How risky is consumption in the long-run? Benchmark estimates from a robust estimator

The long-run standard deviation (LRSD) of consumption growth is a key moment in determining risk premiums under Epstein-Zin preferences. Standard estimators of the LRSD are biased downward and have poor confidence interval coverage, making them overreject the long-run risk model. This paper studies a new estimator with smaller bias and accurate confidence intervals. Standard long-run risk calibrations are still rejected in the data. The LRSD of consumption growth in the postwar sample is estimated to be 2.5% per year with an upper bound to the 95% confidence interval of 4.9%. These values can be taken as benchmarks for future calibrations.