Gravity models are used to understand intra- and international trade flows. Trade costs play a central role in these models, but are not clearly observable. In order to infer these costs, different estimation methods exist. The aim of this paper is to investigate these methods on systematic patterns in their predicted trade costs. By applying the methods to one dataset, the resulting trade cost estimates become comparable. For a given trade elasticity, the inverse gravity framework from Novy (2013a) is found to predict lower values than ratio gravity, used for example by Simonovska and Vaugh (2014). However, when moderating the impact of outliers, inverse gravity produces lower estimates.