I'm doing a research about the relationship between investment risk tolerance and demographic factors such as age, generation, period and control variables.
On my main regression, I used the following model:
Code:
 
 logit highrisk age i.period i.wwgen
In which, highrisk = 1 if respondent willing to take highrisk investment, 0 otherwise; period represents dummies for every 4 years since 1993-2019 (e.g. 93-96, 97-00, 01-04,...); wwgen are the way of generation clarify following Worldwide standards (Z, Y, X, Baby boomers, Silent, GI).
Then as robustness checks, I run the second model:
Code:
 
 logit highrisk agegr year i.usgen
For this model, agegr are age group (u24, u40, u65, u80, u100); usgen are the way of generation clarify following American standards (Depression era, WW2, Post-war, BB1&2, X, Y, Z).
But for the period effects, if I include every survey years in the model, it would be too long. I expect respondents tolerate different level of risk in each year according to the economic events happening in that year. For example, in 2018, people tend to be more risk aversion due to financial crisis.
So I'm posting this to seek for another way to measure period effect (which is survey years)