Dear experts,

I have a variable "Financial socialization" which measures an individual's level of financial learning and experiences from various socializing agents such as parents, friends, spouse, etc. Financial socialization is contextual in nature (for e.g. financial socialization through parents, friends, spouse, etc.). I am using the financial socialization scale proposed by Shim et al. 2010 who measured parental financial socialization (item e.g. My parents openly discussed financial matters with me, my parents taught me strategies for smart shopping etc.).

If I take more than one agent (parents, spouse, and friends as supported by literature), I have to measure financial socialization separately for every socializing agent. But it will also substantially increase the length of the questionnaire. There is no scale to measure financial socialization separately for each agent, so I have to adapt Shim et al's scale. So, I thought of a solution.

In the first question, I asked respondents to rate the influence of each socialization agent (e.g., parents, peers, media) on their financial knowledge/behaviour (on a 5-point high to low scale).

In the second question, I measured the Financial Socialization construct as a whole (overall), referring to all the agents without mentioning any particular person as follows:

Think about the times when all the people mentioned above (your parents, friends or colleagues, spouse, c) shared their financial knowledge with you, did they do any of the following financial activities? (strongly agree to strongly disagree)

items e.g. Openly discussed financial matters with me, Taught me strategies for smart shopping etc.

Now, since I got the weights assigned to each family member (first question), as well as consolidated financial socialization scores from the second question.

Following are my queries:

1. Can I create separate constructs for each agent (parents, friends, spouse) by multiplying the weighted scores of each agent with the consolidated FS score (second question)?

1. Is this weighted scoring method valid and robust? If not, please suggest an alternative.

2. Can I use regression coefficients as weights instead of arbitrary numbers?

3. I do not have any citation for using this method in my context, although there exists mathematical literature which supports such an approach. Do I need to look for literature support in my domain necessarily or not?