Regional earthquake catalogs typically report local magnitudes for most earthquakes. Moment magnitudes, which are commonly used in ground‐motion modeling, are available generally only for larger earthquakes. Earthquake rate forecasts based on local magnitudes must be converted to moment magnitudes for use in earthquake hazard studies. Here, we describe several approximate methods of carrying out the conversion, based on a regression of moment magnitude on local magnitude and its associated uncertainties. They apply to gridded forecasts in the style of regional earthquake likelihood models. The methods vary greatly in computational efficiency, but provide similar results. The simplest method applies a multiplier to the rate in each magnitude bin and spatial cell, assuming that the earthquake rate versus magnitude relation has a Gutenberg–Richter ‐value of 1. This method has been used to convert earthquake forecasts disseminated during the Canterbury and Kaikoura earthquake sequences in New Zealand to moment magnitude.