Large fluctuations in survey data on the income and asset holdings of rural households in Ethiopia suggest that these households are forced to cope with an enormous amount of uninsured risk.
However, these observed fluctuations could be substantially exaggerated due to measurement errors.
To address this issue, we develop a method that can separate true shocks from measurement error in dynamic panel data models.
Using survey data from Ethiopia we estimate a linearized optimal livestock accumulation rule.
Using these data we find that we cannot reject the hypothesis that all random fluctuations in the observed livestock holdings are driven by measurement error.