That is the title of a new paper by Francois Facchini (University of Paris) forthcoming in the Atlantic Economic Journal. The argument of the paper is that econometrics is not going to ultimately explain (at least on its own) the size of the public sector, but it also contains a useful summary on the history of thought. Here is the abstract:

The purpose of this paper is to present and assess the literature about the determinants of public spending. Its originality is the adoption of a methodological perspective. Does econometrics allow economists to discover universal constants for public spending or is it only another way of writing the history of public finance? The economic theory of the size of government includes 23 explanations and 78 explanatory variables. The size of government reflects the preferences of citizens (demand model), the power of politicians and bureaucrats to impose their interests against citizens’ interests (supply model) and the constitutional design that governments face in raising revenue. Nonetheless, the great instability of econometric tests shows, all the difficulties of the science of public finance in finding a constant and discovering real causality between these variables. The analytical consequence of this result is the great futility of the search for a general law of the dynamic of public spending. This futility can be interpreted as a consequence of the complexity of political exchange. The size of the state is a multi-causal phenomenon. The prescriptive consequence is that it is not possible to use quantitative analysis to defend a form of “causal manipulationism” and predict the timing of state reforms and a political strategy to reduce the share of public spending in production. Econometrics is only one way of reporting the history of public finances.

This figure is also quite interesting, it shows concepts studied, the number of studies that confront them, and how frequently they seem to conform to theoretical expectations. Political fragmentation is the most commonly accepted hypothesis at 84%, while the base size of the economy as measured by pre-tax income has been the least likely to conform to expectations (40%):