We study the propensity to save on permanent and transitory income based on a comprehensive household panel data-set from South Africa. We decompose income into permanent and transitory components, and proxy transitory income by weather deviations from thirty-year normal. We evaluate the propensity to save by OLS and by median regressions. By OLS, we find that the propensity to save on transitory income is not significantly different from one, while the propensity to save on permanent income is significantly different from zero. This finding is in alignment with previous studies. By median regression, we find that the propensity to save by permanent or transitory income are significantly different from either zero or one, meaning that the propensity to save represents some fractions of permanent and transitory income. We also evaluate the propensity to save by income quintiles using median regression. We find that the propensity to save from permanent and transitory income are only significantly different in the highest quintile when durable goods are considered as saving, while not significantly different in the lower 80% of the income distribution. In the top 20%, we find the propensity to save from transitory income is significantly higher than that of permanent income, although it still does not reach one.