This study compares two global satellite-based surface radiation budget (SRB) data sets as a means of quality evaluation. One was developed from Earth Radiation Budget Experiment (ERBE) satellite data using the algorithm of Li et al. (ERBE/SRB), and the other was generated by applying the algorithm of Pinker and Laszlo to International Satellite Cloud Climatology Project (ISCCP/SRB) data. The comparison is limited to net surface shortwave radiation (NSSR). The global annual mean of NSSR obtained from the two data sets differ by 15 W m-2, and maximum regional differences exceed 100 W m-2. The differences are investigated in terms of discrepancies in both input and data and algorithm. It was found that large regional differences in SRB are associated mainly with the discrepancies in the input parameters, namely, top-of-atmosphere (TOA) flux and precipitable water. Differences in TOA flux between ERBE and ISCCP are attributed to angular and spectral corrections for ISCCP data and to different sampling in time and space by ERBE and ISCCP. ISCCP/SRB underestimates NSSR by 20--30 W m-2 over some dry regions arising from excessive amounts of precipitable water from the TIROS operational vertical sounder. Deficient treatment of aerosol in the Li et al. algorithm results in too large NSSR for the major deserts. Systematic discrepancies are accounted for by different methods to compute water vapor absorption. According to the line-by-line results, water vapor absorption was significantly underestimated by the Lacis and Hansen parameterization (used by Pinker and Laszlo) and is moderately overestimated by LOWTRAN 6 (used by Li et al.). If both algorithms use the same water vapor absorption scheme and the same input data, their global annual mean of NSSR agree to within 1 W m-2. On the basis of these findings, some recommendations are made for future improvements of the two products. Overall, it appears that the quality of ERBE/SRB is superior to that of ISCCP/SRB (version 1.1). ¿American Geophysical Union 1995