The author shows that the enormous gap between theory and facts in modern macroeconomics can only be eliminated by nonlinear macroeconomic dynamics with the following special characteristics: First of all, only certain group-theoretical invariants generate the correct growth cycles with irregularly varying lengths, not any shastic process as usually applied for this purpose. Furthermore, a special extended value function and generalized human capital are needed for a correct representation of scientific and technological innovation. Finally, the correct nonlinear macroeconomic dynamics are not reducible to microeconomics, for both of the above mentioned reasons.