Many economic historians have constructed their own panels. An important example is the panel of city-, county- and state-level New Deal spending, economic activity, and demographic data constructed by Price Fishback and his co-authors. For example, Fishback et al. (2007) examine the impact of New Deal spending on infant mortality, non-infant mortality, and general fertility rates in major U.S. cities between 1929 and 1940. They include controls for city characteristics, city and year fixed eﬀects, and make use of instrumental variables (voting patterns in years prior to the Great Depression, and Congressional committee assignments, for example) that helped determine the allocation of New Deal spending. They conclude that about $2 million (measured in year 2000 dollars) in additional relief spending was associated with a reduction of one infant death, half a homicide, one suicide, 2.4 deaths from infectious disease, and one death from diarrhea in each large urban area.
Finding a convincing instrumental variable can be very challenging. The researcher needs to find a factor that helped drive the economic events but did not arise simultaneously from within the economic system. When it comes to policy analysis, this can be almost impossible because political decisions are so intricately intertwined with economic forces.