Four equations (arps, semilog, extreme-c, and double semilog) were tested to determine which would produce a curve that best fit data points that represent a trend of declining oil production in 90 California fields. The curves were first fitted to the initial 9 years of the 15-year production data and then extrapolated for 6 years to determine how well the extrapolated curve fit the data points. The generated curves also were extrapolated beyond the last data point to predict rates of future production. No equation was clearly superior to the others in all respects. The semilog equation resulted in the best fit to production data for the first 9 years, but the arps and extreme-c equations produced a better fit for the extrapolated data for the period of 9 to 15 years, and also more realistic production rates when extended to the point where cumulative production equaled reserves. A second study was made to determine if the least-squares method applied to the arps equation would result in more accurate extrapolations. The logarithmic and the nonlogarithmic forms of the arps equation also were tested. For the logarithmic expression, one of the constants reached a value so small or so large that the equation was not applicable to data from 40 of the 90 California fields. The nonlogarithmic arps equation was applicable to neither the same 40 fields nor to a few additional ones. When the data and the arps equation were compatible, the use of the least-squares method as compared with other methods improved the extrapolation accuracy somewhat.