What role for human capital in the growth process: new evidence from endogenous latent factor panel quantile regressions
| Author | Philip Kostov,Julie Le Gallo |
| Date | 01 November 2018 |
| DOI | http://doi.org/10.1111/sjpe.12196 |
| Published date | 01 November 2018 |
WHAT ROLE FOR HUMAN CAPITAL
IN THE GROWTH PROCESS: NEW
EVIDENCE FROM ENDOGENOUS
LATENT FACTOR PANEL QUANTILE
REGRESSIONS
Philip Kostov* and Julie Le Gallo**
ABSTRACT
The estimates for the human capital effect in cross-country growth regressions
have been subject of considerable controversy. We argue that human capital is
intrinsically a multidimensional construct. We construct human capital measure
by combining available alternative proxies via confirmatory factor analysis.
Using panel data endogenous quantile regression methods we analyse the whole
conditional growth distribution by simultaneously accounting for the potential
endogeneity of human capital and country-specific effects. Our results conform
to theoretical expectations and we are able to demonstrate the beneficial effect
of both the measurement approach and the endogeneity correction on the deriva-
tion of theoretically consistent estimates.
II
NTRODUCTION
Cross-country growth regressions routinely employ educational variables as
proxies for human capital. However, the significance of education measures in
growth regressions has been source of considerable controversy. While the
theoretical arguments of why it should be important in driving economic
growth are compelling (Mankiw et al., 1992), the empirical evidence has been
mixed (Krueger and Lindhal, 2001; Pritchett, 2001). Indeed, there are several
potential problems with used educational variables.
The first one is the imperfect nature of such measures. There has been con-
siderable debate on the appropriateness of different proxies for human capital
and the quality of the available data. Different researchers have put forward
new improved databases of such proxies. Examples include levels of educa-
tional attainment, such as the fraction of working age population in sec-
ondary school (see Nehru et al., 1995; de la Fuente and Domenech, 2006;
Cohen and Soto, 2007; Lutz et al., 2007; Barro and Lee, 2010). However,
*University of Central Lancashire
**AgroSup Dijon
Scottish Journal of Political Economy, DOI: 10.1111/sjpe.12196, Vol. 65, No. 5, November 2018
©2018 Scottish Economic Society.
501
these measures present a number of drawbacks (Folloni and Vittadini, 2010),
in particular, not only the quantity but also the quality of years of education
have an impact on the cognitive skills acquired and ultimately on growth
(W€
ossman, 2003). Hence, these proxies should be designed to measure an
intrinsically unobservable variable: the quality of human capital. While it is
conceptually inconceivable that a single proxy would be able to successfully
capture the quality of human capital, empirical research in cross-country
growth regressions has nevertheless traditionally applied a single proxy for it.
On the other hand, using several human capital variables can lead to issues
with multicollinearity: it would be then tricky to disentangle the effect of sev-
eral such measures used in the same model. Owing to the imperfect nature of
such measures, it is therefore unsurprising that such models have been known
to exhibit considerable heterogeneity and non-linearities within the growth
process (see Temple, 1999; Kalaitzidakis et al., 2001; Sianesi and Reenen,
2003; Sunde and Vischer, 2015).
The conceptual model underlying the cross-country growth regression sug-
gests an unobservable human capital variable that could be potentially mea-
sured by a number of different proxies. One possibility to overcome this
difficulty is to consider human capital as a factor variable transforming part
of the design matrix of the growth regression to incorporate a confirmatory
factor analysis structure. Hence, this paper approaches this problem by using
a general structural equation model (SEM) framework. SEMs have been used
in growth regressions to estimate the impact of variables, for which only
imperfect proxies exist such as policy variables (Brumm, 1997), well-being
variables (Cracolici et al., 2010) and various factors such as creative capital,
entrepreneurship or leadership. However, to the best of our knowledge, these
models have not been used to deal with the specific issue of human capital. In
this paper, based on the CANA database (Castellacci and Natera, 2011), we
use the main educational variable ‘Mean years of schooling’ together with
‘Public Expenditure on Education’, and ‘Primary teacher-pupil-ratio’ to con-
struct a confirmatory factor analysis of the human capital factor. Our
approach can therefore be seen as a macroeconomic extension of the
approach set out by Dagum and Slottje (2000) who, at a microeconomic level,
consider human capital as a multidimensional non-observable construction of
personal ability, home and social environments and investments in education
of the household head and spouse.
The second issue with educational proxies and the human capital factor
that is to be derived from them is their potential endogeneity. We propose to
address this endogeneity via instruments. Search for instruments for human
capital in itself is a problematic area as issues of validity of instruments and
identification are difficult to deal with (Temple, 1999). Here, we suggest using
lagged enrolment rates to address the endogeneity problem. Enrolment rates
can be viewed as an aggregate determinant of educational achievement and
therefore should satisfy the exogeneity requirement. However, they only affect
it after a considerable lag of time and therefore should be excluded from
defining human capital directly. Hence, lagged enrolment rates appear to
502 PHILIP KOSTOV AND JULIE LE GALLO
Scottish Journal of Political Economy
©2018 Scottish Economic Society
To continue reading
Request your trial