How did Feldstein (1985) undervalue the optimal level of social security benefits?

In his seminal model (Feldstein, 1985), the government operates a social security system to counter the representative worker's myopia. (i) For a complete myope, he determined a sizable optimal tax rate (and the corresponding benefit level). (ii) For a partially shortsighted worker, he determined another optimum, which was much lower, possibly zero. Departing from Feldstein, I take into account that neither a paternalistic government nor a cautious bank tolerates long-term negative saving, and then even in (ii), the government may choose the first rather than the second optimum. Having revised it, Feldstein's model regains its place in the textbooks.