A long-term experiment was carried out in the dryland of northern China to assess the effects of applications of maize stover, cattle manure and NP (1:0.44) fertilizer on partial nitrogen (N), phosphorus (P) and potassium (K) balances, extractable soil N (SEN), P and K, and soil organic matter (SOM) in a spring maize cropping system, under reduced tillage conditions. The experiment was set-up according to an incomplete, optimal design, with three factors at five levels and 12 treatments, including a control with two replications. Statistical analyses using multiple regression models showed that the partial N, P and K balances were strongly influenced by annual variations in the amounts of soil water at seeding (SWS) and growing season rainfall (GSR). Most treatments had positive P but negative N and K balances. Cumulative P and K balances were reflected in extractable soil P (P-Olsen) and K (exchangeable K), but the weak relationships indicated that the sorption of P and buffering of K were strong. Cumulative balances of effective organic carbon (C) were weakly related to soil organic C (SOC) content after 12 years. Negative C balances were related to decreases in SOC, but positive C balances were not translated into increases in SOC. The analysis of nutrient balances and soil fertility indices revealed that nutrient inputs in most treatments were far from balanced. It is concluded that the concepts of 'ideal soil fertility level' and 'response nutrient management' provide practical guidelines for improving nutrient management under the variable rainfall conditions of dry land areas in northern China.