Two empirical density models based on drag analyses of 22 low-perigee satellites are developed for use in predicting low-altitude satellite ephemerides. A stepwise multiple regression analysis is performed with density at 145 km as the dependent variable, and a number of independent variables chosen to represent variations with solar cycle, geomagnetic activity, latitude, season, day of the year, and time of day. Densities between 120 and 500 km are determined from the hydrostatic law in a simple (but physically realistic) analytic form. The two models compare very well with the 1971 Jacchia model. The main advantage of such models is that they represent a major savings in both computer storage and run time making possible improvements in operational systems.