semi-logarithmic formula assumes that income elasticity varies inversely with the level
of consumption or expenditure, the effect of the adjustment was in nearly every case to
increase the elasticities above the calculated figures; in general however, the adjustment
was relatively greater for the rural than the urban households.

 Having arrived in this way at estimated figures valid both for the rural and urban sec-
tions of the sample, we must next take account of the fact that the data used for the analy-
sis related to the expenditure incurred by the housewife on the different items purchased
and not to their quantity or their value at ex-farm or import prices. It is well-known
that the elasticity of expenditure on a given type of food has in general a higher value than
the elasticity of the quantity purchased, because part of the additional expenditure incurred
by higher- income families is devoted to higher priced varieties of food or to products
whose value includes a higher proportionate element of processing or distributive costs.
The element of quality elasticity may play an important part in determining the total ex-
penditure elasticity, and estimates of its effect can be obtained from surveys (such as the
United Kingdom National Food Survey) in which the elasticity can be separately measured
both for quantities purchased and for expenditure. In the Jamaica survey, no precise al-
lowance for this "quality factor" was possible, but since what is required for purposes of
projection is an estimate of the quantitative elasticity in terms of farm output or imports,
a downward adjustment, determined in the light of general probabilities, was made to the
estimates resulting from the procedure described above. In most cases this second ad-
justment largely cancelled the effect of the first so far as the urban households were con-
cerned, but for the rural sample the elasticities remained on the whole higher than the un-
adjusted figures resulting directly from the analysis.

 If the figures obtained in this way could be taken as valid estimates of the income elas-
ticities of demand for the rural and urban sections of the total population respectively, it
would be sufficient to weigh them according to the rural and urban proportions of expendi-
ture on each commodity in order to calculate average figures applicable to the population
as a whole. This procedure, however, assumes that the survey was in fact representative
of the whole population, an assumption which cannot be taken for granted without question.
The survey was conducted by random methods and was intended to give a representative
sample, but it seems probable that, like most surveys, it failed to give adequate repre-
sentation to the highest income- strata. The average food expenditure recorded in the
survey (J26.4 per capital per year) falls substantially short of the official estimate made
for national income purposes of the total personal expenditure on food, which works out at
an average for 1958 of about J 39.8 per capital. An exact agreement between these two
figures is not to be expected for a number of reasons; in particular the national income
estimate includes the expenditure of non- residents, which must be excluded in order to
give a figure comparable with the survey average, while the latter is subject to errors of
recording which are obviously much more likely to result in under- estimation than in
over-estimation. It is however difficult to explain in these ways more than a small part of
the relatively large discrepancy between the two figures. If the national income estimate
is to be taken as approximately accurate, therefore, it seems necessary to assume that
the highest income-groups among the population, say the wealthiest 10 to 15 percent, were
substantially under-represented inthe survey. The difference would be largely accounted
for if the survey average could be taken to represent the expenditure of say 85 percent of the
population, and if the average food expenditure of the remaining 15 percent were assumed
to have amounted to about J 100 per capital per year or about J 2 per week. This is by
no means an improbable figure it was in fact exceeded by about 14 percent of the two-

 57