Recently large scale transcriptome and proteome datasets for human cells have become available. A striking finding from these studies is that the level of an mRNA typically predicts no more than 40% of the abundance of protein. This correlation represents the overall figure for all genes. We present here a bioinformatic analysis of translation efficiency - the rate at which mRNA is translated into protein. We have analysed those human datasets that include genome wide mRNA and protein levels determined in the same study. The analysis comprises five distinct human cell lines that together provide comparable data for 8,170 genes. For each gene we have used levels of mRNA and protein combined with protein stability data from the HeLa cell line to estimate translation efficiency. This was possible for 3,990 genes in one or more cell lines and 1,807 genes in all five cell lines. Interestingly, our analysis and modelling shows that for many genes this estimated translation efficiency has considerable consistency between cell lines. Some deviations from this consistency likely result from the regulation of protein degradation. Others are likely due to known translational control mechanisms. These findings suggest it will be possible to build improved models for the interpretation of mRNA expression data. The results we present here provide a view of translation efficiency for many genes. We provide an online resource allowing the exploration of translation efficiency in genes of interest within different cell lines (http://bioanalysis.otago.ac.nz/TranslationEfficiency).