I theoretically analyze and empirically identify the optimal default savings rate in automatic enrollment retirement saving plans. I derive a formula for the optimal default as a function of sufficient statistics that can be empirically identified. I estimate individual adherence to the default using exogenous increases in the default rate of OregonSaves, the first state-sponsored auto-enrollment plan in the U.S. I also use survey data to infer the degree of undersaving if workers actively switch to a non-default rate. Combining estimates from administrative and survey data with the optimal default formula, I find the optimal default is 7% of income.