There are alternative definitions of vulnerability to poverty. Most researchers prefer to define vulnerability as the probability of a household or individual falling into poverty in the future. Based on this definition and using household survey panel data from rural China, this paper attempt to assess the extent to which we can measure vulnerability to poverty. The assessment is based on comparisons between predicted vulnerability and actually observed poverty. We find that the precision of prediction, first, varies depending on the vulnerability line; our results suggest setting the line at 50 per cent in order to improve predictive power. Second, precision depends on how permanent income is estimated. Assuming log-normal distribution of income, it is preferable to use past weighted average income as an estimate of permanent income rather than using regressions to gage permanent income. And third, prediction precision depends on the chosen poverty line. More accurate measurement of vulnerability to poverty is obtained with a higher poverty line of US$2 instead of US$1.
