Brazilian Portuguese version of financial risk-taking and tolerance scales: validity evidence within and between measures

Objectives: To construct and validate a psychological measure called the Financial Risk-Taking Scale (FRTakS) and to translate, adapt, and validate a psychological measure called the Financial Risk Tolerance Scale (FRTolS) with a Brazilian sample. Exploratory and confirmatory factor analyses were used to assess evidence of the validity of the scales’ internal structures. We also tested the convergent validity between FRTakS and FRTolS. Method: After construction (FRTakS) and adaption (FRTolS), the instruments were evaluated by expert judges for the relevance of their items to the scales, followed by pretesting. A cross-sectional study was then conducted using a convenience sample of 834 people who responded to invitations sent to a mailing list or to an online invitation on the Brazilian Securities and Exchange Commission website (Comissão de Valores Mobiliários [CVM]). Results: Mean age of participants was 39.27 years (standard deviation [SD] = 10.82), they had high educational level (60.9% post-graduate), were married or living together (60%), and their spending power was 41.36 (SD = 13.27). Exploratory and confirmatory analysis identified two factors in FRTakS (Investment and Spending Money), both with 4 items; and identified a single factor in FRTolS, comprising 7 items. Conclusion: Reliability indexes for the goodness of fit of the factor structure were satisfactory. There was a positive and significant correlation between the FRTakS Investment factor and FRTolS, confirming convergent validity. The results suggest the existence of a two-dimensional factor structure for FRTakS, and a one-dimensional factor structure for FRTolS. The instruments also exhibited convergent validity with each other.


Introduction
The perspective on risk-taking in finance considers it to be an individual inclination, in a given context, to choose risky options over safer ones. The risky option is defined by its multiple possible outcomes while the safe option is defined by a single predicted outcome. 1 Within the aspect of individual inclination, previous researches have pointed out that individual financial risk tolerance is an important variable that influences financial decision-making, especially in the context of uncertainty. Financial risk tolerance is related to how much a person can bear possible losses or to the volatility of a choice. 2 Risk-taking and risk tolerance are directly and positively associated constructs, since risk tolerant individuals tend to engage in more risky behaviors. 3 A theoretical model proposed by Grable 4 predicts that risk tolerance works as a mediator between risk profile, risk perception, risk need, and choice of risky behavior.
In general, there is an ongoing discussion on the difficulties of studying risk-taking, since there is no agreement on standardizing instruments and establishing convergent validity between them. One reason for this difficulty is that risk-taking is a multideterminant phenomenon that is influenced by risk tolerance, the context in which decisions are taken, personality traits, and sociodemographic data such as gender and age. 3,[5][6][7] Use of scales to measure risk-taking has been expanding, with a domain-specific approach that presupposes multiple dimensions of risk related to different contexts, rather than using generalized risktaking scores. Researchers have tried to adapt the concept of the Domain-specific Risk-Taking Scale (DOSPERT) 8 for the Brazilian culture. 6,7 The DOSPERT is a North American instrument that aims to measure risk-taking as risk attitudes, defined as how much an individual would engage in risky behavior; and perceived-risk attitudes, defined as the willingness to engage in a risky activity as a function of its perceived riskiness. The two parts of the scale have 40 and 30 items. Participants rate on 5-point or 7-point Likert scales the extent to which they perceive an item to be risky (from "Not at all risky" to "Extremely risky") and how likely they would be to engage in the activity or behavior (from "Extremely unlikely" to "Extremely likely"). 9 The original scale covered five factors identified based on a review of the literature on risk-taking behaviors: ethical, financial, health/safety, recreation, and social (see the original paper 8 for descriptions of each factor).
The financial factor is subdivided into gambling and investment items. In the original version, 8  and confirmatory factor analysis (CFA) were used to assess evidence of the validity of the internal structure of both scales. We also tested the validity of measures analyzing convergent validity between FRTakS and FRTolS, hypothesizing a positive association. 3 Additionally, the hypothesis is that these instruments will have satisfactory indicators of internal validity and that there will be convergence between them.
Regarding the factor structures of these scales, we do not expect to find the same structure as instruments administered to samples from developed countries, because country conditions may have an influence on risk. 2,13,14 In a study 13

Sociodemographic data
A sociodemographic questionnaire was used in all steps of the research and included questions addressing sex, age, schooling, marital status, and spending power measured using the Brazilian Economic Criteria. 15 We used these criteria as a continuous measure on which the participants could score from 0 to 100 points, depending on their answers; the higher the score the higher the consumption of goods and the educational level of the participant/family.

Risk-taking
The initial instrument consisted of 18 items, originating from the following scales: 1) Eight items from the DOSPERT 9 financial factor, translated into Portuguese. 16 2) Four items from the Specific Risk Propensity Scale. 6 3) We also used three items developed by the authors, but these were excluded after statistical analysis, considering our objective of constructing a new instrument.

3) Three items from the Specific Risk Propensity
Scale -Evolutionary Domains. 7 As mentioned, the factor structure of this instrument did not include a specific factor for financial risk-taking, but we selected three items related to financial topics, which were part of the Competition/Fertility factor. Three items had 75% agreement and were retained. structure of the instruments was tested in part of the sample using EFA and reliability indexes were computed for the items of the resulting subscales. Aiming to verify the factorial adequacy of the data, we performed an unweighted least squares (ULS) factor analysis for both instruments using the oblique rotation method (PROMIN), which assumes a correlation between the resulting factors. 17 We also used the Hull method to attempt to find alternative models, based on balance between model fit and number of parameters. 18 Internal consistency was calculated as the α reliability indices for each of the independent factors.
CFA was conducted using the second part of the database. The method of estimation to verify goodnessof-fit for the models proposed by CFA was weighted least squares mean-and variance-adjusted (WLSMV), designed specifically for non-continuous data that may not be multivariate normal. 17 (Table 1) were named as follows: Spending Money (Factor 1), and Investments (Factor 2).
The second database was analyzed with CFA.
The model being tested was the factor structure that emerged from the exploratory step.
The initial model did not achieve satisfactory adequacy As shown in Figure 1, the correlation between factors was r = 0.001. The reliability indexes were similar to those found for the first sample: α = 0.25 for Factor 1; α = 0.58 for Factor 2. It is important to stress that cross-load removal improved the α indexes, making them achieve 0.61 for each factor. However, without cross loads, the model parameters are compromised.

Psychometric properties analysis for the FRTolS
Hull analysis suggested a one-dimensional solution.
In the EFA (n = 417) with one factor extraction, four  Table 1 for statistical indices).  Greater risk tolerance was associated with greater risk-taking in spending money. Furthermore, when the instruments were analyzed together, the direction of the relationship between risk factors changed ( Figures   1 and 3).

Discussion
EFA and CFA showed evidence of the validity of the internal structures of both instruments. However, the factor structures were different from the initial proposals.
Structure differences may have occurred because country conditions can widely influence risk taking and the risk tolerance. 2,13 The different structures of the risk-taking scale and the risk tolerance scale in different countries are likely attributable to the facts that an individual's developmental history in different parts of    19 We must consider that the robustness of the In quantitative terms, FRTakS followed the pattern of reliability indexes shown in previous research with risk-taking scales in Brazilian samples 6 and had slightly weaker reliability indexes than presented in studies with the original scale. 8,9 This reliability contributes to restricting applications for use of the test. 19

Conclusion
Studies of how people make decisions and the risk they tolerate and are willing to take in these decisions allow us to support strategies or programs designed to guide a decision-making process that suits the individual, institutional, and social levels of the decision-maker. 4 We conclude that these instruments have presented good evidence of validity for use with Brazilian samples.
The results suggest the existence of a two-dimensional factor structure for FRTakS and a one-dimensional factorial structure for FRTolS, and showed convergent validity between the two scales.