Applying the Shapley decomposition to unit-record household survey data, this paper investigates the trends and causes of poverty in China in the 1990s. The changes in poverty trends are attributed to two proximate causes; income growth and shifts in relative income distribution. The Foster-Greer-Thorbecke measures are computed and decomposed, with different datasets and alternative assumptions about poverty lines and equivalence. Among the robust results are: (i) both income growth and favourable distributional changes can explain China’s remarkable achievement in combating poverty in rural areas in the first half of the 1990s; (2) in the second half of the 1990s, both rural and urban China suffered from rapidly rising inequality and stagnant income growth, leading to a slow-down in poverty reduction, even reversal of poverty trend.