This cross-sectional study compared body fat percentage (BF%) obtained from a four-compartment (4C) model with BF% from hydrometry (using 2H2O), dual-energy X-ray absorptiometry (DXA) and densitometry among the three main ethnic groups (Chinese, Malays and Indians) in Singapore, and determined the suitability of two-compartment (2C) models as surrogate methods for assessing BF% among different ethnic groups. A total of 291 subjects (108 Chinese, seventy-six Malays, 107 Indians) were selected to ensure an adequate representation of age range (18-75 years) and BMI range (16-40 kg/m2) of the general adult population, with almost equal numbers from each gender group. Body weight was measured, together with body height, total body water by 2H2O dilution, densitometry with Bodpod and bone mineral content with Hologic QDR-4500. BF% measurements with a 4C model for the subgroups were: Chinese females 33.5 (sd 7.5), Chinese males 24.4 (sd 6.1), Malay females 37.8 (sd 6.3), Malay males 26.0 (sd 7.6), Indian females 38.2 (sd 7.0), Indian males 28.1 (sd 5.5). Differences between BF% measured by the 4C and 2C models (hydrometry, DXA and densitometry) were found, with underestimation of BF% in all the ethnic-gender groups by DXA of 2.1-4.2 BF% and by densitometry of 0.5-3.2 BF%). On a group level, the differences in BF% between the 4C model and 2H2O were the lowest (0.0-1.4 BF% in the different groups), while differences between the 4C model and DXA were the highest. Differences between the 4C model and 2H2O and between the 4C model and DXA were positively correlated with the 4C model, water fraction (f(water)) of fat-free mass (FFM) and the mineral fraction (f(mineral)) of FFM, and negatively correlated with density of the FFM (D(FFM)), while the difference between 4C model and densitometry correlated with these variables negatively and positively respectively (i.e. the correlations were opposite). The largest contributors to the observed differences were f(water) and D(FFM). When validated against the reference 4C model, 2C models were found to be unsuitable for accurate measurements of BF% at the individual level, owing to the high errors and violation of assumptions of constant hydration of FFM and D(FFM) among the ethnic groups. On a group level, the best 2C model for measuring BF% among Singaporeans was found to be 2H2O.