THE SIZE MOBILITY OF EARNINGS*
| Published date | 01 May 1983 |
| Date | 01 May 1983 |
| Author | P. E. Hart |
| DOI | http://doi.org/10.1111/j.1468-0084.1983.mp45002003.x |
THE SIZE MOBILITY OF EARNINGS*
P. E. Hart
I. INTRODUCTION
Economists and statisticians, among others, have been systematically
studying the inequality of incomes for over a hundred years, using
a wide variety of different measures of static inequality. But their
investigations of the mobility of incomes, showing the extent to which
people move up and down the distribution, are of more recent origin,
depending as they do on the availability of longitudinal data on the
incomes of the same people over time. Different measures of mobility
are used partly because different problems are investigated. For example,
Atkinson (1980) concentrates on the intergenerational mobility of
incomes, Hart (1976) measures intragenerational mobility over periods
up to 10 years, King (1980) is concerned with the effects on horizontal
equity of changes in taxation, and Shorrocks (1981) studies the effects
of different lengths of time period on mobility. Much research work
is still in progress and it is too early to judge the respective merits of
the different mobility measures used in these different approaches
to studying the dynamics of income distributions.
The present paper uses measures of mobility based on the product
moments of the bivariate distribution of the logarithms of weekly
earnings of adult males in the United Kingdom 1971-78. This approach
follows from the use of the moments in univariate distributions (after
logarithmic transformation) to measure the inequality of earnings in
1971 and 1978, based on classical statistical analysis.' Most distributions
in economics are summarized by their moments. In any case the
moments are very convenient, they have well-known sampling proper-
ties, they are readily decomposed, and they are scale-invariant in
the sense of being unaffected by replication. The entropy family of
measures of inequality are not scale-invariant and are inconvenient for
comparing inequalities of distributions with widely different numbers
of observations.
A possible limitation of the use of the moments of the logarithms is
that the variance of the logarithms may or may not satisfy the principle
*J should like to thank J. Creedy and M. A. King for their helpful comments on an earlier
version of this paper. I am also indebted to the Department of Employment for providing
Table 1, but they are not responsible in any way for the analysis of it in this paper.
'A more detailed appraisal of the convenient properties of moments is given in Hart (1982),
the discussion paper on which the present analysis is based.
181
To continue reading
Request your trial