is a particularly noteworthy finding because a previous study of the same dataset dismissed the importance of FK (CFPB, 2018b). This finding is somewhat puzzling because FB is often thought to completely mediate the relationship between FK and OFS (e.g., CFPB, 2018b), yet FK had limited relationships with FB relative to those of FS. Thus, we suspect that the relationship between FK and OFS is caused by an association with some uncaptured FB. However, a future study would have to be conducted to determine if this is true.

If we assume that the conceptual model of the CFPB (2018b) is correct, then the relatively strong association of FK with OFS compared to its association with FWB means policymakers must carefully consider their goals when developing policies and interventions. Choosing the ‘wrong’ one can lead to the development of policies or interventions that are not directly in line with their underlying goals. If aiming to improve OFS, it seems an intervention designed to increase FK might be appropriate. If aiming to improve FWB, it seems an intervention targeting FS might be more suitable. Future studies aiming to further improve our understanding of the nuanced differences between various aspects of household finance and how they relate with one another may help policymakers develop solutions that more precisely target their goals. It also may be worthwhile for future studies to consider individual components of FWB. In our analysis, we considered individual components of OFS and FB because the scores were unavailable to us, but we found results that would not have been uncovered had we used an aggregated score. Studying individual components may help reconcile some of the conflicting findings in household finance literature.