We define 'transient poverty' as the component of time-mean consumption pov
erty at household level that is directly attributable to variability in con
sumption; this can be thought of as a measure of vulnerability to falling c
onsumption. The non-transient component then depends solely on mean consump
tion over time,and we call this 'chronic poverty'. Using robust semi-parame
tric; methods and household panel data for rural China we test whether tran
sient poverty is determined by a process that is similar to chronic poverty
. Commonly identified causes of poverty in this setting have weak explanato
ry power for transient poverty and some of the factors determining transien
t poverty do not matter to chronic poverty, or even have the opposite effec
t. Successful policy responses to chronic poverty may still leave considera
ble transient poverty.