The poor fertility of sandy Sahelian soils remains one of the major constraints to pearl millet (Pennisetum glaucum) production in West Africa. On-farm trials under farmers' management were conducted in two rainfall zones of Niger in 1996 and 1997 to evaluate the risk characteristics of six soil fertility restoration options. Stochastic dominance analysis was used to compare the fertilizer treatments tested. Results showed that the farmers' traditional method (no fertilizer control), Tahoua phosphate rock (PRT) alone applied at 13 kg P ha−1 broadcast, and a combination of PRT broadcast at 13 kg P ha−1 and single super phosphate (SSP) hill-placed at 4 kg P ha−1 had the most desirable risk characteristics and were acceptable to risk averse decision-makers in both rainfall zones. At current input–output price ratios, most fertilizer-using farmers would choose the combination of PRT broadcast and SSP hill-placed. If the availability of SSP was limited, some farmers would use PRT alone. The demand for risk efficient alternatives could significantly increase if farmers could bear less than half the fertilizer costs at the current output price, although further research is required to say if a fertilizer subsidy could be justified on broader economic or social grounds.