Dielectronic recombination (DR) is the dominant recombination process for most heavy elements in photoionized clouds. Accurate DR rates for a species can be predicted when the positions of autoionizing states are known. Unfortunately such data are not available for most third and higher-row elements. This introduces an uncertainty that is especially acute for photoionized clouds, where the low temperatures mean that DR occurs energetically through very low-lying autoionizing states. This paper discusses S$^{2+} \rightarrow$ S$^+$ DR, the process that is largely responsible for establishing the [S~III]/[S~II] ratio in nebulae. We derive an empirical rate coefficient using a novel method for second-row ions, which do have accurate data. Photoionization models are used to reproduce the [O~III] / [O~II] / [O~I] / [Ne~III] intensity ratios in central regions of the Orion Nebula. O and Ne have accurate atomic data and can be used to derive an empirical S$^{2+} \rightarrow$ S$^+$ DR rate coefficient at $\sim 10^{4}$~K. We present new calculations of the DR rate coefficient for S$^{2+} \rightarrow$ S$^+$ and quantify how uncertainties in the autoionizing level positions affect it. The empirical and theoretical results are combined and we derive a simple fit to the resulting rate coefficient at all temperatures for incorporation into spectral synthesis codes. This method can be used to derive empirical DR rates for other ions, provided that good observations of several stages of ionization of O and Ne are available.