Jonathan Temple, "Robustness Tests of the Augmented Solow Model", Journal of Applied Econometrics, Vol. 13, No. 4, 1998, pp. 361-375. The file temple.zip contains two data files and five program files. All files are in DOS (CR/LF) format. File: testsol.dat The data are organised in a spreadsheet in space-delimited text format. Unless otherwise noted, the data are taken from Mankiw, Romer and Weil, 1992, "A contribution to the empirics of economic growth", Quarterly Journal of Economics, 107, 407-437. Variable Description MRWNO Identifying code used by Mankiw, Romer and Weil NSAM Dummy variable equals one for MRW's non-oil sample ISAM Dummy for MRW's intermediate sample OSAM Dummy for MRW's OECD sample AFRICA Dummy for sub-Saharan Africa, from Easterly and Levine (1997) LATINCA Dummy for Latin America and the Caribbean, from Easterly and Levine (1997) EASTASIA Dummy for East Asian countries (constructed by author) INDUST Dummy for industrialised countries (constructed by author) GDP60 Real GDP in 1960 GDP85 Real GDP in 1985 GDPG Growth rate (not the dependent variable) POPN Population growth INV Investment rate SCHOOL MRW's schooling variable LNGD Log(n+g+d) where n is population growth and g+d=0.05 LINV Log of investment rate LSCH Log of SCHOOL LGDP60 Log of 1960 GDP LGDP85 Log of 1985 GDP DY LGDP85-LGDP60, the dependent variable File: testsol.csv This contains the same data as testsol.dat, but in comman-delimited format. Files: reverse.txt, impose.txt, dstats.txt, imposec.txt, calcrsq.txt These are the programs used for the Klepper-Leamer measurement error diagnostics, written in the S-Plus statistics language. They can be read into S-Plus using the 'dget' command, for example: calcrsq <- dget("calcrsq.txt") Since the functions call on each other, you should give them the same titles as their filenames, after checking you do not already have any S-Plus objects with these names. The programs are described below. Ideally you should read this description in conjunction with Klepper, Kamlet and Frank, 1993, "Regressor diagnostics for the errors-in-variables model - an application to the health effects of pollution", Journal of Environmental Economics and Management, 24, 190-211. The first step in applying the diagnostics is to carry out the direct and reverse regressions. This is done by the function 'reverse' which takes a list of explanatory variables and the dependent variable as its arguments. For instance you might use: x <- c(x1,x2,x3) rev1 <- reverse(x,y) where x1, x2 and x3 are the regressors. The object rev1 will then contain the direct and reverse regressions, the g-statistic from Klepper and Leamer (1984) together with the indices of the two key f(i) [f(c) and f(e)], the bound M, and the associated d-statistics. If M<1 and the true R-squared must be constrained, this is done using the function 'impose', which is called using the explanatory variables, the dependent variable, and the chosen bound as arguments. The use of this function corresponds to step 4 in the Appendix to Klepper et al. (1993). If it is not reasonable to bound the true R-squared below M, then either of the two key f(i) will have to be bounded below their d(i) value. The d-statistics for a bound M can be calculated using the 'dstats' function, with x, y and the M bound as arguments. A bound on f(c) can be imposed using the 'imposec' function, where the arguments are x, y, M, the index i of the key f(c) and the bound u on f(c). The output of this function corresponds to that discussed in step 7 of the Appendix to Klepper et al. (1993). The last function, 'calcrsq' is called by the others and must be included. Note that the functions will cope with missing data, but not with large numbers of regressors. When the number of regressors becomes too large, the programs will stop with an error message, and will have to be modified: in particular, matrices should be initialised to be a larger size. This is left as a task for the user. Although I have checked the programs fairly thoroughly, they are applied at the user's own risk. I would be interested to hear from anyone who successfully applies this software, or who finds errors in it. Finally, I would be grateful if anyone making use of these programs in their own work acknowledges the source. Jonathan Temple, September 1997 ----------------------------------- Email jon.temple@economics.ox.ac.uk ----------------------------------- References Easterly and Levine, 1997, "Africa's growth tragedy: policies and ethnic divisions" Quarterly Journal of Economics, forthcoming. Klepper and Leamer, 1984, "Consistent sets of estimates for regressions with errors in all variables" Econometrica, 52, 163-183.