Martin Feldkircher and Stefan Zeugner, "The Impact of Data Revisions on the Robustness of Growth Determinants - A Note on 'Determinants of Economic Growth. Will Data Tell?'", Journal of Applied Econometrics, Vol. 27, No. 4, 2012, pp. 686-694. INTRODUCTION Ciccone and Jarocinski (2010) show that inference in Bayesian Model Averaging (BMA) can be highly sensitive to small changes in international income data. In particular, they demonstrate that the importance attributed to potential growth determinants varies tremendously over different revisions of growth data. They conclude that 'agnostic' priors appear too sensible for this strand of growth empirics. In response, we show that the instability they found owes much to a specific BMA set-up: First, comparing the same countries over data revisions improves robustness. Second, much of the remaining variation can be reduced by applying an evenly 'agnostic', but flexible prior. This tutorial replicates the results provided in the note by Feldkircher and Zeugner. Here, we provide the data set ("pwt_replication.RData") as an R workspace, a file that replicates our results (fz_tutorial.R) as well as a short tutorial that guides the reader through the replication (fz_tutorial.Rnw). All the results in our paper have been computed using the R system for statistical computing (http://www.R-project.org/) and in particular the R package BMS. Both the R system and the strucchange package are freely available at no cost under the terms of the GNU General Public Licence (GPL) from the Comprehensive R Archive Network at http://CRAN.R-project.org/ We recommend executing the "fz_tutorial.Rnw" file by typing Sweave("YOURPATH/fz_tutorial.Rnw") in your R-console, where "YOURPATH" is the computer path where the file "fz_tutorial.Rnw" was saved on your desktop computer (e.g. Sweave("C:/projects/fz_tutorial.Rnw") ). This will produce the latex file "fz_tutorial.tex" that contains all results from the paper. Alternatively, the file fz_tutorial.R returns the same output on the R console: In R, run the command source("fz_feldkircher.R") to execute it. The first step -- as explained in fz_tutorial.Rnw -- is to load the data file "pwt_replication.RData" into your R-console. The data are stored in a list object called "dataList" containing three vintages of PWT data: PWT 6.0, 6.1, and 6.2. We proceed by showing how to arrange an index that sorts out the countries that are common across all the vintages: We finally show how to perform the BMA estimation for the three PWT vintages and various coefficient priors. Note that -- for the purpose of illustration -- we have defined an extremely small number of iterations for the BMA MCMC-sampler: burnNr=1000 (for the burn-in draws) and iterNr=5000 for 'counted' draws. For this reason, the produced figures will differ considerably from those in the paper. In the paper, 20 million burn-ins and 80 million subsequent iterations were used. Increase burnNr and iterNr to burnNr=200000 and iterNr=2000000 for results that approximate the paper. FILES R-files: * fz_tutorial.R....replicates the results of Feldkircher and Zeugner * fz_tutorial.Rnw...replicates the results thereby taking a 'step-by-step' explanation for the user. See above how to execute the file. Produces a tex file that contains the replication report. Data files:* pwt_replication.RData...contains data on three PWT vintages. The following files contain the same data as does pwt_replication.RData, but in text format: * pwt60t88: PWT 6.0 data for 88 countries, complemented with data as in Sala-i-Martin, Doppelhofer, Miller (AER 2004) * pwt61t84: PWT 6.1 data for 84 countries, complemented with data as in Sala-i-Martin, Doppelhofer, Miller (AER 2004) * pwt62t79: PWT 6.2 data for 79 countries, complemented with data as in Sala-i-Martin, Doppelhofer, Miller (AER 2004) PDF file: * fz_tutorial.pdf...Output of fz_tutorial.Rnw - contains a 'step-by-step' tutorial to replicate our results. All of the above files except pwt_replication.RData are zipped in the file fz-files.zip. Since they are ASCII files in DOS format, Unix/Linux users should use "unzip -a".