Slope length and slope steepness are the two topographic factors (L and S) used in the Universal Soil Loss Equation (USLE) for soil erosion prediction. These factors are usually extracted from the digital elevation models (DEM). We hypothesized that different DEM resolutions and sources influence extracted topographic factors and therefore soil loss estimation. The objective of this study was to quantify the effect of different DEM resolutions on the topographic factor calculation and consequently soil loss prediction on different geomorphological types in China and United States. We selected three typical landforms of plains, hills and mountains in both countries. We used 1) open source DEMs of 3-, 10-, 30-, and 90-m in the United States, and 2) DEMs generated with 1:10,000 and 1:50,000 topographic maps in China to extract topographic factors using a LS software tool. The Chinese Soil Loss Equation (CSLE) was used to calculate average annual soil loss. The results showed that different DEM resolutions had different effects on the extracted L and S factors. Slope length and the L factor were overestimated as DEM resolutions coarsened, and the trend of the overestimation in different landforms was the same. Slope steepness and the S factor were underestimated as DEM resolution decreased in all landforms. However, the LS factor as a product of L and S factors, which reflects relative soil loss if the other factors are held constant, are less sensitive to DEM resolutions than either L or S factor alone due to partial error cancelation. Thus, the LS factor is not as greatly affected by DEM resolutions as either L or S factor alone. The relative soil loss errors are similar for all DEMs, and “good results for wrong reasons” are obtained with coarser DEMs. Overall results indicate that except for the 90 m DEM for certain occasion, all DEMs may be used to estimate soil loss for all types of terrains for practical application, as far as the relative soil loss errors are concerned. However, in the range of 3–90 m resolutions, for the sake of good practice and solid scientific support, higher resolution DEMs if available ought to be used to minimize estimation errors of L and S factors and consequently to increase confidence of soil loss prediction.