Paul Johnson, Steven Durlauf and Jonathan Temple 1121that ain’t so.” This point applies with some force to almost the entirety of the
empirical growth literature. It is well known that some claims have not survived
later scrutiny, with Levine and Renelt (1992) an especially famous demonstration
of the lack of robustness of some early results. To take a more recent example, few
papers in economics have been as directly influential on policy debates as the study
of foreign aid and growth by Burnside and Dollar (2000). Yet its claims have been
vigorously contested in a series of subsequent papers, and a careful reading of those
papers might suggest that the key hypotheses cannot be reliably tested given the
limitations of the available data.
Such concerns have not prevented the proposals for growth determinants from
increasing with every passing year. Indeed, the number of proposed determinants
is now similar to the number of countries for which data are available. It is hard
to believe that all these variables are central, yet the range of possibilities, the
small number of countries that can be studied, and the arbitrary choices that are
often involved in estimating a specific model, all conspire to make learning about
economic growth unusually difficult.
These and other difficulties have prompted the field to evolve continuously, and
to adopt a wide range of methods. We argue that the statistical tools that have
been applied to growth questions are sufficiently rich that they collectively define
a distinct field, growth econometrics. This chapter provides an overview of the cur-
rent state of this field, updating and revising our earlier survey in Durlauf, Johnson
and Temple (2005). The chapter will survey the body of econometric and statistical
methods that have been brought to bear on growth questions, and provide some
assessments of the value of these tools. In keeping with the rest of this book, the
focus will predominantly be on the application of econometric methods and tech-
niques and their interpretation, rather than attempting to summarize substantive
findings. For an earlier survey with a greater focus on substantive findings, see
Temple (1999), and for an earlier evaluation of the econometrics of growth, see
Durlauf and Quah (1999).
The techniques that have been used in growth econometrics largely reflect the
specialized questions that naturally arise in this context. Consider the identifica-
tion of empirically salient determinants of growth when the range of potential
factors is large relative to the number of observations. The associated model
uncertainty is one of the most fundamental problems facing growth researchers.
Individual researchers, seeking to communicate the extent of support for particular
growth determinants, typically emphasize a single model (or small set of models)
and then carry out inference as if that model had generated the data. But there are
usually other models that have equally strong claims on our attention, and hence
the standard errors will often understate the true degree of uncertainty about the
parameters. Moreover, the decision to report one model rather than another is
often somewhat arbitrary. The need for a more systematic and objective approach,
one that properly accounts for model uncertainty, naturally leads to Bayesian or
pseudo-Bayesian approaches to data analysis.
Bayesian approaches seem especially natural for growth econometrics, given the
paucity of the available data. This represents a major constraint on the scope for