Long-term (over 14 years) experiments on winter wheat (Triticum aestivum L.)–rice (Oryza sativa L.) crop rotations were conducted in Southwest China to investigate phosphorus (P) fertilizer utilization efficiency, including the partial factor productivity (PFP), agronomic efficiency (AE), internal efficiency (IE), partial P balance (PPB), recovery efficiency (RE) and the mass (input–output) balance. The seven treatments were Control, N, NP, NK, NPK, NPKM and NPKSt, representing various combinations of inorganic fertilizers (N, P and K), manure (M) and the application of rice straw (St). Without P application, the soil could supply c. 14·7–22·5 kg P/ha annually and produce, on average, c. 1·8 t/ha wheat and 6·0 t/ha rice. Phosphorus fertilization increased crop yields by 65·5 and 11·4% for wheat and rice, respectively, over the 14 years. The PFP values ranged from 80·2 to 177 kg grain/kg P fertilizer for wheat and from 222 to 255 kg/kg for rice in the NPK treatments. However, the mean AE over the 14-year period was 31·9 and 21·3 kg grain/kg inorganic P fertilizer for wheat and rice, respectively. The mean IE was 214 and 318 kg grain/kg P uptake for wheat and rice, respectively, during the cultivation period. The PPB for the whole rotation system over the 14 years ranged from 0·58 to 0·64. However, the mean RE of P fertilizer was 0·26 (varying from 0·22 to 0·29) in the wheat–rice cropping system over the 14-year period. For every 100 kg surplus P/ha per year, the concentration of soil P extracted by 0·5 m NaHCO3 at pH 8·5 (Olsen-P) would increase by, on average, 4·12 mg/kg in soil. For the wheat–rice cropping system, the current P application rate of 55–65 kg P/ha per year is able to sustain annual yields of about 3 t/ha for wheat and 7 t/ha for rice. This study suggests that, in order to achieve higher crop yields, the P fertilizer utilization efficiency should be considered when making P fertilizer recommendations in wheat–rice cropping systems.