Climate models have persistently simulated greater satellite-era tropical tropospheric warming than observations. As a result of this difference, it has been suggested that models are overly sensitive to anthropogenic forcing and that future projections may be biased. In addition to biases in climate sensitivity, several other factors may contribute to this model-versus-observed difference including deficiencies in the prescribed model forcing, a large contribution of internal variability to the observed trend, and biases in observational records. There is evidence that all these issues contribute to greater simulated than observed satellite-era tropical tropospheric warming. We estimate the impact of each of these factors using machine learning, multimodel analysis, and a suite of complementary measurements. The controversial history of the tropospheric temperature record provides an important reminder that each of these factors: forcing, model response, internal variability, and observational accuracy need to be considered when confronting Earth System Model Trends with observations.