To understand how to promote growth, alleviate poverty, and reduce inequality, we need measures of wealth that are both disaggregated over space and available over long time periods. We need data that is disaggregated over space because poverty and inequality depend on the distribution of wealth within countries, as well as between countries. We need data over long time periods, because many of the important policies that might help achieve these social goals have effects over very long time horizons. Reliable sub-national data on wealth only go back to 1990 for many developing countries, meaning that many important questions cannot be answered using existing data. This research programme will construct the first sub-national measures of wealth for more than 60 countries across the developing world between 1939 and 1990. To construct our measures of wealth, we will adapt and apply novel machine learning techniques for extracting information from high-resolution imagery to a novel digitized, georeferenced dataset created by our team from a previously unstudied archive of 1.6 million aerial photography images: essentially, a historical "Google Earth". Exploiting these rich and novel data, we will provide evidence on important unresolved questions about the determinants of long-run growth: How does institutional change shape a country's growth trajectory? Do investments in infrastructure lead to growth? And what are the long-run consequences of extreme climatic events?