Framing the Retirement Planning Behavior Model towards Sustainable Wellbeing among Youth: The Moderating E ﬀ ect of Public Proﬁles

: This study examines the e ﬀ ect of ﬁnancial literacy, saving attitudes, social inﬂuence, and goal clarity on the retirement planning construct. In addition, it investigates how the public demographic proﬁle moderates these relationships. The questionnaire approach was utilized to collect data by adopting and customizing the measurement scale from previous studies. A systematic random sampling approach was employed on 323 prospective respondents. The outcomes of this study illustrate that all relationships are signiﬁcantly and positively associated with retirement planning using structural equation modeling (SEM). Furthermore, all moderator variables (gender, age, status, income, and education) moderated the relationships. The government should construct a holistic retirement planning model that is based on demographic characteristics. (H5c). Age moderates the relationships between saving attitude and retirement planning . Hypothesis 5d (H5d). Age moderates the relationships between goal clarity and retirement planning


Introduction
The Department of Statistics Malaysia announced that the number of employed people has been increasing from 5.2 million in 1982 to 14.99 million in June 2020 [1]. If this situation is maintained, the number of preretirees will increase in the future. The reality hits when most locals admit that they do not have sufficient savings for their retirement [2][3][4]. This financial problem is not only happening in Malaysia but also in other developed and developing countries. For example, the people in the United States rely on self-directed investment accounts [5]. Half of the retirement assets are independently deposited in those accounts [6,7]. Some employees appoint financial experts or consultants to allocate their savings into retirement accounts, but most of them make their own decisions. The majority of the employees do not know about financial management, which has led to the loss of savings because of nonperforming financial instruments.
In Malaysia, there are two most notable retirement schemes, known as the public pension scheme and the Employees Provident Fund (EPF). As the names suggest, the former scheme is provided for government servants only, based on a defined benefit plan, while the latter mainly caters to the private sector workforce and also those in the public sector who opt for the scheme after they are confirmed in their work positions; these employees comprise more than half of the total labor force in Malaysia. The EPF is a mandatory and defined contribution retirement savings scheme in Malaysia. Both employee affirmative action before their retirement. Therefore, this study will provide a better understanding of the preparedness of working individuals below 40 years old when facing their retirement age in the future. Specifically, we have nine general hypotheses that consist of four exogenous constructs (financial literacy, social influence, saving attitude, and goal clarity) and 5 moderator variables (gender, age, income, marital status, and education). These effects are explained in detail in the following section.

Literature Review
The main challenge faced by aged Malaysians recently is the shortage of savings upon retirement. Therefore, it is important to prepare them with knowledge and awareness about saving for sustainable wellbeing in the future. The terms of sustainability are actually very broad across different areas such as human, social, economic, and environmental. Sometimes, researchers define sustainable wellbeing as sustainable happiness, which means the happiness that contributes to an individual, a community, or the world and does not exploit the happiness of others. Consequently, a critical discussion on this topic is suggested and how sustainability can be related to society. Aligned with this reason, "society" in the current study refers to the employees from the private sector and government companies in order to understand how they will sustain their living after retirement. This should include their lifestyle, food and beverage cost, and healthcare, given their age requirement. The symbolic interaction theory is very relevant to the current study as it aims to assess retirement planning behavior [23]. From that, a model has been modified by the inclusion of several important factors, namely, financial literacy and social influence as exogenous constructs, with public profiles considered as a moderator variable to provide more comprehensive findings. Each factor is discussed in the following section in detail.

Financial Literacy
Financial literacy is the ability to use knowledge and skills to manage financial resources effectively for a lifetime in terms of financial wellbeing [24]. Most previous studies have offered several insights into the reasons for not planning for retirement, which can help people in the future. The findings revealed that working individuals failed to develop any retirement savings plan [25,26]. The primary reason for this poor planning is financial illiteracy. Furthermore, most of them are unaware of fundamental economic concepts during their lifetime and old age. To gain better insight into this particular issue, Lusardi and Mitchell [27] found that a lack of confidence can lead a working individual to make a poor plan. This evidence is proven by Wong and Earl [21] and Kim, Kwon, and Anderson [11] regarding the confidence of an individual's retirement.
Hypothesis 1 (H1). Financial literacy has significant positive effects on retirement planning.

Social Influence
Beshears et al. [28] affirmed that the presence of peer information can influence the working individual's decisions on retirement savings. People can get information and experience from others who have the potential to influence their decisions. Moorthy et al. [3] and Van Dalen [29] pointed out that parental effects and social influence have significant positive impacts on retirement planning. Growing empirical literature has revealed that peoplewith better social network tend to invest their savings [30][31][32][33]. Many studies have revealed that peer information can cause some individuals to become discouraged in contributing more to their retirement savings. Hence, this study intends to prove that peer effects can be another influence on people's retirement planning decisions.
Hypothesis 2 (H2). Social influences have a positive significant effect on retirement planning.

Saving Attitude
The majority of working individuals have trusted the EPF to decide on what or where to invest their contributions in, as long as their savings increase every year [4,34]. Some people are unwilling to face the complexity and difficulty of the investment system, and they are passive in making their investment choices [35]. People tend to get low investment returns since the EPF usually invests in safe investment options. Due to this fact, their retirement income is insufficient to cover their living maintenance for their golden years.

Hypothesis 3 (H3)
. Saving attitude has a positive significant effect on retirement planning.

Goal Clarity
Retirement goal clarity is another psychological factor associated with planning practice in predicting saving tendencies [36]. Theoretically, six psychological scales were initially introduced in measuring saving strength, such as general self-efficacy, future time perspective, financial activation, retirement goal clarity, self-rated financial knowledge, and financial risk tolerance [37]. These factors are assessed to determine the clearness of working individuals in financial goals for retirement, which is highly associated with retirement saving behavior [38,39]. Nonetheless, retirement goal clarity is usually adopted in various disciplines [40,41].

Hypothesis 4 (H4)
. Goal clarity has a positive significant effect on retirement planning.

Respondent's Age
Retirement planning has received much attention from policymakers. Jacob-Lawson, Hershey, and Neukam [42] tested a comprehensive and integrative retirement planning model among a group of middle-aged working individuals. One popular view of financial planning, the successful aging perspective [43], was suggested to test the quality of decisions, which focuses on individuals above 50 years old. Baistaman et al. [44] also addressed the issue that people's age can influence the impact of financial literacy and social influences. Therefore, this former model was redesigned and redevised by economists, sociologists, psychologists, and financial planning professionals to identify variables related to financial planning and saving tendencies that are suitable for individuals under 50 years old.
Hypothesis 5a (H5a). Age moderates the relationships between financial literacy and retirement planning.
Hypothesis 5b (H5b). Age moderates the relationships between social influence and retirement planning.
Hypothesis 5c (H5c). Age moderates the relationships between saving attitude and retirement planning.
Hypothesis 5d (H5d). Age moderates the relationships between goal clarity and retirement planning.

Gender
Over the past few years, many previous studies on retirement planning have examined the relationships of sociodemographic factors (e.g., age, education level, gender, marital status, and housing income) with retirement planning [34,[45][46][47]. Financial literacy can be associated with mathematical skills because it depends on arithmetic capacity [39]. Females often outscored males, although people believe that men are better than women in mathematics skills [48][49][50]. Based on the meta-analysis of a previous study, there was no gender difference in terms of a deeper understanding of mathematical concepts and theory [51]. In terms of the retirement period, women tend to retire earlier than males because they want to provide direct care to their family members [52]. Meanwhile, men are less likely to retire because they have to continue providing financial support for their family members.
Hypothesis 6a (H6a). Gender moderates the relationships between financial literacy and retirement planning.
Hypothesis 6b (H6b). Gender moderates the relationships between social influence and retirement planning.
Hypothesis 6c (H6c). Gender moderates the relationships between saving attitude and retirement planning.
Hypothesis 6d (H6d). Gender moderates the relationships between goal clarity and retirement planning.

Status
Another possible explanation between financial literacy and retirement planning is spousal influence or marital status [29,53]. Spousal influence is a strong factor in retirement decision-making because the choice of a spouse should be supported by their partner. Retirement life without proper planning requires continuous work even though they have reached the retirement age [54]. The lack of retirement planning can cause family difficulties in the golden years [35].
Hypothesis 7a (H7a). Marital status moderates the relationships between financial literacy and retirement planning.
Hypothesis 7b (H7b). Marital status moderates the relationships between social influence and retirement planning.
Hypothesis 7c (H7c). Marital status moderates the relationships between saving attitude and retirement planning.
Hypothesis 7d (H7d). Marital status moderates the relationships between goal clarity and retirement planning.

Education
In addition, there are comprehensive studies that cover the factor of education levels. Most previous studies have found that education level is one of the essential factors that determine the behavior of pensioners when preparing for their retirement [37,[55][56][57]. Joo and Pauwels [55] stated that individuals with higher education tend to be more knowledgeable and confident when planning their retirement income. A higher level of education is positively related to a higher probability of confidence in retirement planning. Therefore, a household with more wealth is positively linked with retirement preparedness.
Hypothesis 8a (H8a). Education moderates the relationships between financial literacy and retirement planning.

Hypothesis 8b (H8b). Education moderates the relationships between social influence and retirement planning.
Hypothesis 8c (H8c). Education moderates the relationships between saving attitude and retirement planning.
Hypothesis 8d (H8d). Education moderates the relationships between goal clarity and retirement planning.

Income
Income and age are correlated in retirement planning behavior [54,58]. Working individuals are motivated to take action for retirement when there is an increase in their age and income. This statement is supported by Hira, Rock, and Loibi [23] and Arano, Parker, and Terry [59], who stated that the planned retirement age is guided by different perceptions of income adequacy.

Hypothesis 9a (H9a). Income moderates the relationships between financial literacy and retirement planning.
Hypothesis 9b (H9b). Income moderates the relationships between social influence and retirement planning.
Hypothesis 9c (H9c). Income moderates the relationships between saving attitude and retirement planning.
Hypothesis 9d (H9d). Income moderates the relationships between goal clarity and retirement planning.
All association hypothesized and tested, presented in Figure 1.

Sample Size and Measures
This study used a questionnaire to obtain information from the respondents. The sampling frame was initially composed of 869 private companies, which had more than 10,000 working individuals. A large number of working individuals was identified, and this study used the systematic random approach in which every tenth company on the list was selected, contacted by telephone or email, and the corresponding worker asked to participate in the survey. The systematic random approach was chosen because it is one of the probability sampling techniques where each unit has an equal chance of probability to be selected. After approaching 63 companies, the enumerators contacted 625 prospective respondents who were under 40 years old. The enumerators were appointed based on their experiences in the fieldwork, and they were trained for one month before the data collection stage to ensure that they could provide informed responses. The respondents include junior and senior executives in business, officers, managers, and chief executive officers (CEOs). Nonetheless, 6 out of 63 companies were excluded from the study because they did not have workers under 40 years old, and another 15 companies declined the survey. Therefore, 42 companies agreed to participate, with a total of 378 prospective respondents. The questionnaires were given to the representatives from each company, and the respondents were asked to return them within one week. Finally, 335 responses were recorded, with a response rate of 88.62%. Only 43 questionnaires were not returned within a week. Then, 12 questionnaires were unusable due to incompleteness and double answers, which resulted in the final sample size of 323 respondents. The number of respondents met the minimum requirement of sample size using the Hair approach. According to Hair et al. [60], the number of sample size can be determined by the number of variables included in a model. Thus, we apply the 10-times rule to obtain the sample size. Using this approach, the total variables in this study is 36, which means that the minimum and maximum range of sample size is (36 × 5) 180 and (36 × 10) 360, respectively.
This study deals with multiple unidimensional constructs for retirement planning, as proposed by numerous researchers, such as the sets of financial literacy [26], saving attitude [61], social influence [62], and goal clarity [63]. This is a first-order construct, which is assessed by interdependent variables. Specifically, financial literacy and retirement planning were assessed by nine reflective indicator measures, whereas saving attitude, goal clarity, and social influence were measured by six reflective indicator measures. In the pilot study, the data were analyzed by exploratory factor analysis and Cronbach's alpha to measure the suitability and reliability of items in the retirement planning behavior model.

Data Analysis Method
This study used covariance-based structural equation modeling (CB-SEM), which has gained prominence in various areas of tourism [64,65], management research [60], advertising [66], and other fields [67][68][69] for analyzing research model relationships. This study selected CB-SEM because the technique is confirmatory in nature [70] in order to test the existing theory. The hypotheses are grounded in causal estimation, where the model has high estimation accuracy. Moreover, the adequate sample size and the use of the probability approach for the sampling technique in the east coast region of Malaysia are comparatively proper. This study used the maximum likelihood estimator with the maximum number of 100 iterations in the CB-SEM algorithm settings. The normality of the data is achieved as the value of skewness is between 0.021 and 0.371, which is less than 3.0 [60]. In addition, the multivariate normality value is 2.150, lower than 50.0, which indicates that the data at hand are normally distributed and meet with maximum likelihood properties.
To further discuss this operationalization, the one-way interaction analysis was performed because the moderating effect becomes one of the main analyses to complement model estimation. This analysis was conducted after running the heterogeneity test (chi-square difference) for every moderator variable, which is also recognized as the prominent approach for moderation analysis [68]. The analysis procedure was set up by splitting the data from different groups of moderator variables and the significance of chi-square values that were obtained by different types of models (constrained and unconstrained models).

Assessment of Common Method Bias
The effects of common method bias have long been discussed in previous research [71]. The researcher defined the measurement process from the beginning phase, in which the content of the item, response format, instruction, the characteristics of examiners, the capability of respondents, and the respondents' motivation are the factors of method bias. The threat of the effect of common method bias has long been discussed in previous research [71]. This study addresses the statistical issues by implementing the common latent factor using SEM to provide consistent results [72]. The results from the common latent factor indicate that there is no method bias in the data. To check the results' robustness in terms of common method bias, this study used the alternative method of Harman's single factor. This analysis indicates that a single factor explains 32.5% of the total variance, which is less than 50%. The result implies that the detrimental effect of method bias did not affect the results. Table 1 shows the demographic representation of the prospective respondents. The data were analyzed using SPSS software to obtain the value of frequency and percent for each group of variables. The majority of the respondents were male, aged between 31-40 years old, and had a bachelor's degree as their highest qualification, and the range of monthly income was between RM4001 and RM6500.  Table 2 summarizes the standardized loadings, average variance extracted (AVE), composite reliability (CR), mean, and standard deviation for each construct and item.

Reliabilities and Validities
The internal reliability represented by CR values is consistently high, which fulfilled the recommended use of a 0.7 threshold value after deleting poor loadings from the measurement models. According to Nasir et al. [73], the acceptable standardized loading in the measurement model is at least 0.60. Specifically, five items from all measurement models have poor loading, in which one item each is from financial literacy, saving attitude, and goal clarity constructs, whereas two items are from the retirement planning construct. Therefore, all the retained indicators exhibit high standardized loadings, which yielded high average variance extracted (AVE) values above the 0.50 threshold, thus, supporting the convergent validity criterion. For discriminant validity testing, this study used the conventional approach as the Fornell and Larcker criterion, which has the best approach to assess discriminant validity in the CB-SEM. All the construct correlation values are lower than the square root function of AVE [67], thus, supporting the discriminant validity criterion shown in Table 3.

Path Analysis
This study follows standard evaluation guidelines to analyze the first-order construct measurement models and the structural model [74]. The first phase assessed the measurement models that focus on exogenous construct measures of internal reliability, construct validity, convergent validity, and discriminant validity, as depicted in Tables 2 and 3. Construct validity can be explained by the global fitness indices. This study used chi-square/df, RMSEA, CFI, IFI, and TLI to represent parsimonious, absolute, and incremental fit to evaluate the measurement of model fitness. The chi-square/df is deemed satisfied when the value is lower than 3.0, and RMSEA is below 0.08. CFI, IFI, and TLI are declared an excellent fit when the values estimated are above 0.95. The measurement models satisfied all the recommended threshold values (chi-square/df = 1.043, RMSEA = 0.011, CFI = 0.997, IFI = 0.997, and TLI = 0.996).
Lastly, the structural model was assessed to test the relationship between financial literacy, social influence, saving attitude, goal clarity, and retirement planning. The results in Table 4 and Figure 2 show the unstandardized and standardized estimates. Figure 2 also shows the result for R 2 values, in which the model is explained by approximately 0.44 or 44%. For path coefficient estimates, it is revealed that financial literacy, saving attitude, goal clarity, and social influence have a positive and significant effect (p < 0.05) on retirement planning. This study concludes that all research hypotheses are supported.

Measurement Invariance
To assess measurement invariance, the analysis used Byrne's (2010) procedure, namely, the heterogeneity test (chi-square difference) for metric invariance before testing the moderating effect. Configural invariance is established because chi-square/df, CFI, and RMSEA fulfilled the recommended values across two different groups (gender, age, income, status, and education). The procedure for metric invariance can be implemented by providing chi-square values from unconstrained (chi-square = 922.276, df = 848) and constrained models (chi-square = 954.686, df = 879). The chi-square difference yielded from those models is 32.41, and p-value = 0.397, which is above the recommended value of 0.05, thus concluding that the model has a partial measurement invariance. This study did not test the scalar invariance because full measurement invariance is unnecessary for a further test of invariance and it does not provide sufficient information [71]. The researcher can analyze the moderation effect to this path model by providing a chi-square value from each group of moderator variables, as shown in Tables 5 and 6. In CB-SEM, there are several approaches that were introduced to perform the multigroup analysis, such as user-defined estimand, the heterogeneity test, pairwise deletion, and the critical ratios for difference test. For this study, we used the heterogeneity test or chi-square difference test to assess the significance result for categorical moderators (gender, age, income, marital status, and education) as it is imperative to understand the significance effect on each group. Using this approach, the value of chi-square for each group is compared by constraining the path of interest to get the value of the chi-square difference test. According to Hair et al. [60], the moderator is statistically significant when the value of chi-square difference is above 3.84. The chi-square difference test is actually the same thing as the value of the z-score [72]. Thus, one can conclude that gender, age, income, marital status, and education were found to moderate the relationships between financial literacy, saving attitude, goal clarity, social influence, and retirement planning. Since both groups are found significant (as depicted in Tables 5 and 6), thus, partial moderation has occurred.

Interaction Effect
The diagrams in Figures 3-22 present the effect of one-way interaction on the impact of four exogenous constructs (financial literacy, social influence, saving attitude, and goal clarity) on the retirement planning for different genders (men and women), age (21-30 and 31-40), status (single and married), education (high and low education), and income (high and low income). For the education variable, we recoded the Diploma and Other group as low education, while Bachelor and Master/PhD was considered high education. In addition, we also recoded the income variable by combining monthly income brackets of less than RM4000 as low income, whereas RM4001 and above is considered high income. We did this because the SEM method cannot handle more than two groups of categorical variables. The analysis of interaction was performed to investigate the role of each moderator variable on the proposed relationships.
Interaction Effect (Moderating Role of Gender)

Interaction Effect (Moderating Role of Age)
The results from Figures 7-10 are similar to previous studies, where the moderator variable of age has an interaction in all relationships. The participants are below 40 years old. Participants in the age range of 20-30 are more literate in finance, with better saving attitudes and goal clarity than those in the age range of 31-40. Therefore, it can be concluded that older participants have more influence in social relationships than young participants. Participants who are in the age range of 31-40 have more influence than participants from the age range of 20-30.     Figures 11-14 show the result of marital status as a moderator variable. The participants from the singles group are ahead in goal clarity and financial literacy constructs. In contrast, married participants outperform the singles group in social influence and saving attitude. Married participants have more commitment and responsibility than the single participants, which led them to spend according to their needs.

Interaction Effect (Moderating Role of Education)
The moderation effects of education (see  occurred in all relationships. Participants with high education were more inclined towards financial literacy, social influence, goal clarity, and a saving attitude. The moderation results are similar to previous studies [25,75,76].

Interaction Effect (Moderating Role of Income)
Participants with low income are better planners than participants with high income, as can be seen from the R 2 results (see . Participants with high incomes are not concerned with retirement planning because they believe that their savings are sufficient for retirement. In summary, all moderator variables partially moderate the effect in a model because both groups (Tables 5 and 6) have a highly significant effect.

Discussion
This study provides empirical insight into the direct and moderating effects of financial literacy, saving attitude, social influence, goal clarity, and retirement planning, with public demographic perspectives. This study complements the conceptual consideration by previous studies [27,77]. These constructs have a positive significant effect on retirement planning, which assumes that the public is aware of retirement issues. Further analysis of the moderating effect between public demographic and retirement planning using the heterogeneity test revealed a significant moderation effect. This study provides insight into the retirement planning model using in-depth analysis, which provides more information about this element.
These findings are stable across gender, age, income, status, and education samples. There is no significant difference between them in the model effects through the measurement invariance in SEM. These results support the generalizability for the findings across groups with distinct gender [77], age [78], income [75], marital status [79], and education [80]. The results suggest that the retirement planning behavior model is robust in terms of demographic differences, and the summary of the research hypotheses is shown in Table 7. On the other hand, the model was also verified by establishing the global fitness index, which implies the suitability of the indicator to assess the role of the constructs. Table 7. Summary of hypotheses testing.

1.
Financial literacy has a significant effect on retirement planning Supported 2.
Saving attitude has a significant effect on retirement planning Supported 3.
Social influence has a significant effect on retirement planning Supported 4.
Goal clarity has a significant effect on retirement planning Supported 5.
Gender moderates the relationships between financial literacy and retirement planning Supported 6.
Gender moderates the relationships between saving attitude and retirement planning Supported 7.
Gender moderates the relationships between social influence and retirement planning Supported 8.
Gender moderates the relationships between goal clarity and retirement planning Supported 9.
Age moderates the relationships between financial literacy and retirement planning Supported 10.
Age moderates the relationships between saving attitude and retirement planning Supported 11.
Age moderates the relationships between social influence and retirement planning Supported 12.
Age moderates the relationships between goal clarity and retirement planning Supported 13.
Education moderates the relationships between financial literacy and retirement planning Supported 14.
Education moderates the relationships between saving attitude and retirement planning Supported 15.
Education moderates the relationships between social influence and retirement planning Supported 16.
Education moderates the relationships between goal clarity and retirement planning Supported 17.
Status moderates the relationships between financial literacy and retirement planning Supported 18.
Status moderates the relationships between saving attitude and retirement planning Supported 19.
Status moderates the relationships between social influence and retirement planning Supported 20.
Status moderates the relationships between goal clarity and retirement planning Supported 21.
Income moderates the relationships between financial literacy and retirement planning Supported 22.
Income moderates the relationships between saving attitude and retirement planning Supported 23.
Income moderates the relationships between social influence and retirement planning Supported 24.
Income moderates the relationships between goal clarity and retirement planning Supported Finally, the supplementary analyses support this conclusion, which indicates that the interaction effects in testing the moderating effect did not distort the results. This study also contributes to the methodological aspects by underlining the importance of supplementary analyses to determine that the researchers have obtained the information in more detail. This study concludes that single women at the age of 20-30, with high education and low income, have high financial literacy and goal clarity in preparing for their retirement savings, which is consistent with Sabri and Juen's [14] findings. When assessing social influence on retirement planning, it is revealed that married women at the age of 31-40, with high education and low income, are more pronounced in these effects. Finally, married women at the age of 20-30, with high education and low income, are more obvious in the relationship between saving attitude and retirement planning. Overall, women are more ready than men to save for retirement.

Conclusions
This study has several implications, and it offers some recommendations for future research. It provides a foundation to further assess financial literacy, goal clarity, saving attitude, social influence, and retirement planning. Moreover, this study offers a mechanism to frame the causal effect relationship with public demographic properties in a model that is tested regularly, with no disturbance effect. Although the results revealed a significant moderating effect in all the relationships in a model, it is revealed that gender, age, income, education, and marital status variables only partially moderate those relationships. Hence, future research should focus on exploring those moderator variables.
Future studies should test these structural properties across different domains, for example, countries with different cultures and socioeconomic characteristics, rather than focusing on the east coast region of Malaysia. The model proposed from this study can be generalized to neighboring countries such as Indonesia, Thailand, and Singapore due to similar demographic characteristics such as culture and attitudes. This study has limitations. First, this study focused on working individuals under the age of 40 to examine their preparedness for retirement savings. Hence, people who are more than 40 years old were not considered in this study. Lastly, this study used a cross-sectional design, whereby the respondents' decisions on their savings were only measured once throughout the study. In future research, the application of latent growth curve modeling or multilevel modeling is more suitable because it can estimate the respondents' decisions more than once.