Measuring Resilience as Asymmetric Mean Reversion

We introduce a new method to measure resilience, defined as the ability of an individual, household or community to recover after a decline in well-being. Our approach measures resilience as the extent to which outcomes recover, against the benchmark of symmetric mean reversion arising from measurement error or random fluctuations. Observing desirable asymmetry in which recovery is larger than would be expected due to mean reversion allows us to measure resilience without having observed the precipitating shocks. We present the method, derive correction factors to account for autocorrelation, and apply the method to data on diet diversity and anthropometry of women and children from Nepal, Bangladesh, and Uganda. Tests introduced in this paper offer a promising approach to identifying groups with statistically significant resilience; observing the presence or need for social insurance, safety nets and other sources of resilience; and assessing the sustained effects of interventions