Suppose you are a mean-variance optimizer in a world with many risky and one risk-free asset Your coefficient of risk aversion A-2and you have found the optimal allocation of your wealth into risky and risk-free assets. Now suppose that suddenly you realize your risk aversion coefficient is really A-3. What do you change? 

You change the fraction of wealth in risky vs. risk-free, but keep the composition of the risky portfolio the same

 You change the composition of risky assets in your risky portfolio, but keep the fraction of wealth in risky vs risk-free the same 

You don't change anything

You change both the traction of wealth in risky vs risk-free and the computation of the risky portfolio