Do Crop Income Shocks Widen Disparities in Smallholder Agricultural Investments? Panel Survey Evidence From Zambia

We investigate whether the effects of negative crop income shocks in one season persist in subsequent seasons due to reductions in crop inputs. If bad seasons cause household cash constraints to bind, and this results in the scaling back of the next season’s production, the next season’s crop income is also compromised, potentially creating a poverty trap. Troublingly, households most susceptible to such a poverty trap mechanism are likely to be those that rely the most on own-farm production and have the fewest sources of liquidity—in other words, the poorest. We use data from a three-wave (2001, 2004, and 2008), nationally-representative survey of smallholder farm households in Zambia to test for the effect of rainfall shocks—interacted with measures of household liquidity—on investment in maize production in the following season. We focus specifically on the ability (or inability) of farm households to invest in own-farm maize production in the form of mineral fertilizer use, improved seed use, and area allocated to maize. We use three liquidity measures: livestock, regular off-farm wage employment, and access to subsidies/loans for fertilizer purchase. A priori, we predict that the presence of such liquidity sources will protect maize investments from negative income shocks in the previous seasons. These liquidity measures may be endogenous to the input decisions; we therefore use panel data methods and an instrumental variables/control function approach. Additionally, we test whether reduced maize inputs do indeed cause reduced maize income and, ultimately, total income. Our results show that the effects of rainfall shocks in one agricultural season persist into the subsequent season in the form of reduced maize inputs. The estimated effect of reduced inputs on the following season’s income, however, is modest. However, we must keep in mind that this estimated modest effect is the average effect across all sampled households. Whether this mechanism constitutes a poverty trap for a particular household depends on that household's overall reliance on farm production, as well as the distribution of rainfall shocks that it faces. For households that rely overwhelmingly on crops and typically experience multiple deficit periods in bad years, even two or three consecutively bad years could still pose a poverty trap. Surprisingly, liquidity—as measured by livestock, salaried household members, and fertilizer subsidy access—does not increase households’ ability to smooth inputs. It is important to note, however, that livestock and salaried household members may not be appropriate liquidity measures for the poorer households in the sample. Given that inputs decrease as a result of (negative) rainfall shocks in previous seasons, and given the ability to observe rainfall shocks over Zambia at a fine scale, input divestment might be predicted, geographically, based on rainfall patterns. On a season-to-season basis, the allocation of resources through programs such as the Farmer Input Support Program (FISP) could be informed by these predictions of input divestment. That way, programs such as the FISP would pose less of a crowding-out threat to existing sources of fertilizer, and more effectively target the neediest communities each season.