THE DEMOGRAPHIC EFFECTS OF INCOME REDISTRIBUTION AND ACCELERATED ECONOMIC GROWTH REVISITED: A REPLY
| Published date | 01 May 1988 |
| Date | 01 May 1988 |
| DOI | http://doi.org/10.1111/j.1468-0084.1988.mp50002007.x |
| Author | C. R. Winegarden |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 50,2(1988)
0305-9049 $3.00
THE DEMOGRAPHIC EFFECTS OF INCOME
REDISTRIBUTION AND ACCELERATED
ECONOMIC GROWTH REVISII'ED: A REPLY
C. R. Winegarden
I. INTRODUCTION
The main thrust of Flegg's meticulous critique of my paper (Winegarden,
1984) is directed at what he believes to be the sensitivity of certain of my
results to changes in specification and estimating method. In response, I shall
show that most of his suggested changes are inappropriate and, further, that
my results are strikingly robust in the face of reasonable variations in specifi-
cation.
Flegg's critique is largely focussed on two specific features of my paper: the
variable used to control for the effects of governmental family planning
programmes and my use of two-stage least squares estimation. I shall address
each of these points sequentially, and then re-estimate my model in order to
compare various results.
II. THE FAMILY-PLANNING VARIABLE
Flegg bases much of his case on a quite peripheral issue - the absolute
magnitude of the estimated coefficient of the family-planning dummy used in
equation 1 of my model. As I noted in the original paper, the estimated
impact of an established family-planning programme is implausibly large, and
probably reflects correlation with unmeasured variables in the error term. In
a study primarily devoted to the effectiveness of government programmes to
reduce fertility, this result would call for further investigation. In the present
context, it should occasion little concern. However, Flegg makes the size of
this coefficient the standard by which to judge a series of alternative specifica-
tions. In doing this, he ignores the fact that the data clearly show the existence
of two regression populations, which are distinguished by a substantial inter-
cept value. Even though the original intercept adjustment is probably too
large in terms of what it is intended to measure, the fact remains that it can be
omitted or arbitrarily diminished only at the risk of generating biased and
inefficient estimates of the other regression coefficients. This, one must
suspect, is precisely what happens in his equations 2-4 in Table 1, and which
he mistakes for instability in the coefficient estimates relating to income
distribution and economic growth. (It will be noted that as the intercept
adjustment is reduced to more acceptable levels, his 1 2 values also decline.)
195
To continue reading
Request your trial