From Risk Preferences to Portfolios: Comparing SCF Risk Scales and Their Predictive Power for Asset Ownership

Abstract

This study compares two risk tolerance scales used in the Survey of Consumer Finances (SCF), namely the long-standing 4-point scale and the newer 11-point scale, to determine which better captures an individual’s investment risk preferences. The analysis includes exploring how each scale relates to household demographics, socioeconomic factors, and ownership of risky versus conservative investments. By utilizing prospect theory, the findings reveal that while both scales effectively measure risk tolerance, the 11-point scale provides a more detailed understanding of differences in asset ownership across risk levels. For financial professionals, these results highlight the value of using a more granular risk assessment tool to better align investment strategies with client preferences, leading to improved client relationships and outcomes.

1. Introduction

Risk tolerance assessment has become a cornerstone of the financial planning industry, serving as a critical tool for crafting personalized investment strategies and fostering client confidence (Hayashi & Routh, 2024; Ingram & Thompson, 2012; Brayman et al., 2017). As the financial landscape grows increasingly complex and client expectations for customized solutions rise, accurately measuring risk tolerance has never been more important (Hayashi & Routh, 2024). Financial professionals rely on risk scales to bridge the gap between clients’ stated preferences and their actual behaviors, ensuring that investment portfolios align with both financial goals and emotional comfort (Ingram & Thompson, 2012).

In recent years, the industry has seen a push toward refining risk tolerance assessments to better capture the nuances of individual preferences (Hermansson & Jonsson, 2021; J. E. Grable & Rabbani, 2023). Traditional tools, such as the four-point scale used in the Survey of Consumer Finances (SCF), provide a general understanding of a client’s risk appetite. However, newer tools, like the 11-point scale introduced by the SCF in 2016, offer greater granularity, potentially revealing subtle variations in risk preferences that can inform more tailored financial advice (Kim et al., 2020). This shift reflects a broader trend in financial planning toward data-driven personalization and evidence-based practices (Kim et al., 2020).

For financial practitioners, understanding the strengths and limitations of various risk scales is crucial. Testing and comparing these tools equip advisors with better strategies to assess client preferences, navigate difficult conversations, and design portfolios that align with both risk tolerance and long-term objectives. As the industry continues to embrace innovation and adapt to evolving client demands, refining risk assessment methodologies will remain central to delivering high-quality, client-centered financial advice.

Purpose of This Study

This study evaluates the performance of two risk tolerance scales used in the Survey of Consumer Finances (SCF)—the long-standing 4-point scale and the newer 11-point scale introduced in 2016—to determine which better measures an individual’s risk tolerance and aids financial practitioners in addressing clients’ diverse needs. By analyzing how these scales relate to investment behaviors, such as ownership of risky assets like equities versus conservative assets like bonds, the study provides actionable insights for improving client outcomes. The analysis incorporates demographic and socioeconomic factors to assess the scales’ effectiveness and examines responses from 2022 to identify relationships in risk tolerance and asset ownership. Prospect theory (Kahneman & Tversky, 1979) is applied as a guiding framework to understand how individuals perceive and respond to risk, highlighting behavioral tendencies that may not align with traditional economic models. The findings have implications for practitioners, including refining risk assessment tools, enhancing client conversations about risk, and improving investment strategies, as well as for academics seeking to advance research in financial literacy and investment behavior.

Previous research has explored this topic by defining risk terms, asset allocation risk profiles, and the relationship between risk and expected reward. This study builds off of the existing literature to compare an individual’s subjective risk tolerance to their portfolio’s risk tolerance. Primarily, we seek to answer the following research question: Which measure of risk tolerance in the SCF more closely aligns with an individual’s portfolio risk tolerance?

2. Existing Literature

As defined in previous research, financial risk tolerance is the willingness to engage in behaviors in which the outcomes remain uncertain with the possibility of an identifiable negative outcome (J. E. Grable & Joo, 2004, p. 73; Irwin, 1993, p. 11). Roszkowski & Davey (2010) differentiated between risk tolerance and risk perception, stating that risk tolerance remains relatively constant regardless of investment performance and market volatility, while risk perception is subject to change. Similarly, Hoffmann et al. (2013) found that despite a drastic change in risk perception during the financial crisis in 2008, riskier investors maintained portfolios with higher-risk investments. Another study differentiated between subjective and objective risk tolerance, with subjective being measured from self-reported risk questions and objective risk tolerance being measured by asset allocation (Chang et al., 2004). The two risk scales in the SCF are considered subjective risk tolerance, as they are self-reported. Measuring asset classes would be considered objective risk tolerance.

It is useful to note that this study does not attempt to incorporate an individual’s risk capacity. The CFA Institute Research Foundation suggests that both risk capacity and risk appetite create an individual’s risk profile, which should be considered when investing (CFA Institute Research Foundation, 2018). Risk capacity “applies to the objective ability of an investor to take on financial risk” (CFA Institute Research Foundation, 2018, p. 3), while risk appetite “is the amount of risk that one is willing to take in pursuit of reward” (CFA Institute Research Foundation, 2018, p. 21). Given the behavioral nature of this study, we will be focusing on the subjective risk appetite. In our context, this is synonymous with subjective risk tolerance discussed previously in the existing literature.

2.1. Demographic Variables Related to Risk Tolerance

Previous research on the topic suggests relationships between several demographic variables and risk tolerance. There is a recurring theme in research that women are more risk-averse in both risk tolerance questionnaires (Gilliam et al., 2010) and actual investment behavior (Brooks et al., 2019). Fisher and Yao (2017) found that the lower risk tolerance demonstrated by women is not a direct result of the difference in gender but instead results from a lower net worth and a higher income uncertainty present in women.

Chang et al. (2004) found that age had a negative relationship with holding riskier assets, as expected. The same study found that individuals with higher levels of education had a higher objective risk tolerance, as illustrated by their investment portfolio. Findings by Yao et al. (2011) support the findings that age has a negative relationship with risk tolerance. An interesting finding to note on the topic of age is one by J. E. Grable et al. (2009, p. 9), which found that “the youngest working adults tend to over-estimate and that the oldest working adults generally under-estimate their financial risk tolerance”. Investor overconfidence based on age has been studied, and similar results have been supported numerous times (Gervais & Odean, 2001). Among older investors (post-retirement age), financial overconfidence appears to increase, not because of growing confidence but because of declining capability (Pak & Chatterjee, 2016).

Another study by Yao et al. (2005) found that Hispanics and Blacks are more willing than Whites to take substantial financial risk, while Whites are more likely than the other groups to take some financial risk. Similarly, Reiter et al. (2023) found that non-White respondents in the 2019 Survey of Consumer Finances, the 2012 National Financial Capability Study State-by-State survey, and the 2018 National Financial Capability Study Investor Survey were generally more likely to report higher risk tolerance levels than White respondents.

Goetzmann and Kumar (2008) found that individuals with a higher net worth are more diversified in their investments. Investors with a higher income and those who have attained more education show higher investment risk tolerance (J. E. Grable & Rabbani, 2023). Shtudiner (2018) found that self-employed individuals have a higher risk tolerance than employees, as well as a higher future-oriented approach to investing.

2.2. Behavioral Factors Related to Risk Tolerance

J. E. Grable (1997) suggested that demographics are not sufficient independent variables to understand the risk tolerance of an individual. Similar to Kim et al. (2020), this study used future expectations of income and inheritance as independent variables. In developing a risk assessment, J. Grable and Lytton (1999, p. 172) suggested that “a person who is less proactive in earning a gain because of limited information, yet still receives a significant payout (e.g., receiving an inheritance) is generally less risk tolerant”.

Other studies have explored objective financial literacy, financial self-efficacy beliefs, and confidence in the economy in the context of asset ownership. Cupák et al. (2020) found a positive correlation between all three of these independent variables and the ownership of equities. Noman et al. (2023, p. 231) analyzed four different waves of the National Financial Capability Study to find that “investors’ risk tolerance is associated more with their subjective knowledge than their objective knowledge”. A study by Nguyen et al. (2016) found that financial literacy has a significant positive correlation with financial risk tolerance.

3. Theoretical Foundation

3.1. Prospect Theory

The framework of prospect theory is a good theoretical model for interpreting this study (Kahneman & Tversky, 1979). With prospect theory being one of “the best available description of how people evaluate risk in experimental settings” (Barberis, 2013, p. 173), we can use it for a descriptive approach to understand some of the behaviors of investors. Barberis (2013) also mentioned many drawbacks of the theory in economic applications, suggesting that, instead, certain elements of the theory could be applied to various economic models. While expected utility theory (EUT) also models risk aversion—through the curvature of the utility function—prospect theory provides an expanded lens by incorporating empirically grounded psychological behaviors such as reference dependence, loss aversion, and probability weighting. These features allow for a more nuanced understanding of investor behavior that may not align with purely rational decision-making models.

We acknowledge that risk aversion is fundamental to both frameworks. However, prospect theory’s descriptive strengths make it particularly useful for interpreting survey-based measures of risk preferences. In practice, individuals may not evaluate risk using stable utility functions, especially in self-reported contexts. Instead, their responses may reflect heuristics, emotional responses, or contextual interpretations—all of which are core to prospect theory’s contributions.

One such element of prospect theory is risk aversion, which suggests that individuals inherently try to avoid unnecessary risks, a concept easily applicable in the context of investing (Kahneman & Tversky, 1979). The expected utility of increased risk allows for interpreting the findings from the standpoint of an investor responding to the SCF. If a respondent does not anticipate higher returns, prospect theory suggests that they will not accept the increased risk (Levy & Levy, 2021). Similarly, J. E. Grable (2008, p. 5) used the expected utility theory to show that “risk tolerance is conceptually linked with risk-taking behaviors”.

Additionally, prospect theory allows for a lens through which to view different types of asset classes and an investor’s understanding of risk descriptives. The wording and presentation of a situation can influence decisions based on perceived risk (Monteiro & Bressan, 2021), a process called framing. The wording of the risk scale may skew the responses, simply based on how the scale responses are presented or framed to the respondent.

3.2. Asset Allocation in the SCF

Previous research has used the SCF to evaluate the asset allocation of different individuals. Since the risk for the equity asset class is higher than the risk for fixed income (Jacquet, 2021), we expect individuals with higher risk tolerance to hold more equities relative to their fixed income holdings. Consequently, individuals who report higher risk tolerances are expected to hold more equities relative to debt investments. The previously mentioned study by Nguyen et al. (2016) found that risk tolerance had a positive influence on growth asset ownership through their structural analysis model.

4. Methods

4.1. Data

The Survey of Consumer Finances (SCF) 2022 wave was utilized in this study. The SCF is a survey administered every three years by the Federal Reserve to U.S. families. Its focus is on balance sheets, income information, and demographic collection. For missing values, data are imputed in five ways by the Federal Reserve and included in the survey data. This survey is administered by a representative from the NORC at the University of Chicago either in person or over the phone. The data collectors attempt to choose random households across all economic strata, although it is voluntary, so certain types of households may be underrepresented.

4.2. Analysis

The risk tolerance question that has been present in the SCF since 1983 consists of four possible answers for respondents when asked how willing they are to take financial risks. The other risk question is presented the same, but the answer options are on a 0–10 scale, with 0 being “Not at all willing to take financial risks” and 10 being “Very willing to take risks”. The coding of these answers differed slightly from the raw data for the purpose of analyzing the data. The four-point scale was recoded to 0–3, with 0 as “Not at all willing to take financial risks”. The 11-point scale was coded similarly, from 0 to 10. This allowed for the baseline of “no risk” to be 0 on both scales.

Table 1. The 4-point and 11-point scale distribution table.

4-Point Scale			11-Point Scale
Response	n	%	Response	n	%
Not willing to take any financial risks	1459.2	13.44%	0	617.6	13.44%
			1	144.8	3.15%
			2	318.2	6.92%
Take average financial risks expected to earn average returns	1839.8	13.44%	3	408.6	8.89%
			4	368.2	8.01%
			5	880.6	19.16%
Take above-average financial risks expected to earn above-average returns	1051.8	13.44%	6	510.4	11.11%
			7	559.4	12.17%
			8	403	8.77%
Take substantial financial risks expected to earn substantial returns	244.2	13.44%	9	117.4	2.55%
			10	266.8	5.81%
n = 4595

shows the distribution for both scales.

All regressions were run without survey weighting. Due to the multiple imputations inherent in the SCF, all data were divided by five for analysis. Repeated imputation inference (RII) was used to account for the five replicates in the SCF (Montalto & Sung, 1996). We acknowledge the SCF’s complex survey design and the availability of sample weights to adjust for potential nonresponse and oversampling. However, we elected not to apply survey weights in our regression analyses for two main reasons. First, our study aimed to assess the internal relationships between self-reported risk tolerance measures and asset ownership within the SCF sample rather than to produce nationally representative point estimates. As such, unweighted regression analyses can yield unbiased estimates of these relationships under the assumption that the model is correctly specified. Second, recent methodological research has shown that including survey weights in regression models does not always improve estimation efficiency and may even reduce statistical power, particularly when the weights are not strongly correlated with the dependent variables (Winship & Radbill, 1994; Pfeffermann, 1993). Following precedents in similar SCF-based behavioral finance studies (e.g., Kim et al., 2020; J. E. Grable & Rabbani, 2023), we opted for unweighted regression models while still accounting for the SCF’s multiple imputation structure using repeated imputation inference (Montalto & Sung, 1996).

4.3. Variable Creation

Two composite variables were created, shown in the table below. These were grouped together in the same asset class for the purpose of analysis. If the particular asset class was owned, it was assigned a 1. If a respondent did not indicate that they owned that asset class, it was assigned a 0. There are more asset classes available in the SCF, but these two were the focus of this study. In

Table 2. New variables.

Variable Name	Asset Class	SCF Variables Included
Bonds_Held	Fixed Income	TFBMUTF + GBMUTF + OBMUTF + SAVBND + BOND
Stocks_Held	Equity	STMUTF + STOCKS

, all variables in the column labeled “SCF Variables Included” are the variables created for the public dataset. The fixed income variables referred to ownership of mutual funds focused on tax-free bonds, government bonds, or other bonds, as well as direct ownership of savings bonds or bonds. The equity ownership variables included both stock mutual funds and direct ownership of stocks themselves. These variables only captured ownership in non-qualified investment accounts, due to the structure of the questions and the SCF itself.

4.4. Multinomial Logit Model (The 4-Point Risk Tolerance Scale)

Let Ri ∈ {0,1,2,3} represent the respondent’s self-reported risk tolerance on the 4-point scale, where Ri = 0 corresponds to “not at all willing to take financial risks”. The probability of individual iii selecting risk level j ∈ {1,2,3}, relative to the base category j = 0, is estimated using a multinomial logit model:

$$\mathit{Pr}\left(R_i=j\right)={\displaystyle \frac{\mathit{exp}\left(X_i^{\prime }{\beta }_j\right)}{1+{\sum }_{k=1}^3\mathit{exp}\left(X_i^{\prime }{\beta }_k\right)}}\mathrm{for}j=\mathrm{1,2},3$$

$$\mathit{Pr}\left(R_i=0\right)={\displaystyle \frac{1}{1+{\sum }_{k=1}^3\mathit{exp}\left(X_i^{\prime }{\beta }_k\right)}}$$

where

• Xi is a vector of covariates for individual i (e.g., age, gender, education, income, subjective/objective financial knowledge, employment status, economic expectations);
• βj is the parameter vector corresponding to risk level j;
• The model reports relative risk ratios (RRRs) for ease of interpretation.

4.5. Ordinary Least Squares (OLS) Model (The 11-Point Risk Tolerance Scale)

Let Ri(11) ∈ [0, 10] represent the respondent’s risk tolerance on the 11-point scale. Given the quasi-continuous nature of this variable, we estimate an OLS regression model as follows:

$$R_i^{\left(11\right)}=\mathsf{\alpha }+X_i^{\prime }\mathsf{\gamma }+{\epsilon }_i$$

where

• α is the intercept term;
• Xi is the same vector of independent variables as in the multinomial logit model;
• γ is the vector of regression coefficients;
• εi is an idiosyncratic error term assumed to be normally distributed with mean zero and constant variance.

Additionally, tax-deferred account composition was added to the non-qualified investment account composition in the new Bonds_Held and Stocks_Held composite variables. This was modeled after Bergstresser and Poterba (2004), who created an asset allocation model of fixed income and equity based on the SCF question on composition of tax-deferred accounts. They combined this information with the non-tax-deferred account questions, as captured by the variables mentioned above (2004). If a respondent answered that they held only fixed income or equity assets in their tax-deferred accounts, the respective variable from Table 2 would be assigned a 1, while the other would be assigned a 0. If the respondent answered saying they held a combination of both, both variables were assigned a 1. Similarly, if a respondent answered the separate question stating that they held combination mutual funds (meaning a fund consisting of both stocks and bonds), both would be assigned a 1.

For any binary variables, results were recoded as 0 or 1. Those variables were as follows: household type (couple = 0, single = 1), gender (male = 0, female = 1), homeownership (no = 0, yes = 1), and expectation for a substantial inheritance (no = 0, yes = 1). Other variables were coded as outlined in the SCF codebook, each with its own multiple-choice response or Likert-type scale. These variables included racial/ethnic status, health status, employment status, current income relative to normal income, and expectation of future income. Most of these demographic variables were chosen because both Kim et al. (2020) and Kuzniak et al. (2015) performed similar studies using these variables. Both studies compared the 4-point SCF scale to another risk scale and used similar variables.

Additionally, we looked at economic expectations and financial knowledge as explanatory variables. Specifically, we looked at the respondents’ 1-year and 5-year expectations of the economy to improve, worsen, or remain the same. For subjective financial knowledge, we performed a similar recoding to the new risk scale to look at subjective financial knowledge, ranking it as 0–10. Objective financial knowledge comprised the three basic financial knowledge questions known as the “big three” (Lusardi, 2019). Both objective and subjective financial knowledge were standardized.

For part 1 of this study, we sought to measure the aforementioned independent variables (demographics, socioeconomics, and confidence variables) against the two SCF risk scales. Two separate empirical methods were used for part 1. The first regression on the 4-point risk scale was run using multinomial logistic regression, with the base category being “no risk”. Since this scale did not appear to have equal spacing between results, it would be inaccurate to process it as a continuous variable instead of a categorical variable. The reported results included the relative risk ratio, or RRR. An ordinary least squares regression was run for the 11-point scale in part 1, as this scale can be viewed as continuous.

For part 2 of the study, we sought to measure the correlation between the risk scales and the asset ownership variables to see which subjective risk tolerance measure had a stronger relationship with the objective risk tolerance of a respondent. All four possible combinations of the two risk scales and two asset classes were analyzed using logistic regressions. For this part, we used the risk scales as independent variables, with the asset class ownership as the dependent variable.

5. Results

5.1. Part 1

In this first regression for the old four-point risk scale, shown in

Table 3. Multinomial logistic regression of 4-point risk category, with no risk as the reference group.

Variable (N = 4595)	Average Risk		Above-Average Risk		Substantial Risk
	RRR	S.E.	RRR	S.E.	RRR	S.E.
Constant	0.412	0.197 *	0.195	0.114 ***	0.222	0.183 *
Respondent Age	0.980	0.004 ***	0.959	0.004 ***	0.965	0.007 ***
Education Level (ref: Some High School)
High School Graduate	1.594	0.283 ***	1.949	0.549 **	1.067	0.31
Some College	2.630	0.459 ***	3.155	0.875 ***	0.861	0.263
Bachelor’s Degree	4.548	0.835 ***	6.892	1.929 ***	1.748	0.536 *
Graduate Degree	6.533	1.286 ***	9.026	2.625 ***	2.361	0.769 ***
Sex (ref: Male)
Female	0.667	0.058 ***	0.489	0.052 ***	0.434	0.072 ***
Marital Status (ref: Single)
Married	1.011	0.096	1.077	0.127	1.387	0.245 *
Race (ref: White)
Black	0.653	0.081 ***	0.746	0.121 *	1.242	0.292
Hispanic	0.356	0.051 ***	0.456	0.081 ***	0.786	0.197
Asian/Other	0.649	0.078 ***	0.608	0.087 ***	1.077	0.225
Children Living at Home (ref: None)
Children Living at Home	0.979	0.101	1.006	0.121	1.134	0.2
Employment Status (ref: Employed)
Self-employed	1.739	0.231 ***	2.243	0.328 ***	3.864	0.767 ***
Not working	0.807	0.168	0.841	0.218	1.242	0.418
Retired	0.799	0.099 *	0.797	0.124	1.330	0.334
Home Ownership (ref: Do Not Own)
Own a Home	1.280	0.145 **	1.307	0.182 *	0.814	0.165
Log of Income	1.163	0.032 ***	1.224	0.039 ***	1.215	0.059 ***
Log of Net Worth	1.036	0.01 ***	1.092	0.016 ***	1.050	0.02 **
Health Status (ref: Excellent)
Good Health	1.025	0.112	0.840	0.102	0.494	0.086 ***
Fair Health	0.896	0.113	0.692	0.105 **	0.449	0.1 ***
Poor Health	0.614	0.135 **	0.745	0.208	0.549	0.217
Inheritance Expectation (ref: Do Not Expect a Substantial Inheritance)
Expect a Substantial Inheritance	1.547	0.217 ***	1.661	0.253 ***	1.655	0.352 **
Current Income Relative to Normal Income (ref: Higher)
Same	0.682	0.095 ***	0.619	0.096 ***	0.473	0.102 ***
Lower	0.637	0.107 ***	0.746	0.141	0.648	0.168 *
Future Income v. Inflation (ref: Income Will Grow More)
It Will Be Less	0.993	0.122	0.879	0.122	0.699	0.132 *
It Will Be Same	0.925	0.112	0.674	0.094 ***	0.437	0.087 ***
1-Year Economic Expectation (ref: Improve)
Worsen	0.768	0.107 *	0.704	0.115 **	0.852	0.21
Same	0.810	0.103 *	0.677	0.104 **	0.860	0.197
5-Year Economic Expectation (ref: Improve)
Worsen	0.863	0.097	0.899	0.119	0.976	0.19
Same	1.098	0.123	1.183	0.157	0.794	0.168
Subjective Fin. Knowledge	1.080	0.046 *	1.282	0.074 ***	1.226	0.102 **
Objective Fin. Knowledge	1.230	0.054 ***	1.267	0.07 ***	1.158	0.095 *

*** p < 0.01, ** p < 0.05, * p < 0.1; R-square = 0.172.

, we can see that the employment status of self-employed is significant in increasing the odds that an individual will take more risk in each category relative to the baseline of “no risk”. We also see a significant increase in the odds for each increasing risk category for those who expect a significant inheritance and those who have received a college degree. We see a significant decrease in the odds that an individual will take more financial risk in females and those whose current income is similar to their normal income. Net worth and income both have a positive correlation with the odds of an individual taking more financial risk, while age has a negative correlation. The multinomial logistic regression revealed a pseudo-R-squared value of 0.172.

One interesting finding is that there is a greater increase in the odds of an individual taking substantial financial risk instead of no risk when looking at subjective financial knowledge than when looking at objective financial knowledge. However, objective financial knowledge increases the odds more than subjective financial knowledge that an individual will take average financial risk instead of no risk. Another notable finding is that an individual’s 1-year economic outlook has a significant negative impact on the odds of an individual taking an average or above-average risk, while their 5-year economic outlook has no significant impact.

Table 4. OLS regression of 11-point risk scale.

Variable (N = 4595)	Coefficient	S.E.	Variable (N = 4595)	Coefficient	S.E.
Constant	4.745	0.428 ***	Log of Income	0.110	0.024 ***
Respondent Age	−0.027	0.003 ***	Log of Net Worth	0.069	0.01 ***
Education Level (ref: Some High School)			Health Status (ref: Excellent)
High School Graduate	0.192	0.158	Good Health	−0.292	0.091 ***
Some College	0.295	0.157 *	Fair Health	−0.532	0.114 ***
Bachelor’s Degree	0.761	0.164 ***	Poor Health	−0.731	0.198 ***
Graduate Degree	0.877	0.17 ***	Inheritance Expectation (ref: Do Not Expect a Substantial Inheritance)
Sex (ref: Male)			Expect a Substantial Inheritance	0.233	0.108 **
Female	−0.799	0.081 ***	Current Income Relative to Normal Income (ref: Higher)
Marital Status (ref: Single)			Same	−0.494	0.112 ***
Married	0.104	0.09	Lower	−0.227	0.14
Race (ref: White)			Future Income v. Inflation (ref: Income Will Grow More)
Black	0.292	0.123 **	It Will Be Less	−0.313	0.104 ***
Hispanic	−0.086	0.133	It Will Be Same	−0.465	0.105 ***
Asian/Other	−0.034	0.109	1-Year Economic Expectation (ref: Improve)
Children Living at Home (ref: None)			Worsen	−0.350	0.124 ***
Children Living at Home	−0.060	0.092	Same	−0.304	0.116 ***
Employment Status (ref: Employed)			5-Year Economic Expectation (ref: Improve)
Self-employed	0.772	0.105 ***	Worsen	0.028	0.101
Not working	0.337	0.195 *	Same	0.074	0.1
Retired	−0.093	0.117	Subjective Fin. Knowledge	0.427	0.041 ***
Home Ownership (ref: Do Not Own)			Objective Fin. Knowledge	0.099	0.041 **
Own a Home	0.113	0.106

*** p < 0.01, ** p < 0.05, * p < 0.1; R-square = 0.253.

shows the results of the OLS regression of the newer 11-point scale.

Similar to the four-point scale, we see that being a female, having the same current income relative to normal income, and respondent age all have negative correlations with increased financial risk tolerance. We also see a positive correlation between being self-employed, being a college graduate, income, and net worth, and increased financial risk tolerance. The same interesting finding is reflected in the significant negative relationship observed between 1-year economic expectations and financial risk tolerance, but no significant relationship between 5-year economic outlook and financial risk tolerance. In this regression, we see that subjective financial knowledge has a higher correlation with increased financial risk tolerance than objective financial knowledge. The linear regression revealed an R-squared value of 0.253. Multicollinearity checks were run, and no issues were found.

5.2. Part 2

For part 2, we see significance in the relationship between ownership of both asset classes and the 4-point and 11-point scales.

Table 5. The summary table of 4-point risk scale logistic regressions on asset classes.

	Asset Class
	Fixed Income			Equity
Variable	Odds Ratio	S.E.	p > |z|	Odds Ratio	S.E.	p > |z|
4-point Risk Scale	1.208	0.068	0.001	1.664	0.088	0.000
Pseudo-R2	0.304			0.378

shows the two regressions run using the four-point risk scale as an independent variable. The results show that a one-category increase in risk tolerance on the four-point scale is associated with a 20.8% increase in the odds of an individual owning fixed income and a 66.4% increase in the odds of owning equity. As equities are a riskier asset class, this result is not surprising, and it is consistent throughout all findings in part 2.

Table 6. The summary table of 11-point risk scale regressions on asset classes.

	Asset Class
	Fixed Income			Equity
Variable	Odds Ratio	S.E.	p > |z|	Odds Ratio	S.E.	p > |z|
11-point Risk Scale	1.046	0.019	0.014	1.166	0.020	0.000
Pseudo-R2	0.303			0.376

shows the two regressions run using the 11-point risk scale as an independent variable. The results show that a 1-point increase in risk tolerance on the 11-point scale is associated with a 4.6% increase in the odds of an individual owning fixed income and a 16.6% increase in the odds of owning equity.

Table 7. The summary table of standardized 4-point risk scale logistic regressions on asset classes.

	Asset Class
	Fixed Income			Equity
Variable	Odds Ratio	S.E.	p > |z|	Odds Ratio	S.E.	p > |z|
4-point Risk Scale	1.179	0.058	0.001	1.558	0.072	0.000
Pseudo R2	0.304			0.378

and

Table 8. Summary table of standardized 11-point risk scale regressions on asset classes.

	Asset Class
	Fixed Income			Equity
Variable	Odds Ratio	S.E.	p > |z|	Odds Ratio	S.E.	p > |z|
11-point Risk Scale	1.135	0.058	0.014	1.547	0.075	0.000
Pseudo R2	0.303			0.376

show the same regressions from part 2 run after the scales were standardized. All results are significant, but we see that the 11-point scale is associated with lower odds increases in asset ownership for both asset classes tested. The odds of fixed income ownership increase 17.9% for each one standard deviation increase in risk for the 4-point scale, but only 13.5% for a one standard deviation increase in the 11-point scale. The odds of equity ownership increase by 55.8% for each one standard deviation increase in risk for the 4-point scale, but only 54.7% for a one standard deviation increase in the 11-point scale. For the four-point scale, this means the spread between fixed income and equity ownership odds is 37.9% (55.8–17.9%). This spread for the 11-point scale is 41.2% (54.7–13.5%).

In the context of prospect theory, we can see the impact of framing for both risk scales. Since equities carry more risk than fixed income, prospect theory’s assumption of loss aversion would suggest that those averse to risk would be less likely to own risky assets. The more drastic odds ratio of equity ownership vs. fixed income ownership suggests that the 11-point scale may better reflect the perception that equities are riskier investments compared to fixed income. The larger spread between fixed income and equity ownership odds on the 11-point scale means that a one standard deviation increase leads to higher odds of holding riskier assets relative to lower-risk assets. This suggests that when using the 11-point scale, respondents were able to provide a more accurate spread of responses to capture the risk tolerance of their portfolios. With 11 options instead of 4, the newer risk scale can inherently be more accurate.

An individual’s understanding of portfolio risk mapped to a risk scale rating may be a product of many variables, although loss aversion likely plays a large role. The 11-point risk scale seems to more accurately capture that increased risk and equity ownership are often closely linked within investment portfolios. As potential losses impact behavior and perceptions more than the proportionate potential gains, the wording of the four-point scale likely skews respondents’ responses to be more conservative, trying to interpret the question in terms of loss aversion. Similarly, respondents who are looking to take below-average risk may opt for the “no risk” option instead of the “average risk” option if they are less likely to choose extra risk when presented with these options. This is a limitation of the risk scale that could be mitigated by adding a fifth category and increasing the scale’s accuracy to match investor behavior (asset allocation).

6. Limitations

Risk tolerance is inherently a difficult topic to research due to its subjective nature. Many individuals may view their risk tolerance as unwavering, while others may view their risk tolerance as a product of market conditions. This difference may lead to inconsistent results when surveying individuals on their risk preferences. An individual’s view on the risk of various asset classes may also create difficulty in surveying financial risk information, as they may not understand that their investment portfolio carries the level of risk that it does. Although the subjectivity of risk questionnaires is a limitation, it also demonstrates the importance of this study. This limitation is more a limitation of risk questionnaires, not necessarily the study, as this study sought to measure the subjectivity of these questionnaires compared to the less-subjective nature of the investments’ risk profile.

There is no granular breakdown of asset allocation available inside qualified retirement accounts in the SCF. The questions presented in the study gather data regarding stock and bond ownership within the retirement accounts, but not an exact breakdown. Information was also collected regarding combination funds, such as target date funds and balanced funds, but the details of the holdings were not collected, meaning that it is unknown if the funds include stocks, bonds, alternatives, real estate, or other holdings. For this reason, these funds were excluded. Trust accounts were also excluded, as they have the ability to hold any type of investment, and a trust account itself is not associated with a certain level of risk. To strengthen this study, an accurate breakdown of the underlying holdings within retirement and trust accounts, including the breakdown of combination funds, would have been useful. Given the focus of this study being stock and bond allocation, another limitation is the exclusion of alternative asset classes, such as real estate and cryptocurrencies. Future studies could include these asset classes, especially with the rise in ownership of some of these assets.

In financial planning, risk measurement tools vary greatly among organizations. While many tools are compliant, they are not consistent with each other. A risk profiling tool from Company A may assign a drastically different risk rating to an investment than a similar tool from Company B. While internally consistent when used to measure both client and investment risk profiles, the variation in measurement makes risk comparison between organizations difficult. This lack of consistent risk measurement between firms presents both a limitation to this study and an opportunity for future research in academic and practitioner settings. While this study examined the associations between risk tolerance measures and investment behavior, it did not aim to make causal claims. To continue this line of research, future studies employing longitudinal data or instrumental variable techniques may help to establish causal pathways. Another key limitation of this study is the inability to address potential endogeneity. While this study examined the associations between risk tolerance measures (i.e., the original 4-point and 11-point scales) and investment behavior, it did not aim to make causal claims. Consequently, findings should be interpreted with caution. To continue this line of research, future studies should expand upon this foundational work by employing longitudinal data or appropriate causal identification methods, such as instrumental variable techniques, to facilitate a more robust exploration and establish causal pathways between risk tolerance and investment behaviors.

The absence of risk capacity in measuring the results is limited in this study. The purpose of this study was not to measure risk tolerance questionnaires against an individual’s risk capacity, but instead to see which of the two risk tolerance questionnaires most closely matched the portfolio. As this study compared the two measures internally, the drawbacks of omitting the risk capacity of an individual were mitigated. However, for a holistic approach to financial planning, both risk capacity and subjective risk tolerance need to be incorporated. Future studies could use risk capacity as an additional variable in the study to strengthen the findings and provide additional applicability to practitioners.

While the 11-point SCF risk tolerance scale offers greater granularity and shows stronger correlations with asset ownership, we acknowledge that such detailed scales carry potential drawbacks. Research by Krosnick (1991) cautions that scales exceeding seven points may increase cognitive burden, potentially leading to higher non-response rates or random error. Additionally, heaping bias—where respondents favor rounded values like 0, 5, or 10—can reduce measurement precision. Our data show some clustering consistent with this effect. These concerns suggest that the improved performance of the 11-point scale may not solely reflect better construct validity but could also be shaped by response behavior. Future research should examine these patterns more systematically, including item-level response distributions and potential endpoint effects.

7. Implications

Both scales appear to capture risk tolerance based on asset class ownership effectively. However, the 11-point scale better captures the disparity in risk between asset classes, suggesting that it is a more accurate reflection of an individual’s risk. This can likely be attributed to the more granular nature of the scale and the wording of the two scales, as interpreted using prospect theory. While both types of scales are used in research and practice, using a scale with more options that is interpreted through quantitative options instead of qualitative options may allow respondents to reflect their actual risk tolerance more accurately.

This research can be applied to practitioners and future academic research. The conversation regarding risk tolerance for practitioners is ongoing due to regularly changing regulations. Creating more specific and accurate risk tolerance questionnaires for clients helps practitioners to not only stay compliant but also to create a better investment portfolio for the clients. Knowing how to guide this conversation with clients both on paper and in meetings ultimately makes a stronger client relationship and allows for the client to be more pleased with their investment performance, regardless of short-term market performance.

Academic research can utilize this research for future studies regarding risk tolerance, both in the fields of economics and personal finance. This has implications for human behavior and investment performance, allowing future studies to use traditional economic theories and behavioral-focused theories. Additional research questions related to capturing risk tolerance of an individual or portfolio, accuracy of risk tolerance based on financial education, and ownership of additional asset classes beyond equities and fixed income would be good next studies for academic research using this study.

We acknowledge that using asset ownership as the primary behavioral benchmark for validating risk tolerance measures presents important limitations. While individuals with higher risk tolerance are generally more likely to hold riskier assets, asset ownership is ultimately a behavioral outcome shaped by more than just risk preferences. Financial capacity, liquidity constraints, time horizons, access to information, and prevailing market conditions also play critical roles in shaping portfolio choices. Therefore, a stronger correlation between a risk measure and asset ownership does not definitively indicate superior measurement of pure risk tolerance. It may instead reflect a broader set of financial circumstances. Prior work, including Kim et al. (2020), also cautions against overreliance on asset ownership for criterion validation and finds inconsistencies, such as decreased ownership levels at the highest reported risk tolerance categories. Future research should incorporate complementary validation strategies, such as experimental risk elicitation, longitudinal investment behavior, or triangulation with alternative psychometric instruments, to more rigorously evaluate and compare the predictive and construct validity of subjective risk tolerance measures.