Ancestry and Long-Run Growth Reading Club: Chanda Comments
By Bryan Caplan
Areedam Chanda kindly sent me the following comments on my post on Chanda, Cook, and Putterman.
Thanks for choosing to discuss
our paper and for your positive reaction.
A few observations.
(a) In your view technology is
the most preferable variable followed by population density and urbanization,
with agricultural history and state antiquity coming last. Further, you observe
that technology does “well” in capturing persistence, while your least
preferable variables do the best, with density and urbanization doing worst.
The fact that urbanization and
density may not do as well is, because we believe that they are poorly
measured. Furthermore, for urbanization the small sample size makes matters
Moving on to the other three
variables, you prefer technology the most but note that your least preferred
variables do better. Having re-examined our results, my own reading is that
technology performs at least as well as, if not better, than state antiquity
and millennia since agriculture. Note that the simple correlations in our
paper indicate that technology, state history and agriculture are very strongly
correlated (compared to their correlations with urbanization and population
I should also clarify that state
history is a “stock” variable that aggregates a measure of the existence of the
state over 50 year periods from 1CE to 1500CE. We apply a 5% decay for past
values. In other words, it is a stronger indicator of the presence of a state
in the centuries closer to 1500, than a 1000 years earlier.
(b) You doubt the Malthusian
theory of limited GDP per capita differences and suggest that slavery is an
indication that living standards are likely to have been above subsistence.
Slavery was not uncommon in pre-industrial times though it did vary by country
or empire. It would be hard to say how much it would be reflected in GDP
per capita differences. If most of the population was still engaged in
livelihoods that only provided a subsistence income, this may not have mattered
so much. In any case, at least indirectly, variations in the existence of
slavery should reflect differences in the power and organization of the state,
or the technology that helped capture and retain slaves. In that sense, I think
variables such as technology and state history are useful measures when GDP per
capita is not available.
Having said, that I now shameless
plug my earlier research with Louis that was published in the Scandinavian
Journal of Economics in 2007. In that paper, we extrapolated GDP per capita in
1500 based on population density and urbanization. Needless to mention, it is
certainly worth exploring further by adding state history and the measure of
again for choosing to cover the series of papers on ancestry and long run development.