Robustness of 5qx estimators under nonstationarity.

When age-specific death rates are available for single year ages any of several formulas may be used to estimate the probability of dying in an age interval, nqx. For 5-year age intervals, however, the simplest formulas are usually not as precise as some that are more complex. If in addition, the population is stable but nonstationary or is subject to fairly regular birth cycles, a formula due to Keyfitz and Frauenthal performs markedly better than others. Tests on real populations also support Keyfitz-Frauenthal, and suggest that actual differences between other competing formulas may be small.