Lung-fei Lee and Jihai Yu, "Identification of Spatial Durbin Panel 
Models," Journal of Applied Econometrics, Vol. 31, No. 1, 2016, pp. 
133-162.

The data files contain three variables and four spatial weights
matrices.

Y is the log value of GDP per unit of labor based on purchasing power
parity and is computed in constant 2005 international dollars. It has
26 countries and 8 time periods, where we stack the data by countries
first, then by time.

X has two columns. The first column is log value of (N+0.05), where N
is the annual working-age population growth rate.  The second column
is log value of investment share (%) of PPP-converted GDP per capita
at constant 2005 international dollar. X has 26 countries and 7 time
periods

W_geo is the spatial weights matrix based on geographic information.

W_Import is the spatial weights matrix based on bilateral imports.

W_Export is the spatial weights matrix based on bilateral exports.

W_ExIm is the spatial weights matrix based on sum of bilateral exports
and imports.

Elements of each spatial weights matrix are obtained by exp{d} where d
is the geographical (munus distances between country capitals) or
relevant trade volume.

All data files are comma-separated ASCII files in DOS format. They are
zipped in the file ly-data.zip. The actual filenames are:

  w_exim.csv  w_export.csv  w_geo.csv  w_import.csv  x.csv  y.csv

Unix/Linux users should use "unzip -a".

Data source: 

Penn World Table 7.1 for Y and X

World Trade Flows Database of Feenstra et al. (2005) for bilateral
import and export flows.


Reference:

Feenstra, R., Lipsey, R., Deng, H., Ma, A., Mo, H., 2005. World trade
flows: 1962-2000. NBER Working Paper Series, 11040.