Do note the latter part of the last sentence, but the entire thesis is interesting:
This paper utilises a unique, purpose-built panel dataset on prominent authors in the UK and Ireland born 1700–1925 to estimate the productivity gains associated with agglomeration of an industry with few capital requirements and no apparent need to cluster geographically. I find the average author experiences productivity gains of 11.94% per annum when residing in London, the only major literary cluster – a gain not associated with living in any of the minor literary clusters. I find evidence of negative selection with respect to productivity, indicating the results are not driven by the self-selection of highly productive authors to London. I find heterogeneity of returns to living in London by birth cohort and Impact Index quartile (a measure of author quality) and that the cohorts who receive the greatest gains from locating in London are those for which there is the strongest evidence of negative selection with respect to productivity.
That is by Sara Mitchell in the Journal of Urban Economics, via the excellent Kevin Lewis.