What to add to Binominal Regression? Demographics + motivations?

#1
So i have a binomial regression for predicting monthly spend on clothes based on shopping motivations; describing 65% of the variance. If I put in some demographics this explnattion goes higher to 69%. But can demographics be added to these models like this to get better results or should they be left out?\

I am investigating the influence of motivations on dependants such as clothes bought.

Also, what % of variance explained is bad, good and excellent (etc.)?
 
#2
Well, as you add predictors to the model whether they are good/bad the r^2 value will always increase. I would test whether the predictors have a statistical significance with the response.
 
#3
Well, as you add predictors to the model whether they are good/bad the r^2 value will always increase. I would test whether the predictors have a statistical significance with the response.
Only two do, age and gender. Does this mean they should be included? Must I run a test MANOVA to determine what should go into the model first?
 
#4
If your theory (or common sense) says that those predictors are relevant to predict you outcome, then add them in the model (regardless of significance of their impact or percent of variance explained). In terms of order, I typically include the focal predictor in the last place (although in practice in hardly ever matters).

Multivariate analysis of variance (MANOVA) is a case when you have several categorical predictors. I don't see how it's relevant here, unless you do have 2+ categorical predictors.

On a side note, I wonder why you choose a binomial regression. What is the underlying logic/justification? This is quite an advanced model, so I am curious what is that your outcome (and its distribution) that you are using such model.