growth. More generally Levine and Renelt (1992) found no robust relation
between growth and the ratio of total government expenditure to GDP, govern-
ment consumption expenditure, capital formation, or educational expenditure.
Education has been understood as ‘human capital’ in growth analysis.
Human capital can directly raise the productivity of labour or facilitate the
absorption of new technology and so raise TFP. The results from empirical
testing remain poor and ambiguous. Pritchett (2001) finds a negativecorrela-
tion between higher school enrolment and TFP growth in developing coun-
tries, a paradoxical finding due to an incentive structure commonly found in
developing countries that diverts talented individuals to privately profitable
(they offer high wages) but socially wasteful activities and to policies that
reduce the demand for skilled labour (such as import tariffs that hinder the
importation of advanced technology from abroad). As has been more memo-
rably expressed, if it is more profitable to be a pirate than an engineer, people
will be pirates and the economy will suffer. Bils and Klenow (2000) find
evidence of reverse causation: economic growth increases the expected return
from and so incentives to acquire education. These ambiguities are reflected in
empirical studies of the education–growth link in Pakistan. Khan (2005) finds
positive and significant results linking literacy rates, average years of school-
ing, and gross secondary school enrolment to GDP growth. Iqbal and Zahid
(1998) find various measures of education including primary, secondary and
high school enrolment,are either insignificantly or negatively related to GDP
growth. Khan (2005) finds that education expenditures have no clear link with
TFP growth.
A theme running throughout this literature is the problem of finding a meas-
ure of human capital (Mankiw et al., 1992:418–19). Education investment
includes direct spending (by the individual,family and state) and indirect cost
(forgone earnings). Much schooling is provided free at the point of use and there
is an inherent difficulty in measuring the output or productivity of teaching and
hence in constructing any numerical measure of ‘education investment’. A
common measure used in much statistical work is the share of the working-age
population in secondary school. This fails to measure the quality of education
(influenced by teacher absenteeism), those instances where enrolment rates
bear little resemblance to actual children in school being educated or the learn-
ing on the job that takes place in the workforce. Hanushek and Woessman
(2009) in recent work have used internationally comparable test data to try and
measure the qualityof education. This data, though avoiding many of the prob-
lems with earlier data, is difficult and expensive to collect so is not yet available
for a particularly broad cross-section of countries. Many of the benefits from
education are non-material, such as female empowerment and reduced fertility
(see Chapter 1 on Sen and ‘capabilities’). Finally, not all expenditure on educa-
tion is intended to generate productive human capital; much is about improving
the ability of an individual to participate in wider social processes.
There are many more general reasons for these weak empirical results. The
most obvious is the statistical problem of correctly measuring the variable of
Growth in the Modern World Economy since 1950 51