12
concentrated geographically and local resource dependence is more likely to be exogenous as
it is decided by the national oil company, Petrobras. It turns out that oil discoveries and
exploitation do not affect non-oil GDP very much, albeit that in line with the Dutch disease
hypothesis services expand and industry shrinks somewhat. But they do boost local public
revenue, 20-25 percent (rather than 10 percent) going to housing and urban development, 15
percent to education, 10 percent to health and 5 percent on welfare. Interestingly, household
income only rises by 10 percent, mostly through higher government wages. The lack of
migration to oil-rich communities also suggests that oil does not really benefit local
communities much. The evidence for Brazil thus offers support for the Dutch disease
hypothesis, but also to waste in local government and corruption (see section 3.3).
3.2. Temporary loss in learning by doing curbs economic growth
A declining traded sector is the appropriate market response to a resource windfall. In itself
this does not justify government intervention, since it is optimal to specialize in one’s
comparative advantage. Why are resource windfalls then perceived to be a problem? One
popular answer is that the traded sector is the engine of growth and benefits most from
learning by doing and other positive externalities, hence non-resource export sectors
temporarily hit by worsening competitiveness are unable to fully recover when resources run
out. This can be demonstrated in a two-period, two-good Salter-Swan model where learning
by doing is captured by future productivity of the traded sector increasing with current
production of traded goods (van Wijnbergen, 1984a) or with cumulative experience
(Krugman, 1987).5 If manufacturing rather than agriculture enjoys learning by doing and the
income elasticity of demand for agricultural goods is less than unity, shifting from
manufacturing towards agriculture curbs growth in an open economy (Matsuyuma, 1992).
Similarly, if human capital spill-over effects in production are generated only by employment
in the traded sector and induce endogenous growth in both traded and non-traded sectors,
natural resource exports lower employment in the traded sector, hamper learning by doing and
thus stunt economic growth (Sachs and Warner, 1995; Gylfason et al., 1999).
With perfect international capital mobility and no specific factors of production, the
wage, the relative price of non-traded goods and the capital intensities in the traded and non-
traded sectors are pinned down by the world interest rate. Higher resource revenue then
induces gradual movement of labour from the traded to the non-traded sector. This reduces
learning by doing and thus lowers the rate of labour-augmenting technical progress, so that
the resource boom permanently lowers the rate of growth. One can show that non-resource
GDP falls on impact after a resource discovery if the traded sector is capital-intensive (see