In this dissertation I include two studies where the identification strategy is based on a natural experiment. Firstly, a change in government policy, pension eligibility age for men in South Africa. Secondly, the introduction of an interface language on the South African Google Search website, as a spillover of that translation work being done for Botswana. In both cases I use the National Income Dynamics Study, the most comprehensive panel data set on South Africa. Since it will be hard to find more relevant data, a full replication - using new data - will prove to be difficult. As such, I believe it is all the more important for research to be as transparent as possible, making it at least as reproducible as can be. The third and fourth chapter were written together with Victor Kummritz. We take a new algorithm for trade flow decomposition and implement this is the R language. The functions that we created for this implementation were included in an open source R package which was published on the Comprehensive R Archive Network (CRAN). Chapter 3 is the paper that accompanies this R package. In Chapter 4 we use the package and apply this to a previously unavailable dataset TiVa, with a specific focus on developing economies.