Cem Ertur and Antonio Musolesi, "Weak and Strong Cross-Sectional Dependence: a Panel Data Analysis of International Technology Diffusion", Journal of Applied Econometrics, Vol. 32, No. 3, 2017, pp. 477-503 SUPPLEMENTARY APPENDICES The supplementary appendices -- Ertur_Musolesi_WSCD_APPENDICES.pdf -- provide additional results. Supplementary Appendix A provides a detailed analysis of the time series properties of the main variables under investigation through application of second-generation panel unit roots tests that allow for cross-sectional dependence. Supplementary Appendix B provides additional estimation results by focusing on the issue of slope heterogeneity. DATA The data used are contained in the spreadsheet files data_Ertur_Musolesi.xls and longlat24.xls. The file data_Ertur_Musolesi.xls contains the data of the main variables for the 1971-2004 period. Our main source is the CHH data set. This data set is a balanced panel of 24 countries observed over the period 1971-2004. Our measures of TFP and domestic R&D capital stock come from this data source. The average number of years of schooling used to construct our measure of human capital is taken from Barro and Lee (2013). Finally, the distance between two countries is calculated as the spherical distance between capitals. The spreadsheet file longlat24.xls contains longitudes and latitudes used to calculate the spherical distance between capitals. The definition of the variables and the order of the countries are provided below The ASCII text files data_Ertur_Musolesi.txt and longlat24.txt contain the same data as their .xls counterparts. All four files are zipped in the file em-data.zip. Labels and definition of the variables country: country number year: year (1971-2004) ltfp: Total factor productivity in logs lrd: Total domestic R&D capital stock in logs lhc: Human capital in logs lwerd1: Foreign R&D capital stock in logs calculated using geographic distance lsfbiw: Foreign R&D capital stock in logs calculated using bilateral import weights as proposed by CHH lsflp: Foreign R&D capital stock calculated using bilateral import weights as proposed by Litchtenberg and van Pottelsberghe de la Potterie (1998) dum_g7: Dummy variable, 1 for G7 lrdg7: lrd*dum_g7 lrdnog7: lrd*(1-dum_g7) lwerd1g7: lwerd1*dum_g7 lwerd1nog7: lwerd1*(1-dum_g7) lhcg7: lhc*dum_g7 lhcnog7: lhc*(1-dum_g7) lsfbiwg7: lsfbiw*dum_g7 lsfbiwnog7: lsfbiw*(1-dum_g7) lsflpg7: lsfp*dum_g7 lsflpnog7: lsfp*(1-dum_g7) m: Imports as a percentage of GDP mlsfbiw: m*lsfbiw mlsflp: m*lsflp mlwerd1: m*lwerd1 mlsfbiwg7: m*lsfbiwg7 mlsfbiwnog7: m*lsfbiwnog7 mlsflpg7: m*lsflpg7 mlsflpnog7: m*lsflpnog7 mlwerd1g7: m*lwerd1g7 mlwerd1nog7: m*lwerd1nog7 Country Order 1 United States 2 Japan 3 West Germany 4 France 5 Italy 6 United Kingdom 7 Canada 8 Australia 9 Austria 10 Belgium 11 Denmark 12 Finland 13 Greece 14 Ireland 15 Israel 16 Netherlands 17 New Zealand 18 Norway 19 Portugal 20 Spain 21 Sweden 22 Switzerland 23 Iceland 24 Korea ESTIMATION The zip file em-programs contains the programs used to obtain the main estimation results presented in tables 3-6. It also contains the data formatted in a proper form to get these estimates. There are three subfolders: alpha, CCE, Spatial. The subfolder alpha contains the programs and data used to obtain the estimates of alpha, the exponent of cross-sectional dependence (table 3). The GAUSS program ERTUR_MUSOLESI_JAE_EXPONENT.prg allows the estimation of the bias-corrected version of alpha given by equation (13) in Bailey et al. (2015) for the main variables under study: ltfp, lrd, lhc and lwerd1. The file ERTUR_MUSOLESI_JAE_EXPONENT.prg uses the program code_ord_holm.src which was made available by Natalia Bailey, George Kapetanios, and M. Hashem Pesaran in the Journal of Applied Econometrics Data Archive at: http://qed.econ.queensu.ca/jae/datasets/bailey002/ The files ltfp_res.txt, lrd_res.txt, lhc_res.txt and lwerd1_res.txt contain the (reshaped) data in a matrix form N*T. The subfolder CCE contains the programs and data used to obtain the CCEP and CCEMG estimates (tables 4 and 5). The GAUSS code CCE_firstrow allows the CCE estimation of equations 16 to 20 in Ertur and Musolesi (2016) corresponding to the first row of tables 4 and 5. It uses the code that was made available by Takashi Yamagata and that can be found at: http://www.econ.cam.ac.uk/emeritus/mhp1/ppfiles/CCEgauss6_22Aug08.zip The files first.xls contains the data for such an estimation. This data set contains a subset of the variables contained in the file data_Ertur_Musolesi.xls and minor changes are needed to get the all the other results presented in tables 4 and 5. The subfolder Spatial contains the programs and data used to estimate the spatial error model (equation 24 in Ertur and Musolesi, 2016) with the QML approach by Lee and Yu (2010). The MATLAB program ERTUR_MUSOLESI_SEM_JAE.m allows the estimation of the spatial error model for all the specifications with the results presented in table 6. This program uses the following MATLAB codes - dexparc.m and distarc.m generate the weight matrix - sem_panel_FE_LY.m computes the spatial error model estimates - prt.m and prt_fe.m allow printing the results of estimation The file spatialti24.txt contains the data sorted by time first and then by country. The file longlat24.txt contains longitudes and latitudes. Note to Unix, Linux, and Mac users. The files em-programs.zip and em-data.zip contain both text and binary files. Do *not* use "unzip -a" to convert text files out of DOS format, since that will destroy the binary files. Please address any questions to: Antonio Musolesi Department of Economics and Management, University of Ferrara --Italy and SEEDS email: antonio.musolesi [AT] unife.it