Methods to identify genes and variants for complex diseases generally assume that effects of alleles are additive, meaning that a homozygous genotype confers twice the risk of a heterozygous genotype. Genes under recessive selection exhibit different population dynamics, especially in populations that have undergone dramatic bottlenecks followed by re-expansion, such as the European population [1]. The majority of genome-wide association studies (GWAS) have assumed additive models; however, the few studies that have tested recessive models have discovered recessive associations [2]. We have developed a novel method to quantify the magnitude and mode of selection of all protein-coding genes across the human genome by comparing European population sequencing data from the Exome Aggregation Consortium (ExAC) (N=35,000) with simulated evolutionary histories for both additive and recessive alleles, This method could inform model choice by identifying genes and pathways likely to be under recessive selection. We find a variety of biologically meaningful categories enriched in the predicted recessive class, including glycoproteins (Benjamini P=6.3x10-13), immunoglobulin domains (P=0.023), and inflammatory response (P=0.0052). The enrichment for inflammatory genes in particular suggests that many complex diseases with inflammatory components may be under recessive selection, such as Crohn's disease and rheumatoid arthritis (RA). In the case of Crohn's disease, we find that genes with large and well-validated effects are predicted to be under recessive selection, while genes implicated by GWAS studies have no enrichment for recessive selection. Similarly, RA loci discovered by GWAS show no enrichment for recessive selection, despite the fact that RA is known to involve pathways that are highly enriched for recessive selection according to our method. This highlights the need for using recessive models in GWAS, and the potential usefulness of our catalog of mode of selection as a tool for gene prioritization.