4 A sectoral analysis
is that the former set of firms are financing investment mainly via the domestic banking
system, while the later has other financial instruments available, such as issuing equity or
commercial paper, or borrowing on the international capital market. It is often found that
the strength of the link between financial development and output growth differs substantially
between these two groups. This difference across sectors is quite pronounced in middle income
countries and emerging markets, but less prevalent in industrial economies.
The data set of Hoffmann (1965) includes detailed information on the sectoral aggregate ac-
counts of Germany and allows us to do such a decomposition. We focus on six main subsectors
of NDP, the industrial sector, mining, agriculture, trade, transportation and services.
Figure 4 shows the impulse response functions that were generated from bivariate VARs,
including the respective measure of output and our bank lending variable. As in the previous
section, we generate the impulse response functions from a Cholesky decomposition, where
the bank lending variable is ordered at the second position in the VAR.
We find that in all sectors there is a positive reaction of output to an unexpected shock
in bank lending. In all sectors, except for the trade sector, this reaction is also statistically
significant at the 5% level. However, the variance decomposition in table 5 shows that the
shocks coming from the banking system are of quite different importance for the various
sectors of the economy. The insignificant trade sector is least affected by banks. Shocks from
the banking system explain only up to 4.9% of the forecast uncertainty of the trade sector.
Interestingly, shocks from the banking system also show little impact on the industry and
mining sectors, with values of 9.3% and 5.7%. This finding is interesting, as it challenges
the conventional wisdom that the industrial revolution was substantially accelerated by the
parallel development of the banking system. On the other hand, we find that the sectors
agriculture (up to 17.9%), transportation (up to 25.5%) and services (up to 25%) were most
affected by shocks in the banking system.15
The structure of German exports - that was also recorded, although not on an annual
basis, by Hoffmann (1965) - suggests that the industry sector was indeed the most tradable
in Germany. In 1910-13, final goods had the largest share in total German exports - textiles
(12.3%), metal and machinery (21%) as well as chemicals (9.9%) - followed by raw materials
15Note that the significance level of the variance decomposition is very low in general. Our robustness tests
in the following section will show, however, that the contributions of banks to the forecast error variance
are also significant at conventional levels, when using the alternative banking indicator.
11