It is not the variance of income but the conditional variance of innovations in the income process that determine the degree of precautionary saving. We estimate this process at both the micro and macro level and construct the variances of the innovations explicitly.
In the absence of panel data on individual households we develop a grouping estimator that allows us to estimate this income process with conditional heteroskedasticity from repeated cross-section data.
Construction of the relevent conditional variances and substitution into a consumption growth equation yields the expected positive effects - precautionary saving increases consumption growth by depressing consumption - consumers delay spending in the face of uncertainty as the life-cycle model predicts.