Who saves more, the naive or the sophisticated agent?

We consider a class of additively time-separable life-cycle consumption-savings models with iso-elastic per period power utility featuring resistance to inter-temporal substitution of θ with linear consumption policy functions. The utility maximization problem is dynamically inconsistent for almost all specifications of effective discount factors. Pollak (1968) shows that the savings behavior of a sophisticated and a naive agent is identical with logarithmic utility (θ = 1). We extend this result by showing that the sophisticated agent saves in any period a greater fraction of her wealth than the naive agent if and only if θ ≥ 1, irrespective of the discount function.