important than tropical location. Bloom et al. (1998) examine data for 77
African and non-African countries from 1965 to 1990 and find that openness,
institutional quality and public savings have a positive and significant impact
on economic growth. Their results show that being located in Africa is impor-
tant and that Africa experienced on average 2.1 per cent slower growth than
non-African countries between 1965 and 1990. They find the key features
accounting for this slower growth are: the greater share of Africa’s land area
being in the tropics; and the large proportion of Africa’s population being
located more than 100 km from the coast. A third study by Warner (2002) finds
that the most important geographical variables are tropical location, remote-
ness from the coast or a river, and having mountainous terrain.
There are still significant statistical problems. Factors of production can
migrate, and this has implications for statistical work. If labour and capital are
mobile they would tend to leave areas of poor geography or low incomes.
Over time factors would continue to migrate until the benefits of being
located in areas of ‘good geography’ have been competed away, implying that
there would no longer be any identifiable statistical relationship between
geography and income. It is possible to overcome this problem. As many
institutions tend to be reasonably similar within countries, regional data
provides a way to test for the effects of geography and climate holding those
institutions constant. As factor mobility is greater across regions than coun-
tries we would expect to see first, stronger effects of geography on output
density (the amount of output produced in a particular area) than output per
capita and second, smaller effects of geography on output per capita in
re gional data. Regional data on large and diverse nations (Brazil, China, India
and US) finds that both of these hypotheses are correct and shows that geog-
ra phy has a significant effect on output density using regional/sub-national
data (Warner, 2002).
Early work on geography by Sachs and others used single countries as data
points so that India and the US were considered ‘coastal’ despite their size and
diversity and the fact that large areas of land were located far from the coast.
The crudity of this early work has been overcome by using ‘gridded data’. This
method divides the world into almost 20,000 data points rather than the 150+
country observations previously used. This approach allows us to use more
finely tuned geographic data (including climate, location, distance from
markets or seacoasts, and soils). The results confirm the importance of geogra-
phy and find a significant positive link from temperate, climate and costal loca-
tion to high economic densities (Nordhaus, 2006). Economic density is here
calculated as output per capita per square kilometre.
In Sub-Saharan Africa there is a very high concentration of land in the trop-
ics. Only 19 per cent of the entire continent’s population are within 100 km of
the coast, a quarter of the population are in landlocked countries, and Sub-
Saharan Africa is far from core European markets. In India the mass of popu-
lation is in the landlocked North-Central Ganges valley, a long way from the
coast,and India is partly tropical. The US has a relatively high (38 per cent)
234 Patterns and Determinants of Economic Growth