The impact of data revisions on the robustness of growth determinants—a note on ‘determinants of economic growth: Will data tell?’

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. (2010) show that inference in Bayesian model averaging (BMA) can be highly sensitive to small changes in the dependent variable. In particular they demonstrate that the importance of growth determinants in explaining growth varies tremendously over different revisions of Penn World Table (PWT) income data. They conclude that 'agnostic' priors appear too sensible for this strand of growth empirics. In response, we show that the instability found owes much to a specific BMA set-up: the variation in results can be considerably reduced by applying an evenly 'agnostic', but flexible prior. Keywords: Bayesian model averaging, Growth determinants, Zellner's g prior, Model uncertainty. JEL Classifications: C11, C15, E01, O47. * The opinions in this paper are those of the authors and do not necessarily coincide with those of the Oesterreichische Nationalbank. We would like to thank Terms of use: Documents in

Habit formation in consumption: A meta-analysis

“…As a robustness check, we also study alternative prior setups. To this end, we employ the benchmark g-priors for parameters suggested by Fernandez et al (2001) along with the beta-binomial model prior for the model space, which gives each model size equal prior probability (Ley & Steel, 2009); we also use the data-dependent hyper-g prior suggested by Feldkircher & Zeugner (2012), which should be less sensitive to noise in the data. the posterior model probability (that is, how well the model fits the data relative to its size).…”

Growth econometrics for agnostics and true believers

“…This is examined by Feldkircher and Zeugner (2012), who find that changes in the sample play an important role, not least because the countries which move in and out of the sample tend to be African ones that are especially likely to be outliers. Nevertheless, even after restricting attention to a fixed sample, some instability remains.…”

Measuring the Income Elasticity of Water Demand: The Importance of Publication and Endogeneity Biases

“…Since we use the bma package in R (Feldkircher & Zeugner, 2012), our BMA does not estimate all of the 2 32 possible combinations of models but uses Markov Chain Monte Carlo samplers that propose the candidate models to be estimated (estimating all of the models would take several months). The estimated BMA coefficients, posterior means, are averages of the coefficients across all of the models, weighted by the posterior model probability.…”