I can't help on your basic question, but polynomials beyond the cube level are generally uninterpretable and highly unstable. The tails really whip around due to the data dependencies; if the data change even a little the coefficients can be very different. I think that Frank Harrell's Regression book has a discussion on it.