Consider two regression models. One is a natural cubic spline with 15 degrees of freedom. Another one is a degree-14 polynomial (with 15 degrees of freedom). What would we expect to see in terms of the fit around the boundary (where predictor values are very small or very large.)