Marcos Sanso-Navarro, María Vera-Cabello, and Domingo P. 
Ximénez-de-Embún, "Human Capital Spillovers and Regional Development", 
Journal of Applied Econometrics, Vol. 32, No. 4, 2017, pp. 923-930

Building on their theoretical model, Gennaioli et al. (2013, GLLS
hereafter) study the influence of geography, natural resource
endowments, institutions, culture and human capital on regional
development. With this aim, these authors construct a new database by
collecting cross-sectional information at the highest level of
administrative division or its closest counterpart. The dataset has
an extensive coverage: 1,569 regions from 110 countries which, taken
together, represent 74% of the world's surface and 97% of its GDP.
This newly constructed database is publicly available at Professor
Shleifer's website:

http://scholar.harvard.edu/files/shleifer/files/regions_data_web_july2012.xls

An exhaustive description of the data can be found in the paper:

Nicola Gennaioli, Rafael LaPorta, Florencio Lopez-de-Silanes, and
Andrei Shleifer (2013) "Human Capital and Regional Development",
Quarterly Journal of Economics, 128(1), 105-164.

The variables considered in our replication exercise are (Table IV in GLLS):

Income per capita, in natural logarithms (lnRegGDP)
Years of education (YearsEd)
Population, in natural logarithms (POPij)
Temperature (temp)
Inverse distance to coast (invdistcoast)
Oil production per capital, in natural logarithms (lnoilpc)
Institutional quality (Xhat)
Trust in others (trustothers)
Number of ethnic groups, in natural logarithms (NumberEthnicGroups)
Years of education 65+ (YrsEducOld)

These variables are included in the file svx-data.csv (144 KB), an
ASCII file in DOS format.

Following expression (8) in our paper, we incorporate the spatial
dimension into the data set constructed by GLLS with a shapefile
containing regional boundaries. The main source from which this
geospatial information has been extracted is the GADM database of
global administrative areas for countries and lower levels
(www.gadm.org). Alternative sources of information were used only for
five countries (see Footnote 1). The administrative division for
regions in this database does not coincide with that considered by
GLLS. In order to match the two data sets, lower-level administrative
divisions in the GADM database were merged for some countries (see
Footnote 2). It was also necessary to remove regions for which no data
were available and to reshape some spatial units. Finally, individual
shapefiles were merged into a single one.

The geographical information is included in th file svx-regions.csv
(855 MB), an ASCII file in DOS format. Both CSV files are zipped in
the file svx-data.zip. Unix/Linux users should use "unzip -a". The
svx-regions.csv file uses the well-known text (WKT) representation of
coordinate reference systems to describe regional boundaries.

The files svx-regions.csv and svx-data.csv can be joined using a
geographic information systems (GIS) software through the "RegionID"
field.

Marcos Sanso-Navarro
marcossn [AT] unizar.es