The New Growth Evidence

The New Growth Evidence
by
Jonathan Temple
Journal of Economic Literature
March 1999
The Outline
•
•
•
•
•
The evolution of the growth literature
Six important empirical questions
Methodology and Complications
The Evidence
Conclusion
The Literature
• Patterns of economic growth and
development important question but ignored
until recently. Why?
– No reliable cross-country data;
– Technology important but no adequate theory that
would explain its evolution.
• Last two decades progress on both fronts.
Six Empirical Questions
• The world income distribution.
• Is there convergence as Solow model predicts?
• Are there diminishing returns to K and H? How
fast do they kick in?
• Factor accumulation vs. Technological Progress?
• Explaining the long Run differences in growth.
• The Long Run equilibrium? How will the WID
look like in 2155?
Empirical Methodology
• Historical case studies
• Cross-country growth regressions
growthi = constant + b*Xi + ei
• Growth accounting
Dy/y = DA/A + aDK/K + b)DH/H
+(1-a-b)DL/L
Issues & Complexities
• PPP-adjusted GDPs
– Prices vary across countries  a $ does not afford
same standard of living everywhere.
• Data problems (availability, accuracy and
measurement)
• Outliers
– Visual inspection and elimination
– Cook’s test
• Model uncertainty
– Robustness checks
Issues & Complexities (cont.)
• Endogeneity
– Y  X (solution: lag X)
– But, if Z  X and Z  Y, then lagging won’t help
– Instrumental Variables (IV): find some other Z that is
related to X but not to ei. Not always easy.
• Measurement error
– In general, may create attenuation bias.
– Not always, though…
What Have We Learned?
• Table 1: confirms the wide disparities in the
world income distribution (WID).
• There does not seem to be “convergence” as
Solow model predicts. Figure 1.
• What about conditional convergence?
The Evidence (cont.)
• Conditional convergence
– Most studies find that y converges to its steady state at
the rate of 2 % per year. This is consistent with
Decreasing Marginal Product (of K and H).
– If there is conditional convergence, what does it say
about factor accumulation vs. productivity (or the role
technical change)?
The Evidence (cont.)
– Mankiw, Romer, and Weil (1992) show that 80 % of the
variation in y across countries can be accounted for
with differences in ik ih and n.
– But their approach has its problems.
– Klenow and Rodriguez-Clare (1999) look at primary
enrollment rates and find that it is more like 50 %.
– Some argue that Japan and East-Asian NICs cannot be
solely explained by factor accumulation. Hsieh (2002).
The Evidence (cont.)
• The World Technology Frontier
– Theory predicts that technologies should be
transferable across countries. If so, what should
income per capita trends look like?
– Indeed, Evans (1996) finds that the growth of
income (among rich countries) stays within
some bounds.
The Evidence (cont.)
• What drives this convergence?
– If A’s vary across the countries, then DMPK
and DMPH cannot be driving it alone; it also
has the be A transfers across countries.
– Ventura (1997) has shown that if countries trade
with one another, then the DMPK property
applies at the global level; if must be A
transfers that drive convergence.
The Evidence (cont.)
• Inputs and Growth
– K accumulation and Growth:
• Some supporting evidence but remember theory
says effect should be temporary.
• Investment is endogenous to the environment; Y I
or Z Y and Z  I. Thus, results may be biased.
– H accumulation and Growth:
• MRW find that Y = AK1/3H1/3L1/3 is about right.
The Evidence (cont.)
• Since then, other have raised questions about this
large role.
• Part of the problem: H is much broader than school
enrollment; it covers experience, training, health,
LE, etc. Another problem:
– Proxies for H: years of schooling (Y  h)
enrollment rates (stock vs. flow?)
• Micro studies could be helpful here. How?
Mincerian regressions
The Evidence (cont.)
– R&D and Growth:
• Why is R&D important? R&D  DA/A > 0.
• Microeconomic studies support that the returns to
R&D are significant at the firm level.
• At the macro level, there are spillover effects (i.e.
externalities +/-). Need to account for them.
The Evidence (cont.)
• R&D productivity has fallen in the OECD in the
post-WWII era despite improvements in I, H,
openness to trade, etc. (Jones, 1995a, 1995b).
• Wider Influences on Growth
– Population Growth:
n  Dy/y ?
• A slight negative effect is the consensus, but should
we worry about Dy/y  n ? Not so much. It would
be a bigger issue with n  y .
• Where does this negative effect come from? High n
seems to lower k, h and labor productivity. The last is
somewhat of a puzzle, but Hanushek (1992) shows
that high n  lower student achievement.
The Evidence (cont.)
• Sometime people test
fertility  Dy/y ; with
that Dy/y  fertility a bigger concern? Why?
– Trade and Growth:
• Many of the typical problems: outliers, endogeneity,
specification errors, etc.
• How to quantify trade friendliness of the economy is
another problem.
– Two standard measures are OPENNESS and BLACK
MARKET PREMIA.
– Sachs and Werner (1995), updated by Wacziarg and HornWelsh (2004), create dummies for trade liberalization; find
strong evidence for the role of trade lib. On growth.
The Evidence (cont.)
• Causality a big deal here; Rodrik (1998). So it may
pay to look at natural experiments; Ben-David
(1994) EU integration.
– Finance and Growth:
• Early growth theories ignored it; assumed financial
development would emerge in a growing economy
almost as a “side show.” Recent evidence rejects this
(non)treatment; Levine (many papers). Of course,
gotta deal with reverse causality properly.
The Evidence (cont.)
– Government and Growth:
• A charged issue but evidence mixed at best.
• Hall and Jones (1997) do show that higher G 
lower Y. But this is not about growth; it is about level
of Y. Endogeneity is an important issue in these tests
but in Hall-Jones spec, this biases results to
understate the negative effect of G on Y.
– Infrastructure and Growth:
• Telephone networks and electricity capacity seem to
boost growth. Easterly and Rebelo (1993) show that
the gov’t share in I in public transport and
communication also help.
The Evidence (cont.)
– Inequality and Growth:
• Evidence suggests a robust negative link; one
exception is Forbes (2000) but she studies mediumrun growth.
• But why the negative link?
– Traditional argument was Political Economy, Persson and
Tabellini (1994).
» Easy to test: does inequality raise taxation?
» Evidence not supportive.
– Other channels: fertility, education, politico-social
stability.
The Evidence (cont.)
– Polity and Growth:
• Difficulties in measurement.
• Some focus on democracy and growth; evidence is
weak but Barro (1997) suggests an invereted-U
relationship. Why?
– Growth and Welfare:
• Not clear; Easterly (1997) suggests the effect on
welfare is mixed. Dollar and Kraay (2003) suggest
growth helps the poor.
Conclusion
• Poor countries are not catching up to the
rich ones. WID has become more polarized.
• Countries converge to their own steady
states but with lots of uncertainty. This
suggests that both DMPK/DMPH and
differences in technology adoption play a
role in convergence.
Conclusion
• Solow model factors help to explain growth
but many others not in the model also have
been found to play roles (i.e. finance,
inequality, R&D-driven technological
change).
• Big government? Openness to Trade?
Democracy? Jury is still out…