The US FCC mandates the testing of all mobile phones to demonstrate compliance with the rule requiring that the peak spatial SAR does not exceed the limit of 1.6 W/kg averaged over any 1 g of tissue. These test data, measured in phantoms with mobile phones operating at maximum antenna input power, permitted us to evaluate the variation in SARs across mobile phone design factors such as shape and antenna design, communication technology, and test date (over a 7-year period). Descriptive statistical summaries calculated for 850 MHz and 1900 MHz phones and ANOVA were used to evaluate the influence of the foregoing factors on SARs. Service technology accounted for the greatest variability in compliance test SARs that ranged from AMPS (highest) to CDMA, iDEN, TDMA, and GSM (lowest). However, the dominant factor for SARs during use is the time-averaged antenna input power, which may be much less than the maximum power used in testing. This factor is largely defined by the communication system; e.g., the GSM phone average output can be higher than CDMA by a factor of 100. Phone shape, antenna type, and orientation of a phone were found to be significant but only on the order of up to a factor of 2 (3 dB). The SAR in the tilt position was significantly smaller than for touch. The side of the head did not affect SAR levels significantly. Among the remaining factors, external antennae produced greater SARs than internal ones, and brick and clamshell phones produced greater SARs than slide phones. Assuming phone design and usage patterns do not change significantly over time, we have developed a normalization procedure and formula that permits reliable prediction of the relative SAR between various communication systems. This approach can be applied to improve exposure assessment in epidemiological research.