Intercropping is known to increase the efficiency of land use, but no meta-analysis has so far been made on the yield gain of intercropping compared to sole cropping in terms of absolute yield per unit area. Yield gain could potentially be related to a relaxation of competition, due to complementarity or facilitation, and/or to the competitive dominance of the higher yielding species. The contributions of competitive relaxation and dominance were here estimated using the concepts of complementarity effect (CE) and selection effect (SE), respectively. We compiled a dataset on intercropping of grain-producing crops from China, a hotspot of strip intercropping in the world. We quantified the yield gain and its components and analysed the contribution to yield gain of species traits (C3, C4, legume, non-legume), complementarity in time and nutrient input. Total yield in intercrops exceeded the expected yield, estimated on the basis of sole crop yields, by 2.14 ± 0.16 Mg ha−1 (mean ± standard error). Ninety percent of this yield gain was due to a positive CE while the remaining 10 % was due to SE. The net yield gain increased with temporal niche differentiation (TND) which is the proportion of the total growing period of the crop mixture during which species grow alone. The mechanism underlying yield gain shifted from competitive dominance of the higher yielding species when there was more overlap in growth period between the two species, to competitive relaxation when there was less overlap, while competitive relaxation remained the major component of the yield gain. The yield gain was substantially greater in intercrops with maize than in intercrops without maize, but there was no difference in yield gain between systems with and without legumes. The yield gain increased with nitrogen (N) input in maize/C3-cereal intercrops but not in cereal/legume intercrops, illustrating the ability of legumes to compensate for low N input, and highlighting the need for N input for high productivity in intercropping systems without legumes. Yield gain did not respond to phosphorus (P) input. We conclude that competitive relaxation is the main contributing factor to yield gain in the investigated Chinese intercropping systems, which were mostly relay strip intercropping systems. The underlying drivers of yield gain were related to presence of maize and species complementarity in time, but we did not find strong evidence for the selection effect.