They didn't optimize on the right thing. My models indicate that, depending on market returns, many scenarios have me paying MORE tax but being able to spend MORE by taking SS early. They also don't disclose the rates of return they used, calling them "proprietary" (gimme a break)
You hit upon a real weakness in the paper. The authors don’t explicitly say what asset allocations they used in their simulations, but do hint that it was fairly conservative. There are many other assumptions that must go into their models that were not explained, leaving me uncertain as to what to make of the paper. However, there was only a median difference of $31,000 in lifetime wealth between when people actually chose to take social security and what their “optimal” time would have been in the simulations. Since no one can know beforehand when their optimal age to take SS is, it leaves me wondering what the significance of the paper is.