1 Motivation
1 Motivation
In this paper we re-evaluate the hypothesis that bank lending was a key factor in the growth
process in 19th century Germany and that it has been instrumental in financing the industrial
revolution. This hypothesis has been developed, among others, by the influential economic
historian Alexander Gerschenkron (1962). This conventional view has been adopted by most
researchers and has triggered a literature that discusses the benefits of close bank-firm rela-
tionships that were said to be typical of Germany at the time. A survey on papers arguing
along these lines is given for instance in Guinnane (2002). In a notable exception, however,
Edwards and Ogilvie (1996) challenge this view and point out that large universal banks that
serviced the big industrial firms contributed only a small fraction to total bank lending. They
argue that universal banks were primarily engaged in organizing the issuance of new shares,
but hardly contributed to financing long-term investment by credit.
In this paper, we employ a new data set to reinvestigate whether there has been a positive
effect of bank lending on growth and whether indeed the industrial sector - or possibly other
sectors in the economy - benefited most strongly from the development of domestic credit
in Germany. This data set was initially recorded by Walter Hoffmann (1965) for the sample
period of 1870-1912 and includes a detailed sectoral disaggregation of output. It therefore
allows us to trace the effect of the rapid increase in bank lending on net domestic product, as
well as on the sectoral structure underneath it.1 In our paper, we focus on the main subsectors
manufacturing, mining, agriculture, transportation, trade and services.
In the empirical analysis, we use a VAR framework to trace the effect of an unexpected
shock in aggregate lending on domestic product and its subsectors. From the VAR coefficients,
we generate impulse response functions in two different ways. On the one hand, we use
generalized impulse response functions. These can be computed without prior knowledge of
the contemporaneous causal relationships among the variables. On the other hand, we use
a Cholesky decomposition that was proposed by Tornell and Westermann (2005) and that,
using an appropriate ordering, can be interpreted as structurally identified in the context of
a theoretical two-sector growth model with credit market imperfections. As output, in the
1 In addition to the historical interest in the German industrial sector, the importance of sectoral information
when analysing the effects of financial deepening on growth, has been emphasized, among others, by Rajan
and Zingales (1998) and Tornell and Schneider (2004), as aggregate measures on output often mask deep
asymmetries in sectoral output dynamics.