Front Cover
 Table of Contents
 The review of budget analysis and...
 The consumption function in Jamaica:...
 Analysis of data from household...
 Appendix I. The "adding-up"...
 Appendix II. Tables 1 to 22 and...
 Back Cover

Title: Consumers' expenditure in Jamaica
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00087184/00001
 Material Information
Title: Consumers' expenditure in Jamaica an analysis of data from the National accounts (1832-1960), and from the Household budget surveys (1939-1958)
Series Title: Consumers' expenditure in Jamaica
Physical Description: 148 p. : graphs ; 24 cm.
Language: English
Creator: Taylor, LeRoy
Jamaica -- Dept. of Statistics. -- National accounts
Jamaica -- Dept. of Statistics. -- Household expenditure survey
Publisher: Institute of Social and Economic Research, University of the West Indies, Jamaica
Place of Publication: Mona
Publication Date: 1964
Subject: Consumers -- Jamaica   ( lcsh )
Cost and standard of living -- Jamaica   ( lcsh )
Genre: federal government publication   ( marcgt )
bibliography   ( marcgt )
non-fiction   ( marcgt )
Spatial Coverage: Jamaica
Statement of Responsibility: by LeRoy Taylor.
Bibliography: Includes bibliographical references.
General Note: Publication date stamped on cover.
General Note: Includes analysis of responses from the Household expenditures survey, 1958.
General Note: "Similar to ... the author's thesis ... University of London."
 Record Information
Bibliographic ID: UF00087184
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 05076554
lccn - 79303780

Table of Contents
    Front Cover
        Front Cover 1
        Front Cover 2
        Dedication 1
        Dedication 2
        Preface 1
        Preface 2
    Table of Contents
        Table of Contents 1
        Table of Contents 2
    The review of budget analysis and theories of consumption
        Section 1
        Section 2
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
    The consumption function in Jamaica: Historical and survey data
        Page 38a
        Page 38b
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
    Analysis of data from household expenditure survey, Jamaica, 1958
        Page 84a
        Page 84b
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
    Appendix I. The "adding-up" conditions
        Page 102a
        Page 102b
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
    Appendix II. Tables 1 to 22 and charts 1 to 7
        Page 114a
        Page 114b
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
    Back Cover
        Page 150
        Page 152
Full Text








To Pierre, Gloria


The matters discussed in the following pages of this monograph really
emanated from posing the question: are the methods of economic reasoning,
the predictions of economic theories fashioned for industrial, advanced
communities relevant also for less developed economies such as Jamaica? We
have found, by considering one particular aspect of economic thought, namely,
the theory of consumers' behaviour, that such methods and such predictions
are relevant, though the extent of their relevance depends on (1) the internal
consistency of the theoretical model, and (2) the degree to which we, in
specifying our model, can anticipate, and hence make allowances for dis-
crepancies between the theoretical and measured magnitudes.
These points are well illustrated in the debate concerning whether we
should use absolute income (Keynes), relative income (Duesenberry) or
permanent income (Friedman) in specifying the relationship between aggregate
consumption and income (Section I). We have generally tended to favour the
Friedman hypothesis in this debate and although some ambiguities emerge in
applying the model to survey data of Jamaican consumers, the predictions of
the permanent income hypothesis tend on the whole to be meaningful in the
Jamaican environment(Section II). At the same time we recognize the a priori
plausibility of the Duesenberry hypothesis, but we have not been able to carry
out any convincing test of its implications.
In explaining the behaviour of components of the Jamaican consumers'
budget (Section III) our discussion has been inspired by the aim of finding a
single specification of the Engel curve which, on theoretical and statistical
grounds, best fits the data. A subsidiary aim has been that of discovering the
effect of variations in household size and consumption. We have found that
with a few modifications the "linear law" of expenditure gives a sufficient
explanation of the behaviour of many commodity groups. Our general im-
pression is, however, that no simple two-parameter Fngel curve is capable of
giving both a necessary and sufficient explanation of the variation of ex-
penditure on each and every item in the Jamaican consumer's budget. Differ-
ences in location and related factors differences in social class between
households, and of course differences in "tastes" also affect expenditures,
though perhaps not as significantly as variations in income, household size
and household composition.
The present monograph is similar to a thesis accepted for the internal
degree of Doctor of Philosophy (Economics) by the University of London.
I am very grateful to my supervisors Prof. R. G. D. Allen, Prof. A. W. H.
Phillips, Mr. J. Thomas, Mr. A. D. Knox and the late Mr. H. S. Booker who in
my graduate years at the London School of Economics and Political Science
gave sympathetic guidance to my efforts. I am particularly grateful to Prof.
Allen whose comments and criticisms have helped considerably in moulding
all sections of the present work. To Prof. Phillips I am deeply indebted for
many helpful suggestions in all my graduate studies.
Mrs. Mary Manning Carley provided invaluable editorial assistance and
for this I am grateful.
I wish also to thank Miss Sylvia Bayliss for typing and checking the

Le R. O. T.




1. Introduction ... . ... ... ... 1

2. Analysis of family budget data: The "classical" approach 4

3. The problem of defining a "homogeneous" group of households 6

4. Engel curve analysis: the "classical" approach ... ... 14

5. The nature of household income in Jamaica ... ... 18

6. Theories of the consumption function ... .. ... 20

7. Note to section I. The consumption function: Pure theory ... 33


1. Aggregate consumption expenditure in the national accounts,
1832 1960 ... ............ 39

2. The question of errors in the national income data and related
problems ... .. ... ... ... 50

3. Aggregate consumers' expenditures: household budget data,
Jamaica, 1939 1958 ... ... ... ... 54

4. The economic implications of the analysis of aggregate
consumption (budget data) ... ... ... 65

5. The components of total consumers' expenditure in Jamaica 70

6. Summary ... ...... ... ... 83

CONTENTS Continued



1. Differences in expenditure patterns ... ... ... 85

2. Alternative forms of Engel curves ... ... ... 87

3. The effect of household size and composition on the Fngel
formulation ... . .. .. .... 92

4. Testing alternative hypotheses about expenditure behaviour 98

5. Possible effects of "social class" and other factors ... 99


The "adding-up" conditions ... .. ... 103


Tables I XXII ... ... ... ... .. 115

Charts 1 7 .. .. ... ... ... 142






This monograph is an enquiry into the factors which determine con-
sumers' expenditure in Jamaica. It is an empirical enquiry in the sense that
we are not only interested in listing the numerous factors which are expected
to affect the way in which the consumer disposes of his income, but we wish
as well to measure the influence exerted by the most important factors.
The study divides itself conceptually into two parts. In one part we look
at the components of totalexpenditure and try to explain the variation of each.
In the second part we try to explain the variation of aggregate expenditure
In consumption economics broad "sociological" and demographic factors,
such as the location, age and sex composition of consumer units, as well as
income, are recognized as important determinants of the pattern of consumers'
expenditures. However, because of the elusiveness of the more sociological
determinants of consumers' preferences it has not been found possible to in-
corporate them explicitly into the theory of consumers' behaviour, which
theory stresses the importance of the more economic variables such as income
and prices. It would nevertheless be a great mistake to attempt to measure
the influence of "economic" factors on the pattern of consumption without
paying a good deal of attention to the influence of location, etc. All the well-
known studies in this field have had to face the problem of incorporating
some, at least, if not all of these factors which affect consumers' behaviour.1
The whole point of the matter is that: "The theoretical analysis of dis-
tribution of expenditure postulates a group of families with the same needs
and complex of preferences and with differing incomes."2 Families in differ-
ing localities and of varying sizes and composition, even if they have the
same income when observed simultaneously, are likely to show different con-
sumption patterns. It is obviously impossible to satisfy experimentally the
rigorous postulates of the theory; but because the theory is not implausible,
granted its assumptions, it seems a useful exercise to try to see how in fact
consumption does behave in response to changes in income, other things
being held, however imperfectly, constant. Numerous studies along these
lines have been carried out in other countries, but such studies are still in
their infancy in the West Indies generally.
In the present study I wish to see what patterns of consumption emerge
from a study of historical and survey data on Jamaica. The historical data
and the subsequent analysis will be presented in Section II of this thesis.
The survey data are analysed in detail in Section III.
The main body of survey data dealt with in this thesis is composed of
the responses from a sample survey of about 1,200 households interviewed in
Jamaica between July and October 1958. The households were taken from all
over the island and from all income groups. The survey was divided in such
a way as to represent households from three locational strata:
1. The Kingston Metropolitan Area.
2. The Main Towns.
3. The rest of Jamaica (Rural Areas)

R.G.D. Alien and A.L. Bowley, Family Expenditure: a Study of its Variation,
London, 1935;S.J. Prai andH.S. Houthakker, The Analysis of Family Budgets, Cam-
bridge, 1955.
Allen and Bowley, p. 18.


On the basis of the estimated population of these three strata (as well
as other considerations) it was decided to interview 400 households in
Kingston, 130 in the Main Towns and 720 in the Rural Areas, care being
taken to secure adequate geographical dispersion of the households within
each strata.

Interviewers were required to complete two schedules of weekly ex-
penditure for each household. On Schedule A information was requested about
the household's expenditures on certain items over the past month, quarter,
or even year, as well as expenditure over the past week. The records on
Schedule B refer exclusively to expenditure over the past week. The total
response to the interview by strata was as follows:

KoM.A. Main Towns Rural


No. of households in sample 4D2 402 143 143 738 738

No. of households tabulated 324 306 122 119 676 662

Refusals. 38 56 9 10 52 58

Not completed for other reasons 40 40 12 14 10 18

The total number of households actually visited in each area was much
greater than the number of households in each sample (See below).
K.IM.A. Main Towns Rural
No. of households visited 4,024 724 2,316
The reason for this is that enumerators were not, as in previous surveys,
given a list of the names of heads of households to be interviewed, but were
given instead the sampling fraction determined beforehand for each
sample area. They were then required to list all the households in the area
and collect information from a sample of these households selected accord-
ing to the sample fraction (Pages 4-6 of the Household Expenditure Survey
1958 shows the distribution of households in the sample by parishes and by
sample areas within each parish).
The information requested was of the usual kind: the value, quantity,
and sometimes price of items .purchased, received as gifts, or produced at
home; the age, sex, occupation, education, income, etc., of the head and
members of the household. The household was, as in the Population Surveys,
defined as consisting of "a person or group of persons, living and house-
keeping together".3 The definition covers maids who live on the premises,
boarders and lodgers, except where these latter have all their meals else-
where. This definition is therefore equivalent to the definition of "family"
and almost but not quite equivalent to the "spending unit"-4
3Household Expenditure Survey 1953.4. Jamaica, Department of Statistics,
p. 5.
4M.H. David. Family Composition and Consumption. Amsterdam, 1962. p. 27.


The occupation, work-status and education recorded for each household
refer to the head of the household only. But enumerators, in so far as they
were able, recorded the income of all income-earning members of the house-
hold. The age-range and sex of every member of the household was also
recorded. The age-ranges are rather wide and this makes it difficult to con-
duct a very detailed investigation of the effect of age and sex on household
consumption. Infants of age 4 years and under are distinguished from child-
ren between 5 and 14; adults of both sexes are divided into three age groups
(1) 15-24; (2) 25-59; (3) 60 and over.
The difficulties and limitations of family budget data are well recognized.
Clearly the usefulness of such data is dependent upon the ends they are re-
quired to serve. Representativeness may not be absolutely essential if we
wish to conduct an analysis of the way consumption changes when income
changes.5 On the other hand a number of errors arises in the interview situa-
tion which give a distinct bias to the results; in particular the tendency for
households to exaggerate expenditure and to give an under-estimate of their
actual income position has often been stressed. There is no sure way by
which biases of this sort may be overcome, nor in fact is there any method
by which we can gauge their magnitude (but cf. later discussion of Friedman).
In the Jamaican surveys respondents are requested to recollect purchases of
food and other very current items of consumption over the past week and to
recollect also purchases of durable goods and services made over the past
month, quarter or year. Errors of omission are obviously likely to arise in this
procedure through genuine inability to remember all the items purchased,
especially those which the household is not in the habit of purchasing regular-
ly. On the other hand "due to unconscious telescoping of events......",6
items purchased before the beginning of the seven-day period are likely to be
An attempt is made to overcome these sources of error by field-checks
in which a senior officer undertakes the field enumeration for certain items
of expenditure only so as to avoid the burden on the respondent of repeating
the entire recollection of items. Further, wherever it was possible the prices
recorded for food items were checked at the groceries and shops in the
vicinity. In spite of these precautions it is still not possible to say how great
is the bias deriving from this source, though it is not expected to be very
In the 1958 survey, as in the previous survey, information concerning
non-purchased food was collected in the Rural sample. Subsistence consump-
tion affects the cash purchase, especially of root crops and fruit and vege-
tables. The value of these items has been included in total household ex-
The complete list of items on which expenditure was recorded is given
in the Report.7 We are not interested here in studying expenditure on each
of the items as listed; what we wish to do in analysing the Jamaican budget
data is to derive some conclusions about the factors which determine the
behaviour of broad components of household expenditure. Originally the

5 Prais and Houthakker, Ch. 4.
6Household Expenditure Survey, 1953-4, p.8.
7Household Expenditure Survey, 1958, pp. 85-98.


total list of items had been compressed to about 27 groups; even so there
still remained a number of groups where average household expenditure was
quite insignificant. In order to obtain significant results from our analysis
we have further compressed the number of commodity groups to 16 (See
Section III).


Before embarking on a study of household expenditure it is well worth
pausing to consider seriously what we expect a study of such information to
show and how we expect to go about isolating the factors which operate
simultaneously on the data we are investigating. A study of previous work
in this field carries us a long way in this discussion.s
The early and, in many respects, path-breaking work by Allen and Bowley
on working-class families in Britain was aimed at discovering "how far the
expenditure of individual families, or of groups of families, can be described
by rules and formulae" and how far such rules could be related to postulates
of economic theory. The study was also aimed at accounting for "the varia-
tions from the averages that result from the different choices of individual
Assuming prices to be constant, as they are for a "homogeneous" group
of families at roughly the sane moment in time, it was found possible to
describe fairly accurately the behaviour of expenditure on particular items
by a linear equation in which total family expenditure was the single-determin-
ing variable. A relationship of this form has two powerful characteristics to
recommend it: it is simple and it satisfies the "adding-up" criterion. On
the other hand "the linear expenditure law" becomes an inadequate general
formula if the households or families under investigation show wide dif-
ferences in respect of total income or expenditure. No matter what formula is
used, however, we shall always find that the expenditure of each household
deviates to a greater or less extent from the rule, and the problem of explaining
this variation remains. The methods discussed by Allen and Bowley are still
Having defined a "homogeneous" group of families (which definition
we shall shortly be discussing) when there is detailed information on each
budget, we can write for the tth family -

Yit = kiet + ci + vit

where Yit is expenditure on the ith item et = total family expenditure k and
c are constants to be determined by least-squares, and vit, whose mean is,
by assumption, zero, is the difference between actual expenditure and ex-
penditure predicted by the "linear law". A study of the behaviour of Vit
thus enables us to decide the extent to which we have or have not eliminated
variations in expenditure deriving from different tastes. Two considerations
need bearing in mind. First of all it is necessary to see whether the actual
distribution of residual expenditures approaches symmetry. If the distribution
Allen and Bowley, op. Ci t; Prais and Houthakker op. ci t; Milton Friedman, A Theory
of the Consumption Function,. rinceton, 1957.
Allen and Bowley, p. :1.


is fairly symmetrical, we wish to see, secondly, how closely it approximates
the normal distribution. A graphical comparison of the numbers of positive
and negative values of yt, with the corresponding numbers of values of vt
above and below its mean, enables us to decide on the improvement in
symmetry. Using the Chi-squared test it is then possible to assess the good-
ness of fit of the residuals to the normal curve of error. If after applying both
these tests we find that the residual expenditures are neither symmetrical nor
normally distributed, then we have not succeeded in accounting for tastes:
in effect neither the one nor the other of our definitions of "homogeneity"
and "income" is sufficient to explain the observed variation in expenditure
on the item.
The more recent work of Prais and Houthakker employs the same general
approach as that outlined above. If there is any fundamental departure from
Allen and Bowley it lies in the decision to treat the relationship between ex-
penditure on a specific item and total expenditure as basically non-linear
rather than linear. o The chief advantage of the non-linear model is that it is
now possible to study as a unit a sample of households differing widely in
respect of income. On the other hand, with the kind of non-linear formulations
they adopt, there is no certainty that the least-squares estimates of the para-
meters will satisfy the "adding-up" criterion."' Nevertheless, in studying
the behaviour of expenditure on particular commodities it may be wiser not to
adopt the same form of equation for all commodities, provided that the data is
sufficiently reliable to give meaning to the results derived from the various

Where the distribution of income is narrow a single formulation may be
sufficient to describe every item of expenditure, and if this single formulation
happens to be a polynomial with at least one-term linear in income, then the
"adding-up" criterion will also be satisfied. 12 In any case, having fitted the
equation, it is advisable to examine the distribution of residual expenditures
and to test this distribution for homoscedasticity and normality. For if the
residual expenditures do not even approximately satisfy these conditions,
we have not randomised variations in tastes between the different households,
and this difference will be due, not only to inappropriate formulations of the
Engel curve, but more importantly to the definitions of the "income" variable
and the factors which we consider describe a "homogeneous" group of families
or households.

Prais and Houthakker, p.47 ff.
Prais and Houthakker point out on p.8 3 that the satisfaction of the "adding up" criterion
implies that all the c's in the equation
yit= kiet+ vt+ ci

be zero. None of the c's canbe negative since the equation Is only defined for positive values of
y and e.
Prais and Houthakker, p.84; J.L. Nicholson "Variations in Working Class Family Expen-
diture", Journal of the Royal Statistical Society, Vol. CXII, Part IV, 1949.



Location is clearly a factor determining the pattern of household con-
sumption. This factor is very important in Jamaica, where location happens
to be associated with other factors which, in their own right, affect the dis-
position of household expenditure.
Urban Jamaican households for example consist largely of wage-earners,
whereas the rural household contains mainly own-account farmers. The re-
gularity of employment in urban areas is higher than in the rural areas owing
largelyto the seasonal nature of most of the agricultural products of Jamaica.
Owing to seasonality, too, the pattern of prices paid for the food items at the
same time of the year is different as between rural and urban households.
Fven within the rural area itself, the prices paid for the same food item can
vary sharply from household to household at any moment of time.
The rural area further defines households of a social milieu which con-
trasts quite definitely with that of the urban areas. The number of persons per
household increases as one moves from town to country, and this size factor
operates at all age levels. The typical rural household seems to contain a
larger proportion of children than its urban counterpart, and this indicates that
the average age of members of the households is probably less in the rural
The typical household head in the Jamaican countryside is the own-
account farmer. He is usually the owner of his own home and property, un-
like the typical urban wage-earner; and although his income is almost half
that of his urban counterpart, his cash consumption is not as high by virtue
of "subsistence consumption", i.e. home-grown and home-consumed food, etc.
The distribution of households by income in rural and urban samples seems
also to point to greater inequality in the rural area. Part of the evidence on
this score is shown in Chart 1. The diagonal in this Chart represents a
situation of equal income distribution. It may be noted that every point (other
than the end-points) on the line describing the income distribution of rural
households is further away from the diagonal than the corresponding point
on the line describing the urban income distribution. It is true- that we are
here comparing different years, but evidence from the 1958 Sample Survey of
Household Expenditures, as we shall see in Section II, brings out the same
feature when the distributions for the same year are compared. Tables I III
(Appendix) illustrate further some of the structural differences which we have
been discussing.
The result of these structural differences deriving from household loca-
tion can be seen in the average expenditure of households in each area
(Table IV). The average urban (Kingston Metropolitan Area) household spends
consistently more than the rural household on all items, or rather, groups of
items of expenditure; and with the exception of the item rent etc., so does the
urban(Main Towns) household. In percentage terms, also, the urban household
devotes a larger proportion to rent, transportation, personal and miscellaneous
expenses. In view of the higher regularity and higher average level of income
of the urban household, it; is easy to see why the absolute expenditures should
be higher; and because food becomes a less urgent item of expenditure as the
"standard" of living improves it is understandable that the proportion spent
on food should be lower for the average urban household than for the average
rural household. The observed behaviour of rent, transportation, personal and
miscellaneous items is also expected from what we have said earlier, except


that the higher level of prices for most of these items in the towns is also re-
flected in the figures. This price effect is most important in the case of rent
etc. On the other hand, locational differences seem to have no significant
effect on proportionate expenditures on household equipment, fuel and clean-
ing and tobacco.
Itis clear, then, that fairly significant differences in expenditure patterns
result from differences in the location of households in Jamaica. This change
in patterns is accompanied by a change in the average level, character and
distribution of income. The income effect is most noticeable in the case of the
behaviour of absolute and proportionate expenditure on Food and Clothing. 13
Proportionate outlay on these items declines as the household becomes more
and more urbanised, while the taste for expenditure on Personal Items (cinemas,
books, magazines, sports, gambling etc.) increases noticeably as we move
from the country to the town. The only items for which rural and urban house-
holds seem to have similar tastes are expenses connected with household
maintenance and expenditure on Tobacco.
Since the urban household reveals, on the average, a different pattern of
expenditure which, as we have implied, is partly due to structural differences.
deriving from locational differences, we should study urban and rural house-
holds separately; thus our first task in defining "homogeneous" groupsof house-
holds in Jamaica is to classify them by urban and rural. Having done this we
still have to distinguish between households of different sizes and composition,
for these factors in themselves affect the disposition of household income.
Itis customaryin analysing budgets to make allowance for varying house-
hold composition by using the concept of the "Unit Consumer" or "Equivalent
Adult". These methods are simply means of expressing every household
member in terms of a single unit, usually in terms of the calorie requirements
of an adult male. It is then implicitly assumed that the relative needs of
each individual in terms of calorie requirements are equivalent to the relative
expenditure undertaken by the household on his account. Sydenstriker and
King, in a fundamental article of 1921, questioned this assumption.14 These
researchers were faced with the problem of finding "a reasonably accurate
mode of determining the relative income status of each family" in a group of
families residing in seven cotton-mill villages of South Carolina. It was re-
cognised that a classification of families by total family income, or per caput
income, would ignore important differences in the composition of the families.
Clearly a scale based on the actual expenditure by or on behalf of individuals
of varying ages and sex was indicated. But no such scale was then in ex-
istence; scales had previously been constructed on the basis of physical re-

This is inkeeping with the "laws" uncovered by Ernst Engel and others which state that
the proportionate expenditure on food and necessities declines as the standard of living of the
household increases (See Prais and Houthakker, p.79).

1E. Sydenstriker and W.I. King, "The Measurement of the Relative Economic Status of
Families", Quarterly Publication of the Anerican Statistical Association, vol. 17.
Sept. 1921, pp. 842-857. For a good review see W.E.Ogburn"A Device forMeasuringtheSize
of Families Invented by E. Sydenstriker and W.I. King" in Methods in Social Science: aCase
Book ed. by S.H. Rice, Chicago 1931; also R.M. Woodbury, "Economic Consumption Scales and
Their Uses"', Journal of the nAerican Statistical Association, Vol. 39, Dec. 1944,
pp. 455-468


quirements of individuals. Ernst Engel, as early as 1895, had employed scales
based on varying weight per centimetre of height and had suggested that the
food required by a child under the age of one should be selected as a unit and
named a "Quet" after Quetelet. Rowntree and others had rated familiesin a
similar fashion.
The scale which King and Sydenstriker chose in the first instance was
the Atwater scale. This scale, developed by W. O. Atwater and others in the
United States Department of Agriculture, defined as its unit the calories needed
to "keep a normal young man who is engaged in moderate muscular work in a
proper state of health and vigour". 1 The problem was then one of seeing to
what extent relative family requirements, according to the Atwater scale,
diverged from relative total money expenditures on each individual.
With the expectation that differences in age and sex between families would
have similar effects on food expenses as on food needs, measured in terms of
calories, it was decided to use the Atwater scale as a preliminary measure of
family size (food expenses being a large proportion of total expenditure).
Each individual was, according to his or her age and sex, rated in terms of
the U.S.D.A. unit, renamed by King and Sydenstriker the Adult Male Unit
(A.M.U.). (Apparently in the survey detailed ages were recorded).16 These
fractions, when summed for each family, provided the measure of relative
family size. The total household income per A.M.U. was then used to classify
the families into income groups.
It was decided to approach the problem by deriving scales for food and
non-food separately and by weighting the two scales thus derived to form a
composite scale of relative money expenditures.
The scale for food, called by the authors a "fammain" (food for adult
male maintenance) was derived as follows:
The total food expenditure per A.M.U. was calculated for each family
already arranged by income per A.M.U. Within each income group the
families were classified into sub-groups according to the proportion of
females in the household. The average expenditure per A.M.U. for food in
each of these sub-groups was calculated and each average was expressed
as a percent of the average food per A.M.U. for the entire income group.
The indices so derived for each income group were averaged over all
income groups each index being weighted by the number of A.M.U.'s in
the income class.
Had the relative importance of females in determining food costs been
adequately reflected in the Atwater scale, any changes in the sex-composition
of the households should have had no effect on the average food expenditure
per A.M.U. between sub-groups. On the contrary it was found that average food
expenditure per A.M.U. tended to rise slightly as the proportion of females
increased. This suggested that the relative importance of females in determin-
ing food costs had been under-estimated by the Atwater scale, and an upward
adjustment to this scale was appropriately made.
The next step was to make adjustments in the scale for age. Keeping the
same income classification, the households in each income class were now
separated according to whether they did or did not contain children. Other
families were discarded as being irrelevant. The families with children were
arranged according to-the ages of the children, and families composed entirely
of adults were classified according to the average age of the members. A com-
parison of the average expenditure on food per A.M.U. per household was then

15Sydenstriker and King, op. cit., p. 844, ff.
16Ibid. p. 849


made between each group of households. In each income class this average
was found to be materially less for the "older" household than for the
"younger" household and to be slightly higher for the children than was pre-
dicted by Atwater and others. For each income class a separate index series
was obtained, and an overall average index was computed by weighting the
separate indices again by the number of A.M.U.'s in each income class. The
average index so derived supplied a means of adjusting the Atwater scale for
age in addition to sex. The term "fammain" was then defined for any given
class of people as "a demand for food of a money value equal to that
demanded by the average male in the given class at the age when the expense
for his food reaches a maximum".17 The new scales therefore are intended to
show "the actual relative expenditures for food for persons of differentsexes
and various ages" rather than their relative requirements in terms of calories.
Having derived a new scale in terms of relative food costs, King and
Sydenstriker went on to try to derive similar scales for non-foods. The ex-
penditures of 140 families on non-food items were investigated. For persons of
each sex within a given income class, expenditure allotted on each item was
plotted against age. A smoothed curve was fitted to the points, and these
curves were combined in a weighted average over all income groups and all
items for each sex separately. These final curves were then smoothed, and
indices derived therefrom. 8 The two series for food and non-food were then
combined in proportion to the average sums spent by all individuals on food
and non-food separately. The resulting scales differ from the "fammain"
scales and were called "ammain" scales, the unit being defined for a given
class of people as "a gross demand for articles of consumption having a total
money value equal to that demanded by the average male in that class at that
age when his total requirements for expense of maintenance reach a maximum".
The "ammain" scales were also found to differ quite substantially from the
Atwater scale.
The article by Sydenstriker and King is important because it represents
the first serious attempt to answer the question: how does household com-
position in fact influence household expenditures, instead of the more nutritional
query: how should consumption be affected by household composition?19 Yet
this approach to family budgets was ignored for many years until Prais and
Houthakker "independently" revived it. In the interim Professor Allen held
that the problem of deriving appropriate unit consumer scales was well-nigh
insoluble and that we should not in fact attempt to combine families varying in
composition,2o Nicholson in 1949 attempted a solution by trying to measure
the "cost of a child".21 His method was to study the variation of expenditure
on particular items among families arranged according to the number and ages
of children in the family. He found that he became increasingly involved in the
problem of defining families at a given "standard" of living. He attempted to

1Ibid., p.847
The results of this process for the families interviewed by Sydenstriker and King are de-
monstrated in Table 11, p.853 of their article already referred to.
Prais and Houthakker, Ch.9.
R.G.D. Allen, "Expenditure Patterns of Families of Different Sizes", in Studies in
Mathematical Economics and Econometrics, ed.by O. Lange, F.McIntyre and Th.Yntema.
Chicago, 1942.
Nicholson, op. ci t.


find a solution here by selecting some commodity or group of commodities, ex-
penditure on which could be said to vary with the standard of living of the
households, independently of the number of children. "At any given standard
of living, total household expenditure must increase with the number of
children".22 By comparingthe expenditure patterns of households at the same
"standard" of living but with varyingnumbers of children it should in principle
be possible to arrive at estimates "of that part of total expenditure which could
be explained by the presence of one or two children".23 Clearly the results
are going to depend crucially upon what group of commodities are selected to
define the "standard" of living, so-called. It was apparently found that among
working-class families in Britain in 1937-38, expenditure on clothing provided
the most acceptable measure, and at the average standard of living so defined,
the sum of 26 shillings of the 92 shillings total expenditure by families having
two children was devoted to the children.
Whether the difference between the average expenditure of a family with
one more child than another at the same "standard" of living, is clue ex-
clusively to the child rather than to chance or taste factors is a highly deba-
table question; it is recognized24 that the varying needs and tastes ofthe
different families might in themselves affect the comparison. Secondly, the
data are often not sufficiently accurate nor reliable to define at all adequately
families at similar standards of living. Kemsley has in addition pointed out
that serious losses of efficiency result when this method, which may be called
the "subtraction method", is used on sub-groups.25
Many other attempts have been made to estimate the effect of household
composition on household consumption. 26 The method suggested by Frown is
probably the most interesting.27 lie shows that by grouping thedatainto types
depending on the age and sex composition of the household, it is possible to
derive specific expenditure scales for various types of individuals without
having to make any prior assumption concerning individual needs or require-
ments. Keeping family size and composition constant he shows that by first
calculating the effect of income on consumption we can use the coefficient so
derived to obtain a seriesof estimates of the effect of varying ages and sexes
of individuals on food expenditure. The problem of deriving unit consumer
scales thus becomes a posterior rather than an anterior exercise.

If x = Expenditure on the it commodity by the rth household
M = total net income of the r household
n. = the number of persons in the rt household belonging to
Jr th
the j age-sex group

2Nicholson, op. lit. p.375, ff.
2W.F.F. Kemsley, "Estimates of Cost of Individuals from Family Budget Data", Applied
Statistics Vol. 1, 1952.
For a brief review of these attempts see Prais and Houthakker, Ch. 9.
J.A.C. Brown, "The Consumption of Food in Relation to Household Composition and In-
come", Econometrica, Vol. 22, 1954, pp. 444-460.


yij are coefficients depending on the ih commodity and the
j age-sex group assumed constant for all households.o(j are coefficients de-
pending only on the jt age-sex group and again assumed constant for all
households, we may write
x.ir M
n fi _n
j r r
Now for any given sub-sample of households, say the k6, which is "homo-
geneous" in respect of the age and sex of its members, the denominators of
both sides of the equation remain unchanged as we move from household to
household within the sub-sample. So we may, for the kt sub-group, write

ik= ij njr

k = jr

The semi-log and double-log forms of equation (shortly to be discussed) of the
Engel curve for food items then become respectively: 28

Semi-log 2) (_X) r u+
Lr = ik a + fik bi log ak + ik r

S ik ai ik bi log ak + iik bi log Mr + 3ik uir

i.e. xir = ik + hik log Mr + ir

where ik estimates ik ( ai bi log ak)

hik ik bi
2 2 2
s ,, 2ik 0'
Double-log3) xir = (ik ci ak i) M di Lir
r e

i.e. ir= Pik M qike wir

2The original equations may be written
Semi-log a) _i a + b log r + Uir
X n X1 n r

Double-log b) xir C dI
=^'* ci -"" r
Sij jr n P. ir

See Brown, P. 447.


where Pik estimates (ik ci ak -di)

S d.

2 ,, o2
S Ui

(For the original formulation of eq-1) in terms of semi-log and double log
transformation see note 28).
The constants in the simplified form of the semi-log equation do not
enable us to obtain the "pure" income effect on consumption within a group of
homogeneous families, as the f' sik are unknown. But "the double-logarithmic
form is to be commended in that, by applying it to a group of households of the
same composition but of varying household incomes an estimate may oe ob-
tained of one of the parameters of the Engel curve which is completely free of
the effects of household composition". It will be noted, however, that the error
term enters into the equation in a multiplicative and not an additive manner.
From the original double-log equation we can immediately derive

Mt. -{ =;

Having estimated di and calculated a measure for the standard of livingof
each household29 i.e. Mr it is now possible to estimate the 8f 'sij byor-
ja j njr

dinary least-squares from the equation 4).
The method just outlined is indeed useful and eminently reasonable as
long as we remember that the error term in the double-log expression will not
be distributed with constant variance, and weight the calculations accord -
ingly. Another point in its favour is that it avoids the prior assessment of
individuals according to some presumed scale of equivalence on the basis of
needs or requirements.
As already indicated, Prais and Houthakker employed a method which in
principle is similar to that used by Sydenstriker and King in 1916- 1917. With
the help of modern computers the former were further able to obtain unit-con-
sumer scales for individual components of the budget and they were also able
to try different forms of equations for the Engel curve. But perhaps their main
contribution was to expose some sort of "theory" of equivalence scales.
On thepreliminary assumption that "consumptionper person depends only
on the level of income per person''', that in fact there are no economies or
diseconomies of scale in. consumption, they show that this implies that-
(1) the Engel curves for households of different sizes will most likely cross if
the range of income per person is sufficiently wide; and
(2) that "luxuries" and "necessaries" can be defined equivalently in
terms of household size and household income. These implications derive from
defining household expenditure as a function homogeneous of degree one in
household income and size.3o
Brown uses a slight modification of income per person.
Prais and Houthakker, p.89.


The "homogeneity" hypothesis, however, cannot explain enough of the
items in the budget, and so it has to be modified. This is done by developing
urit-consumer scales for each person and for each commodity and expressing
household expenditure in terms of these units (cf. equation 1 in Brown's art-
Now, whenever total expenditure is used to represent household income,
as it usually is in family budget studies, it is intuitively clear at any rate that
there must be a definite relationship between the consumption scales and the
"income" scales. For if the sum of expenditures is always defined as "in-
come", then any change in income as defined must always be exactly equiva-
lent to the sum of all changes in expenditure. As a corollary to this if income
remains unchanged, then the aggregate effect of changes in household compo-
sition should be zero. Denoting expenditure on the ith item by vi, total expen-
diture by v and the number of persons of type t by nt, these conditions can
be written respectively as


L ant

It can then be shown, after making appropriate substitutions, that the "in-
come" scale is equivalent to a weighted average of the specific consumption
scales (summed over commodities and persons), with weights approximately
proportional to the expenditures on the commodities.31 The mathematical
result is certainly not alarming but the results are repeated here to show the
similarity to the approach of Sydenstriker and King.
The method actually used to derive estimates of the consumption scale
is fully described by Prais and Houthakker. It is similar to Brown's by virtue
of the equations used. It is dissimilar in that it involves an iterative process
by which an initial guess concerning the parameters in the equation relating
consumption per unit consumer to income per unit consumer is constantly re-
vised until B2 in the derived equation connecting expenditure (adjusted for
income changes) and the various types of persons is maximised.32
Clearly the method suggested.by Brown would minimise the iteration
and guess-work involved by providing reliable estimates of the parameters
in the income equation, which estimates can be obtained by classifying tne
households initially into groups homogeneous in terms of age and sex com-
position. The method used by Brown possesses an added advantage in that it
is possible to decide on the degree of homogeneity achieved in each sub-group
by testing the constancy of the regression coefficient in the double-logarithmic
form of equation for the various sub-groups.
In defining a "homogeneous" group of households we have been led to
consider what is in fact the very.core of the analysis of budget data, namely,
the problem of deriving some sort of scale which reflects the relative money
expenditure of households with varying needs. In discussing these matters,
too, we have had to mention the Engel curve before being fully introduced to
the ideas behind this curve and the theory on the subject. It is to this topic
that we now turn.
Prals and Houthakker, p.129.
Prais and Houthakker, p.134- 136.



The Fngel curve is a curve which shows the relationship between the
variation of expenditure on, or consumption of, a particular commodity and the
variationof household income. The relationship is named after Ernst Engel, a
pioneer in the field of family budget studies.
In his studies Engel had found that as the level of household income in-
creased the expenditure on different items of the budget changed in such a
way thatthe proportionate expenditure on the more urgent needs (such as food)
decreased, while the proportion devoted to luxuries and semi-luxuries in-
creased. This finding became known as Engel's law.33 It is interesting to note
that this is one of the cases in economics where the observed regularities of
human behaviour were discovered years before a theoretical apparatus was
developed to explain them. 34 For it was not until well into this century, that
a rigorous theory of consumers' behaviour was detailed to explain the broad
features of consumers' behaviour.35
Assuming that the whole set of purchases of a consumer can be arranged
by him in a consistent and unequivocal order of preference, the well-known in-
difference curve analysis shows how the consumer reaches an equilibrium when-
prices or real income, or both, change. Being based on marginal analysis, the
theory assumes that the purchases of any commodity are infinitely divisible
and can vary continuously. Consumer equilibrium is defined as that situation
where the marginal rates of substitution are equal to the ratios of prices of
goodsprevailing on the market. This situation must be constrained by the fact
that the sum of expenditure must not exceed the total expenditures or income
(i.e. the "budget restraint").
Numerous criticismsand modificationsof this simplified view of the world
have been suggested, but in essence it still remains the explanation which
economists generally give when queried about consumers' behaviour, and it
does remain the theoretical basis of most empirical investigations of demand
ignoring dynamic features for the moment. What we have discussed in the
previous sub-section are some of the problems which arise in applying this
theoretical framework to family budget data.
The theoretical analysis just outlined refers to the behaviour of a single
family or individual with a fixed scale of preference. The predictions of the
theory are therefore applicable to budget data only in so far as we are able to
isolate a group of families or households with similar preference patterns. It
is for this reason that the question of defining a "homogeneous" group of
families is of the first order of importance in studies of this kind. If we are
able to isolate such a group of families, the theory suggests that expenditure
on any particular item in the budget is a function of total expenditure; prices
may be assumed constant, because we are considering a cross-section of
Allen and Bowley, p.7.
G.J. Stigler, "The Early History of Empirical Studies of Consumer Behaviour", The
Journal of Political Economy, Vol. LXII, 1954, p.95-113.
J. R. Hicks, Value and Capital (2nd Ed.). Oxford, 1946, Ch. 1-3; Allen and
Bowley, Ch. II.


families at a moment in time and in a given place.36 The statistical identi-
ficationof homogeneity in this sense is most easily undertaken by considering
the distribution of residuals from whatever form of Engel curve we fit to the
data. If we have isolated a homogeneous group of families, then the variance
of these residuals about the curve should be normal and approximately, at
least, homoscedastic. That is to say, if the pattern of preferencesis not strictly
constant the differences from the average should at least be normally dis-
tributed and with approximately constant variance. If this relationship cannot
be shown, then the theory is not applicable to the data.
When individual family records of expenditure are available we may proceed
by considering the Engel curve for each family and the relationship between
these curves and the curve for the group as a whole.37 It can then be shown
that if the preference scale of each family is linear the Engel curves for all
the families in a homogeneous group will also be linear and parallel. The two
coefficients in the linear equation will depend only on the fixed market prices
and the constants of the linear preference scale The demonstration in the
"two-good" case proceeds as follows38:
Let the marginal rate of substitution R be defined as the ratio of two
linear expressions in xl and x2, which are the amounts of the two goods
purchased. Then if p,, and p2 are the respective prices of these goods and
e the total expenditure we have
al + all x + a12 x2 P1
a2+ a21 xl+ a22 x2 P2

2) p1 X1 + P2 x2 = e

where the a's are constant
Denoting p1 x1 = el

P2 x2 = e2
we can simplify eqnl), and by substituting eqn2) into this simplified ex-
pression we get
3) el = kle + Cl

4) e2 = k2e + c2

where a22 a12

(P2) 2 P1 P2
all a22 a12 a21
+ -

(P) 2 (P2)2 P1P2 P1P2

Whatever price differences still remain can be treated separately in an analysis of qual-
ity variations. See Ch. 8, prais and Houthakker.
Allen and Bowley, p.109 ff.
38 d owley, i d.
Allen and Bowley, ibid.


Cl =

c2 =

all a21

(Pi)2 P1 P2
all a22 a12 a21
--- --- +
(P1)2 (P2)2 P1 P2 P1P2
a2 a1

Pl P1
all a22 a12 a21

(Pl)2 (P2)2 PP 2 P1P2

al a2
P1 P2

all a22 al2 a21
-- +-
(P1)2 (P2)2 PlP2 PlP2

It is thus seen that the constants k and c vary in value as tastes and
market prices vary, but are constant when these factors are constant. It is
easily seen further that k is in fact equal to 1, and c = 0, which in
effect means that if each k is greater than zero no one k can be greater than
unity, and if each c is non-negative all the c's must be zero, otherwise the
sum of positive c's must equal the sum of negative c's.
The problem of applying this theoretical apparatus to budget data is, as
already pointed out, that tastes vary from family to family even among a
"homogeneous" group and the question arises, how systematic is this
variation in inherent tastes? When we fit a linear expenditure curve to such
data, since the lines for each family are parallel to each other, the variance
of the residual expenditures will be the same as the distribution of the
constant in the linear equation (i.e. c) about its mean. The distribution of
this constant about its mean will be normal if tastes are norrrally dis-
tributed between the families. Further, if tastes are uncorrelated with the
level of income the variance of c for any given item will be constant over
all income ranges and so will be the variance of the residual expenditures.
In other words normality and homoscedasticity of the distribution of resi-
dual expenditures will be obtained if tastes are distributed normally and
independently of the level of income. Conversely "...if the observed dis-
tribution of residuals is not normal, and particularly if it is not symmetrical,
then it is probable that there are some biased factors in the variation of
tastes of which we have taken no account". 39
Having found the constants in the linear equation it is possible to
assess the relative order of urgency of different items in the budget for
different families. If the linear Fngel curve is approximately satisfied we
have for any given item the relationship
v = ke + c

Allen and Bowley, p.121.


where v is now expenditure by any family on the given item. The average
expenditure over all families for this item may be written as
v= ke+ c

and the ratio of this average expenditure to the average total expenditure
may be written as
It is these averages which are used as weights for the Retail Price Index
and which are summarised in Table IV. This proportion, w, is related to the
constants k and c, which relationship we can observe by combining v and i7
in the following manner:-
v- v= k(e ) + c- c
i.e. v= ke + (v ke)
= ke + (,w k) e.

so apparently, (w k)e = c

The proportion of total expenditure devoted to any item is

v = k + (w k) e/e

so as total expenditure increase, v will decrease if w > k, but will increase
if ~ < k. These conditions, it may be noted, are equivalent to the definitions
of "necessaries" and "luxuries". A "necessary" is defined in the case of
thelinear Engel curve as an item whose curve has a positive intercept (c> 0),
for in this case the proportionate expenditure on the item declines as total
expenditure increases. Likewise a "luxury" is defined as an item whose En-
gel curve has a negative intercept (c < 0), for under this condition the propor-
tionate expenditure on the item increases with total expenditure. A "necess-
ary" can now also be defined when w > k, and a luxury when w < k.

There are a few important points of departure from Allen and Bowley
in the work of Prais and Houthakker. The most noticeable departure is the
decision to treat the Engel curve as basically non-linear. This decision is
based on two grounds. When individual budgets range over wide levels of
income, homoscedasticity and normality in the distribution of residual ex-
penditures are no longer approximated by employing linear formulations.
Secondly, the restriction on the coefficients of the linear equation might not
be considered to be a fair representation of the behaviour of households at
the extremes of the income distribution. None of the c 's can be negative if
all the values of expenditure are (of necessity) non-negative. Nor is it
feasible to assume that there is no saturation point to the consumption of
any one of the commodities, an assumption implied in the linear formulation
by the fact that the sum of the regression coefficients must add up to unity.


A hyperbolic curve does make allowance for the possibility of saturation
points; *on the other hand if all Fngel curves are hyperbolic "...there would
be some level at which the whole of income would not be spent".40 It has
to be decided, therefore, whether one wishes to use a single type of Engel
curve-for all commodities which satisfies the adding-up criterion, but might
not fulfil conditions imposed by theory on the distribution of the residuals;
or whether we wish to adopt different types of Engel curves for different
commodities without the assurance that the curves will in fact add up. Prais
and Houthakker adopt the second line of approach. They experiment with
several forms of non-linear equations and find that the double-logarithmic
(constant elasticity type) and semi-logarithmic types of Fngel curves best
explain the data with which ;hey deal. However, the relationship between
the Engel curve of a particular family and the Engel curve of other families
within a defined homogeneous group is no longer clear.41
What has undoubtedly occurred is that by adopting essentially a be-
haviouristic approach to budget analysis, the question of verifying or re-
futing the specific predictions of consumer theory has been relegated to a
secondary position in the discussion. "If all households in a survey had
the same preferences, were faced by the same prices, and reacted to income
and price changes instantaneously, then observations of expenditure at
different levels of income would all lie on the Fngel curve which is common
to all these families. None of these assumptions is entirely fulfilled in
reality...."42 The problem then becomes essentially one of understanding
how households actually behave, and only slightly in testing the predictions
of theory. In the present study we are not primarily interested in the investi-
gation of Engel curves for detailed items in the budget, but in obtaining an
idea of consumption patterns by observing the behaviour of expenditure on
broad groups of items. With this aim in view it seems sufficient to adopt
the single-equation approach of Allen and Bowley rather than the more in-
tricate exercise of Prais and Houthakker. On the other hand it is necessary
to bear in mind the strictures of the linear expenditure law and to try non-
linear formulations of the Fngel curve as well. The Fngel curves adopted
for analysing the Jamaican budget data are described in Section III.


As stated at the beginning of this section we are also interested in
studying the relationship between consumption and income in the aggregate.
For this purpose we intend employing evidence from both national income and
household budget data in Jamaica. In view of the doubts surrounding the
income evidence from household surveys, however, some detailed discussion
of the general nature of this evidence would seem to be indicated.
It has generally been assumed that, in surveys relating to Jamaican
families, whenever income information is requested this information is likely
to be highly unreliable for a number of reasons.43 One of the main reasons
derives from the difficulty of assessing the income of the large number of
Prais and Houthakker, p. 83.
The situation does not become any clearer when we study, the discussion on grouping
(p. 89 ff. ), nor the "Theory of Many Consumers" on p. 11 ff.
Prais and Houthakker, p. 11.
4Household Expenditure Survey, Jamaica, 1958, p. 14; Rural Household Expen-
diture Survey, Jamaica, 1956, p.9.


small-farmers and own-account workers who inevitably comprise the majority
of such a sample. Further, in both rural and urban areas, this difficulty is
enhanced by the fact that members of the household have several sources of
earned and unearned income. Another difficulty is the general tendency to
understate income for reasons sometimes connected with income tax evasion
in the case of high income earners, and sometimes, on more psychological
grounds, in order to make things appear worse than they really are so as to
incur sympathy and support from official sources. There is also the problem
of seasonal variations in income which makes assessment in the case of rural
households particularly difficult, especially when the survey lasts for a short
period during the year and relies on the respondents' ability to recall the size
and sources of various incomes in the past.
In view of all these difficulties enumerators in the Jamaican surveys were
requested "...to record incomes in the form of specified ranges... and to write
down the actual incomes whenever this was possible". (The average size of
the income range was about 20/- per week). This rule was applied to each
earning member of the household and to each source of earned income. In con-
structing the tables of income distribution which form the basis for Chart 1, it
was assumed that the average income of those single-valued income figures
could be used to represent the average income of each earner within the given
income range. This assumption depends crucially on the relative proportions
of those reporting "exact incomes" to those reporting "ranged incomes". It
also depends, of course, upon the accuracy of the "exact incomes" them-
selves. Evidently the proportion of those reporting "exact" incomes to the
total number of income earners within each range was sufficiently high to
warrant making the calculations.44
The question of accuracy of the income information is well worth con-
sidering in general. We have already mentioned that the surveys were collected
over a short interval and relied on the respondents recalling to memory former
purchases and receipts. A respondent who received varying and irregular in-
come in the course of a year is most likely to report his average income below
the "true" arithmetic mean of all incomes he received unless his income
expectations are high and roseate, which is unlikely if he is an irregular in-
come earner. And if, as is usually the case, the "true" average income of
this respondent happens to be low, it would not be surprising to find that the
sum of his individual expenditures is in excess of his reported income. 45
It may be noted further that whereas the location, income and source of in-
come of all earners in the household were reported, in the case of occupation
and industry only those of the head of the household were recorded (with a few
exceptions). Unless therefore we make the assumption that all income earners
within a household follow the same occupation as the head, the possibility
of analysing income and expenditure patterns by occupation and industry
seems restricted. If, however, we adopt the line of argument which says that
the chief bread-winner in the household is the source of decision-making
withinthe household, then by studying the distribution of his or her occupation
within the sample it is possible to come to some conclusion regarding the
effect of occupation on consumption. It is possible that this line of argument
The number reporting "exact" incomes was not given in the reports but it is possible to
obtain these figures from the individual budgets.
In this connection see Friedman.


applies witn some force in certain types of households, more often the small
household consisting of an adult male, an adult .female and one or two small
children. As the average age and size of the household increase, we should
expect the source of decision-making within the household to become more
diffused among the members but perhaps less so if the. head of the house-
hold happens to be a female! The increase in household size and age may
occur as the result of adding members not connected by blood to the original
family nucleus, in which case the decision-making functions become even
more widely diffused. We have not found it possible, however, to carry our
empirical analysis very far in this direction.


Keynes in the "General Theory..." had placed great emphasis on the in-
fluence of income in determining consumption, one of the two major components
of aggregate demand. He defined a "propensity to consume" as the functional
relationship between "...a given level of income in terms of wage-units, and
... the expenditure on consumption out of that level of income". 46 For psycho-
logical and other reasons he argued that this function would be fairly stable
and that the increment in consumption for any given increment in income (i.e.
the marginal propensity to consume) would be less than one, especially when
we are considering short-run changes.47 The marginal propensity to consume
turns out to be a crucial factor in the determination of the volume of employ-
ment: "...employment can only increase part pass with an increase in in-
vestment; unless, indeed, there is a change in the propensity to consume".
Following the statement of this relationship numerous attempts were made to
estimate its empirical validity. 48 Use was made of both time-series and cross-
sectional data, and true enough, a high correlation between net income and
consumption was found in most studies. But estimates of the income
coefficient were widely different in the United States some estimates were
as high as 0.90 and as low as 0.60, with heavy concentration around 0.75.49
Such wide estimates rendered the predictive power of the relationship well-
nigh nil. Naturally enough, attempts were made to include other variables
in the relationship and to estimate from a structure containing other macro-
economic relationships along with the consumption function. Perhaps the
only certain implication of all these empirical studies is that, if undeflated
figures are used, both average and marginal propensities to consume tend to
be higher than when the variables are deflated.so At the same time these
J.M. Keynes, The General Theory of Enployment, Interest and Money. New
York and London, 1936, p.90.
Keynes alsoenvisagedthat the marginal propensity to consume might change over the
course of the business cycle,, as well as be subject to possible trend factors.
See Ruth P. Mack, "Economics of Consumption" in A Survey' of Contemporary
Economics, Vol. II, ed. by B.F. Haley,, Chicago, Illinois, 1952, pp. 39-78 for an excellent re-
view of these empirical essays.
Ibid. p.65.
Ibid. p.66.


studies point to numerous factors other than income which ought to beincluded
in the consumption function. It was pointed out by Duesenberrysl and others
that relative income rather than absolute income helped to explain better
certain observed constancies in the relationship between consumption (or
savings) and income, over time. Friedman further suggested that the importance
of relative income status was subsumed under the general problem of income
expectations, which latter factors are the crucial determinants of consumption.
Both of these suggestions must be viewed as alternative hypotheses to the
original one of Keynes, which states the consumption function in terms of the
absolute level of current consumption and income. Several other interesting
suggestions have been made,52 but in the pages that follow we shall be mainly
concerned with examining these two.
As indicated in the previous paragraph, certain observed constancies
in the average and marginal propensities to consume, over long periods of
time, led investigators to cast doubt upon the Keynesian formulation of the
consumption function. If the marginal propensity to consume were constant
and less than one, how is it that, over long periods of time, the percentage of
income saved revealed no change, in spite of the fact that a substantial rise
in real and money income had occurred? Put another way: if the marginal
propensity to consume is less than the average propensity to consume (as
was empirically found when the variables were measured in absolute terms)
how could the average propensity to consume remain constant over time?53
Yet this in fact was what an examination of available data revealed.54
Because this empirically-observed constancy in the average propensity
could not be explained entirely by errors and imperfections in the data, an
earnest search for alternative formulations of the consumption function was
initiated.55 Duesenberry argued that the weakness in theKeynesian hypo-
thesis lay in the failure to consider the interdependence of consumers' pre-
ferences. Considerations of this sort led him to formulate the aggregate con-
sumption function, not in terms of the absolute level of income, but in terms
of the position of consumers in the income distribution of the community:
"... for any given relative income distribution, the percentage of income saved

51J.S. Duesenberry, Income, Savings and the Theory of Consumer Behaviour. Cambridge, 1949.

52For which suggestions see Mack, op. c t., p.66.
53If C = a+ bY
where C is consumption (real or money)
Y is income (real or money )
then the average propensity to consume is
C, and the marginal propensity to consume is A C = b
y AY
where b < 1
and C> b

we can define the elasticity of consumption with respect to income as
77 AC Y AC Y

that is, rc ,y is equal to the marginal propensity divided by the average propensity. This makes 7c ,y
necessarily less than unity, the implication therefore being that given proportionate changes in income
lead to less than equal proportionate changes in consumption at all points on the curve. C must therefore
decline. The results thus not only contradict the assumptions, but violate the observed constancy in the
average propensity to consume, even when large absolute changeshave occurred in income (cf. Friedman,
p. 41).
54Friedman, p. 44.
55Mack, p. 66; Duesenberry, pp. 1-5.


by a family will tend to be a unique, invariant, and increasing function of its
percentile position in the income distribution". 56 The aggregate savings pro-
portion will therefore be "... independent of the absolute level ofincome".
Further theobserved tendency for annual data to showvariation in the savings
ratio is to be explained by the fact that consumption relations are not in fact
reversible in time, as earlier formulations of general demand theory had
assumed, but that the response of consumers to changes in income are
different depending on whether that income has increased or fallen. The
justification for assuming consumer preferences to be interdependent lies in
sociological and psychological features of "local" culture. Frequent contact
with superior goods tends, for any given family, to increase as consumption
expenditures increase. "When that occurs, impulses to increase expenditure
will increase in frequency and strength, and resistance to them will be in-
adequate. The result will be an increase in expenditure at the expense of
saving".57 Duesenberry calls this the "demonstration effect". The fact that
people know about the existence of superior goods, it is to be noted, is not
sufficient to cause this effect. They must also have frequent contact with
So manyobservers havecommented on theexistenceof this phenomenon in
Jamaica, that it seems worthwhile to consider its relevance in some detail. It
has been argued, for instance, that the desire to emulate or to attain to a level
and pattern of consumption displayed by either the American or British con-
sumer or by the former elite of Jamaican society, viz. the white planters, is
a basic drive, especially among urban and middle-class consumers. This is
one of the explanations usually given for the so-called "credit inflation", the
excessive amount of hire-purchase business and the large number of American
cars, high-quality household durables, etc. which are found around the homes
of most middle-class families in modern Jamaica. It is further argued that this
"demonstration effect" has fast slipped down from the middle-class to the
mass of the population. Moreover because these demonstrably "superior"
goods are imported, this explains the pressure on the balance of payments,
which pressure has been building up ever since the end of the second World
There is no doubt that there is some truth in these observations. Jamaican
society is no different from other western societies in the force of the drive
among consumers to exhibit, at any rate, higher standards of living. This drive
is carried to such extremes that if there were a law in respect of bankruptcy
applicable to families as such, so many would be found to spend so much out
of borrowing that they would have to be classed as bankrupt. But there are a
number of other ways to the attainment of high status in the Jamaican commun-
ity. As in other communities occupational success, membership in certain ex-
clusive clubs, family connections, the attainment of high educational qualifi-
cations, etc. are all means of attaining relatively high prestige. But these
means themselves depend to a large extent on the amount of assets previously
acquired by the individual. On the other hand present "demonstrative" con-
sumption is a means of obtaining the same goal and this method is especially
democratic when credit is easy.
Mobility within Jamaican society is, further, not as serious a constraint
today as it was in theold colonial days. In those days before the second World
War and prior to adult suffrage (which came in 1944) it was not easy to rise in
the social ladder. Most Jamaican institutions were then managed largely by
Duesenberry, p.3.
Duesenberry, p.27.


British or expatriate personnel. This was especially true of the civil service.
There was then no doubt that colour differences coincided with differences in
status and all that went with it. The revolts in 1938 were an expression by
Jamaicans of their dissatisfaction with the existing immobility of the social
structure as well as a labour dispute. The more articulate members of slave
descendants in Jamaica were really demanding and eventually won not equality
of status as such, but a wider distribution of the means of attaining this
equality. It is this feature of our history which, I think, explains some of the
present anomalies in Jamaican society. The cry in the thirties was not for the
enactment of a civil rights bill, a bill against social or racial discrimination
or any such egalitarian legislation, it was a cry for the opportunity to
be equal, irrespective of race, colour or creed. It has become painfully
obvious within recent years that such a simple goal is not enough, unless we
can ensure a wide distribution of employment opportunities as well. The
attainment of this latter goal cannot be granted by a colonial office but must
be planned for by the people who have the reins of government.
The problem of economic development then must be solved before the
ideal of "equal opportunity" equal possibilities of obtaining the status
symbols can prevail. And the tardiness in solving the problem (some head-
way has been made) is the main explanation for the frustration demonstrated
recently by some of the poorer members of the Jamaican community, the in-
crease in emigration in recent years, the tendency for families to spend more
than their earnings and the consequent pressure on the balance of payments.
The actual empirical implication of the idea of the "demonstration effect"
is that ... after some minimum income is reached, the frequency and
strength of impulses to increase expenditure for one individual depend entirely
on the ratio of his expenditures to the expenditures of those with whom he
associates". 58 Frustration arises when the consumer has to reject expenditure
because of the unfavourable comparison of his income with the income of
those whose expenditures he wishes to emulate. Again, making allowance
for the interdependence of preferences: "Attitudes towards future consumption
depend on current consumption standards. Current consumption standards ...
are influenced by other people's consumption behaviour, and desires for
future consumption will be influenced in the same way".s5 By making these
assumptions about individual consumer behaviour, we may state the general
consumption function for each individual as follows:

C = f(Y1/, Y2/R, ......... Y n/R; A/R; r .... r) .
where C is consumption expenditures of the given individual; R is a weighted
average of the consumption expenditure of other individuals, the weights
being the importance attached by the given individual to the consumption of
each of the other individuals; Y1, Y2 .. Yn are expected incomes; A repre-
sents current assets and the r's represent interest rates. It is then pos-
sible to see that only a change in the income distribution will affect the
proportion of income consumed, other things being equal. For if all incomes
increase by the same proportion, every individual will increase his con-
sumption in accordance with the absolute income hypothesis, but on the
present formulation this will mean that the denominator, R, in the present
expression will also increase and the increase in C will not be as great as

58Duesenberry, p. 33.
Duesenberry, p. 34.


"when each consumer took his neighbour's consumption as a datum".
The fact of the matter is that, although the absolute level of income of each
individual has increased, his relative income position has not altered. His
tendency to increase savings is, therefore, in terms of the "demonstration
effect", curtailed by the inducement "... to give up some of his increased
saving in favour of further increases in consumption".6 When summed over
all individuals the net effect is to stabilise the observed savings proportion
and what is the other side of the coin the average propensity to consume.
This is essentially a conclusion derived from considering two equilibrium
conditions, but it is quite in keeping with the observed tendency, at a moment
in time, for families with higher relative incomes to have a higher savings and
consumption ratio than families at the lower end of the income distribution.
The details of the hypothesis as stated are not open to rigorous testing.
but Duesenberry was able to confront some of its implications with available
data. 61 He finds in general that his hypothesis is not inconsistent with com-
parisons of families grouped by race and location; it appears also to be
consistent with the behaviour and variation of the savings proportion over
time. 62
If non-income factors over time impart an upward bias to consumption,
then it is possible to explain the observed constancy in the savings pro-
portion by the Keynesian absolute income hypothesis. Duesenberry finds that
with most of the factors liable to have this effect their behaviour over the
period investigated was not such as to warrant the observed constancy in the
average propensity to consume. There is therefore a prima facie case for
not rejecting his hypothesis.
Further, by employing the idea of the irreversibility of the short-term
aggregate consumption function, he is able to throw light on the observed
variation in the savings proportion over the course of the business cycle:
"... there does not seem to be much doubt that past income has an influence
on current consumption and saving".63 The highest level of income attained
in the past is particularly important in determining current levels of saving
and consumption.
Starting from a rather different position Friedman develops what he calls
the "permanent income" hypothesis, which, he claims, includes the"relative
income" hypothesis and the "wealth-income" effect (associated with the
names of Pigou and Haberler, to mention only two) as special cases. By
distinguishing between "permanent" and "transitory" components of income
and consumption he is able to define a stable relationship between permanent

Duesenberry, p.38.
6Duesenberry, p.48 ff.
In making comparison over time it is necessary to make allowance for the influence of
non-income factors on the savings ratio. The main factors referred to usually in the theory of
consumers' demand are (1) the interest rate, (2) the relation between current and expected in-
comes, (3) the distribution of income, (4) the age distribution of the population, (5) the rate of
growth of income, (6) changes in attitudes rc.ards savings, (" 1 "t+anisation, (8) the introduction
"f tc-w products, and so on. See Duesenberry, op. cit. P 57 and PP. 41-46-
Duesenberry, p.85,


Thus if
y = p + Yt

and c = + c

where y and c represent measured income and measured consumption respec-
tively, and the subscripts p and trefer to permanent and transitory components,
the permanent income hypothesis states that

c = k(i, w, u)y

where i represents the rate of interest, w represents the ratio of non-human
wealth to income; u refers to "the consumer unit's tastes and preference for
consumption versus addition to wealth;"64 k is a constant which indicates
that, if the variables within the brackets are constant, the ratio of the per-
manent components of consumption to income will also be constant.
By adopting this method of approach Friedman is able to show that a
marginal propensity to consume of less than one, when i, w and u are fairly
constant, is no reason for rejecting the formulation but ought rather to be an
inspiration to look for other means "... of establishing a correspondence
between the theoretical constructs and the observed magnitudes".65 Hence
the distinction between permanent and transitory components.
The notion of "permanence" which Friedman employs is equivalent to
the notion of expectation in statistics, i.e. a mean of means. "Transitori-
ness", on the other hand, implies randomness or ... factors likely to be
treated as "accidental" or "chance" occurrences ..." The effect of making
this distinction, it is true, is to give a great deal of flexibility in application
and allows various types of material evidence bearing on consumption and
savings to be included in the investigation. The investigator has considerable
latitude in the choice of which factors shall be considered permanent and
which not. The study of the consumption function is thrown firmly back into
the speculative field instead of becoming purely a statistical exercise in find-
ing ways and means of obtaining efficient, unbiased estimates of parameters
in stochastic equations of simultaneous equations models. "Ve are going to
treat consumer units as if they regarded their income and their consumption as
the sum of two ... components, and as if the relation between the permanent
components is the one suggested by our theoretical analysis. The precise line
to be drawn between permanent and transitory components is best left to be
determined by the data themselves, to be whatever seems to correspond to con-
sumer behaviour. 66
Inorder to apply thehypothesis to observations, however, further specific
assumptions concerning the joint probability distributions of the components
are necessary if the hypothesis is tobe refutable. The preliminary assumptions
which Friedman makes are that the correlations between transitory and per-
manent components, and among transitory components themselves are zero:
errors of measurement are presumed to be contained within transitory com-
ponentsof both income and consumption. The form of the equation is specified
Friedman, op cit. p.26.
Friedman, p.20-21.
Friedman, p.23.


in terms of logarithms or natural numbers. In order to see what the hypothesis
involves let us assume a linear relation to be estimated from budget or time-
series data. That is to say:
c = o< + !3y
As is well-known the least squares estimate of / is given by

b= (c- c ) (y- y)

S(y -- )2
and the estimate ofais
a = c by

but c = CT + ct,

Y = Yp + yt' and

c = k.y

and if we assume zero correlation between the permanent and transitory com-
ponents as well as between the transitory components themselves, by
substituting in b we get

b = ,( yp )2
= k. P
y(y- 7)2 y
That is to say b is an estimate of the product of the "true" relationship be-
tween consumption and income and "... the fraction of the total variance in the
group contributed by thepermanent component"; b will therefore depend on the
extent to which observed changes in income reflect changes in "permanent"
income, and the proportion of "permanent" income devoted to consumption.
The first is measured by Py and the second by K 67 If we now substitutethis
value of b into a, we get
a= -- kPy t + k(1-Py) y

and the elasticity of consumption with respect to income is given by

7 = dc y = b. y = k. P. y
dy c c Y -"

If the mean values of income and consumption for the observations are taken
to represent "permanent" income and "permanent" consumption, i.e. the
transitory components average zero, then we have

y = YP
: = CP

which gives c = k. y
v 1
or =
67 c k
Friedman, p. 32. ff.


and the elasticity of consumption with respect to income at the mean levels of
consumption and income now becomes

n= k.P y


By similar reasoning it can be shown that if the regression of income on con-
sumption is considered
i.e. if y = a + b c
we have b = 1 P

c (c- ?)2

Further a = t -1Pcc + 1(- P) c
T k
and the elasticity of consumption with respect to income is given by

dc y 1 y k y
dy c b c Pc c

and if t= ct = o then at the mean levels of income and consumption
If instead of a linear formulation of the consumption function we adopt
a logarithmic form (implying a multiplicative specification between consump-
tion and income) then, denoting the parameters and variables of the function
by capital letters, we now have as analogues of the linear case

C= K(i, w,u) + Y

Y =Y +Yt
P t
C = Cp+ C

Cov. Yt, Y Cov. Ct, C = Cov. Y t C

and applying this model to the data we again have

C = o + fy

and the least squares estimates of the parameters P andocgiven by

B = (C -) F)
y(y- y)2

A = C By


If we now substitute the two identities into B we get

I (C-C)(Y-1'Y) = Z(Cp+ t- CP- c (1+ Y C- p- Yt

and since zero correlations are assumed between the two transitory components
and any permanent component with any transitory component, we are left with

(C-C) (Y-Y) = (C Y C Y C Y + C Y )
=- Cp (Y -Yp) Cp (Y- Yp)

= (C C ) (Y Y )

Substituting C == K + Y in B
we get 1 ( + Y }) (Y Y)

S(Y- Y)2

S(y y 2

Likewise A = C + B (I + Y)

=K + Y+ +C- P (Y + Y)
p y p t

=K + C t-YtP + Y (1-Py)

The elasticity of consumption at the mean logarithmic values of income and
consumption is then given by

= dc = B = P
For the regression of the logarithm of income on the logarithm of consumption
we have

A =-K + Y C P + C (1-P)
t tc p c

and 77 =1

Since the elasticity of consumption with respect to income is constant every-
where on each line the logarithmic formulation does not need the assumption
that the mean transitory components of income and consumption are zero. This
is one of its attractive features.


It is not too difficult to see, graphically, the different implications of the
Friedman hypothesis as compared with alternative formulations of the con-
sumption function.

0 Yl


In this chart a linear formulation of the hypothesis in terms of measured in-
come and consumption is adopted. This is shown by the line c = a + by. It
is again assumed that ct= yt = 0 and that

Cov..yt,yp= Cov. ct, c = etc. ... = 0

The point of average income, D, on the line therefore represents the average
level of permanent income for the group of consumers under consideration.
Since there is zero correlation between permanent and transitory components
of income, the group of consumers whose average level of measured income is


Yo is most likely to have an average permanent income of less than : ...
the winners in any particular set of races may well be better on the average
than the losers but they are also likely to have had more than their share of
good luck".68 Since further zero correlation is assumed between transitory
components, the consumption of that group will be the same thing as their
average permanent consumption which is given by the equation.
CP = k. yp.

Had the average transitory component of income for the group been zero, yo
would be the same as yp, and cp would be measured by the distance yoE. But
the mean transitory component for the group is positive. Mean permanent in-
come is therefore less than mean measured income, consequently the average
measured consumption is yoF which is less than y0E. Similarly, for consumers
whose income is below y, say at the level yl, the average transitory component
of income is likely to be negative; average permanent income is greater than
average measured income, and consequently ylT is greater than ylM. At the
mean level of income all transitory components average zero by assumption,
so measured income is identical to permanent income and measured consump-
tion is equal to permanent consumption. The regression line and 0 E therefore
intersect at D.
The point G defines a permanent income level yp, which is associated
with a permanent consumption yoF -- which, as we have seen, is the average
consumption of the group at income level y,. Ypo is therefore the average per-
manent income of the group at Yo. Similarly yp is the average permanent in-
come of the group at yl.
So ot = Yo -Ypo

Yl t = Yl Yp
and the ratios
Yo Ypo Y- p 1
ro ; r
Yo y y1 -

measure the fractions of the deviation of measured income from the mean in-
come attributable to yo and y t respectively. If rois the same for all income

ro = r1 = r2= ...= -y
and the "true" regression line will itself be linear. From this it can be seen
that if Py is zero, average consumption is the same for all income groups. On
the other hand if P = 1, IF coincides with OE.
The line OH is the 45 degree line which measures the equality of con-
sumption and income. To the left of point J, measured savings are negative
and to the right of J, measured savings are positive. The ratio of average
measured consumption to average measured income decreases as average
measured income increases. The permanent income hypothesis therefore yields
conclusions which are in keeping with the general pattern revealed by data on
consumers' expenditures.
It must be noted, however, that problems of some magnitude arise in es-
timating by least-squares the relationship assumed. Stress ought to be placed
particularly on the least-squares' assumption that the dependent variables are

68Friedman, p.35.


measured without error, and that the regression coefficients define the re-
sponseof the dependent variable to given fixedvalues of the independent vari-
ables. Since the permanent income hypothesis is based crucially on the
assumption that all the variables are subject to errors of measurement as well
as other types of errors summarised in the "transitory" component, the ques-
tion of "errors in the variables" becomes a prime issue. A particularly lucid
account of how this problem can and does often arise in samples of economic
data is given in Beach.69
Nevertheless by using the mass of information relating to consumers'
expenditures in the United States, Friedman is able to show, like Puesen-
berry, that the assumed relationship is not inconsistent with the' observed
behaviour of consumers taken, either at a moment in time or over time, granted
the plausibility of the assumptions and data relating to i, w, u etc. It is
therefore worth the trouble to see at what points the two hypotheses differ.
On the face of it both hypotheses seem to be similar in that they are
capable of explaining the observed constancy of the average propensity to
consume over long periods of time and the observed tendency for families
with high income to save a greater proportion of their income than families
with low income. Friedman, however, points out that the relative income
hypothesis "... in effect regards measured income as permanent income; it
reconciles the observed stability of the average propensity to consume over
time with the tendency for the propensity to decline with income at any time
by allowing for the effect of a rise in income on relative income that cor-
responds to any given absolute income".70 This weakness he argues is quite
significant in terms of the predictions of each hypothesis. If in comparing
farm and non-farm families, for instance, we consider relative income to be
important there is no reason why its impact on the two samples of families
should be different; savings among farm families should be the same as
savings in non-farm families at the same percentile position in the income
distribution; the marginal propensity to consume and the income elasticity of
expenditure should also tend to be the same for farm and non-farm families.
Yet the data shows that this is decidedly not the case. 71 If on the other hand
"... relative measured income is important as an index of the ratio of per-
manent income to average income, there is every reason why it should have
a different effect for farm and non-farm families".72 The notion of permanent
income therefore apparently has greater predictive power than the idea of
relative income.

69E.F. Beach, Economic Models, New York, 1957, pp. 148-152. See also Richard Stone. D.A. Rowe, W.J.
Corlett, R. Hurstfield and Muriel Potter, The Measurement of Consumers' Expenditure and Behaviour in the
United Kingdom, 1920-1938, Cambridge, 1954, as well as R. Stone, "The Analysis of Market Demand" inJournal
of the Royal Statistical Society, Vol. CVIII, 1945. There are as yet no sure methods by which "errors in the
variables" can be overcome, since little is known about their original distribution. However, it does appear that by
transforming theobservations prior to estimation, it impossible to minimise the effect of most errors likely to arise
in practice. For this purpose first-differencing and/or logarithmic transformations have often been employed.
(This is another attractive feature of the double-logarithmic statement of the hypothesis.) If no such method is em-
ployed, it is highly probable that if the regression coefficient is at all an estimate of the true parameter, it will
tend to be a biased estimate, and probably biased downwards.
70Friedman, p. 167.
71Friedman, p. 63.
72Friedman, p. 168.


It is now necessary to see the extent of the increase in predidtive power
of both these hypotheses over the absolute income hypothesis of the Keyne-
sian type.
In an article by Tobin,73 it had been argued that continuous budget
data, and data on consumption patterns in different cities seemed to lend
support to the absolute income hypothesis, whereas time series data and
data on the savings patterns of Negroes and Whites seemed to favour the
relative income hypothesis. In view of which conflict, Tobin suggested a
modified version of the absolute income hypothesis in which the amount of
financial resources(in addition to income) is introduced as a variable likely
to affect consumption.74
Friedman finds that Tobin's analysis of the data is ".... incomplete
and inefficient". Further he finds that the data relating especially to inter-
city consumption patterns are not such as to provide a critical test of the
two hypotheses. In particular he points out that where the distribution of
income is not markedly different between the samples under comparison,
any observed discrepancies between the predictions of the absolute and
relative income hypotheses are rrost likely due to chance, since on theore-
tical grounds we expect a difference in the predictions only when the samples
differ widely in their income distribution. In effect, then, Tobin's evidence
favours more the relative income hypothesis than Tobin himself thought.
SThe permanent income hypothesis on the other hand is apparently un-
able to explain differences between communities not explained by either
the relative or the absolute income hypotheses.76
It may be recalled that in our earlier discussion of the demonstrationn
effect" we had outlined the argument in terms of the Jamaican environment
and had implied that this effect was an operative factor in explaining why
the savings ratio was low in countries with low real income as opposed to
countries with higher levels of real income per head. It may be noted that
thisis also the implication of the Keynesian hypothesis. In contradistinction
to both these hypotheses the notion of permanent income implies that a
country with a relatively low level of real income may have either a high or
low savings proportion. This is because the savings proportion depends
more upon the effects of a change in income than upon the absolute level
of income: ".... a rise (in income) expected to continue tends to raise per-
manent income relative to measured income and so to raise consumption
relative to measured income ... wether this effect shows itself is likely to
depend critically on the source of the rise in real income" 77. We may very
well find that self-financed growth gives rise to high savings ratio if the en-
vironment secures a high rate of return on domestic capital, for "... the high
rate of return to savings is an offset to the high ratio of permanent to current
income and may well be more important" 78 Conversely externally financed
growth may give rise to a low observed savings ratio.
J. Tobin, "Relative Income, Absolute Income and Saving" in Money, Trade and Ec-
onomic Growth. Essays in honour of J.H. Williams, ed. by D.M. Wright et al, New York,
1951 4
57Friedman. p. 170.
76Friedman. p. 174 also pp. 179 ff.
7Friedman. p. 181--182.
Friedman. p. 234.


Likewise we can see that a great (observed) inequality in incomes has
no effect on savings in so far as this inequality reflects only differences
in permanent income. "What is favourable to a high savings ratio is no
inequality per se but uncertainty......", and in a capitalist system this
uncertainty must be such that it does not reduce the rate of return on cap-
ital79 ".... the fluidity introduced into relative income status, the emergence
of fresh possibilities of moving from one class to another of possibilities
of large gains and large losses over short periods of time .... these changes
are favourable to the savings ratios".80
The counter argument here if one wished to maintain unimpaired the
idea of the "demonstration effect" would be to reply: agreed, so long as
transitory components of income do not help to determine permanent (or
transitory consumption), for if this is the case great inequality of income
deriving from "transitory" factors will cause permanent consumption to be
high and consequently yield a low savings ratio. The purchasing of durable
equipment, for instance, as the result of a windfall not likely to be repeated
will be maintained if the unit who experiences the windfall suddenly finds
itself in a position where the prolonged maintenance of such purchasing
becomes important for reasons connected with social esteem. If such atti-
tudes prevail in a community, we may very well find the savings ratio to be
low where, in terms of the Friedman hypothesis, it ought to be high.
At the same time the permanent income hypothesis leads to a type of
long-term aggregate consumption function which is eminently reasonable.
The formulation is one in which current consumption is largely determined
by past incomes.81 The use of this type of lagged consumption function
has proved to be fruitful in the analysis of economic fluctuations.


In developing any theory it is important to define clearly the variables
which enter into the analysis. Unfortunately in developing a theory of aggre-
gate consumption any attempt at precise definition of the main variables is
surrounded by innumerable obstacles. The difficulty here stems mainly from
the fact that if the theory is to be of any use in an empirical sense, thevary-
ables have to be defined in a "dynamic" rather than a "'static" context.8 2
The importance of this last statement will become clearer as we expose the
ideas which economists have put forward in order to explain the behaviour
of aggregate consumption.
In pure "dynamic" theory the calculation of income is a problem of find-
ing "... some sort of standard stream of values whose present capitalised
value equals the present value of the stream of receipts which is actually in
prospect".83 This way of calculating income stems from defining it as the
maximum value which an individual can consume during any given period

7Friedman. p.235.
80 Friedman. p. 236.
81Friedman. p. 229.
82The word "dynamic" here refers to the method of comparative statics rather than to the
"dynamic systems" developed bypostvKeynesian economists such as Professors Phillips,Harrod,
Domar, etc. and Hicks himself. See R.G.D. Allen, Mathematical Economics, (2nd Ed.) Lon-
don, 1959, especially Ch. 1--9. See also Gardner Ackley, Macroeconomic Theory, NewYork,
1961, pp. 249- 251.
83Hicks, p. 184.


(say, a week) with the expectation of being as well off at the end of the
period as he was at the beginning, 84 Implicit in this definition is the notion
of constancy of income over time, regardless of what in fact happens to be
the expected stream of receipts.
Under stationary conditions, where the consumer unit is completely
certain that his receipts in each of a definite number of time periods will be
constant, and that prices and the rate of interest will also be constant, money
income as defined will also be constant in successive time periods. If the
price level varies in accordance with expectation, other things being equal,
income will be constant in the real sense only. If the rate of interest varies
in accordance with expectations, other things being equal, income will also
vary in money terms "... in such a way as to make the capitalised money
value of all future values (in the standard stream) constant from week to
week". 85
It is these ideas of income which in pure theory enter into the deter-
mination of consumption. The gap between the theoretical notion of income
and measured income is easily seen when we note that, in the real world,
prices and interest rates rarely remain constant over time, and further than
that, expectations are not always realized. The relationship between ob-
served consumption and observed income, therefore, could hardly be inter-
preted as defining the theoretical relationship between consumption and
income. By making a few assumptions, however, it is possible to obtain some
insights into the relationship to be expected between the observed magni-
tudes. If we denote
C, C, C3'...... C = consumption in successive years (1, 2 ....... n)
Y1, Y2' Y3 ....... Yn = income in successive years (1, 2, ..... n)
i = the rate of interest
p = the price level
u = the consumer unit's tastes and preferences,
then our theoretical consumption function is given by
C = f (W)
where W is some function of Y1, Y2, .... etc. whose shape is dependent on
variations in i, p and u. We now wish to make certain predictions about fc.
Traditional analysis of f has quite sensibly proceeded by considering two
fundamentally different sets of conditions.86
Under the first set of conditions it is assumed that the individual isin a
situation of completeobjective and subjective certainty. He is then faced with
the problem of choosing ... a most-preferred collection of streams of com-
modities, out of the various collections of streams which (he) could expect to
be able to purchase out of a given expected stream of receipts". 87

84Hicks, p.172.
85Hicks, p. 184
86Friedman, Ch. 11.
I. Fisher, The Rate of Interest, New York, 1907, Ch. VI and VII.
1. Fisher, The Theory of Interest, New York, 1930, Ch. X and XI
Hicks, Ch. XVIII.
Modigliani and Brumbeng, "Utility Analysis and the Consumption Function: An Interpreta-
tion of Cross-section Data, Ch. XV of Post Keynesi an Economics, ed. by K.K. Kuri-
hara, New Brunswick, 1954.
Hicks, p.227.


It would be entirely unrealistic to assume that the expenditure plan of the in-
dividual consumer were complete in all details for all periods extending into
the future, for this would mean that the consumer knows at any given time ex-
actly what his pattern of wants will be in the future.88 However, while this
might be a manifestly absurd assumption on which to build a theory of present
and future consumption, it does not seem entirely unrealistic to use the
assumption in explaining current consumption. Given then the current pattern
of wants of the consumer, the prices of consumer goods and services and the
individual's expectations of future prices, the consumer plans to make money
expenditures El, E2, E, ... out of his current income Rl and his expectedin-
comesR2, R3, R4, ... in the following weeks.39 For any given individual, it is
obvious that these streams of income and expenditure cannot be carried on in-
definitely. Let us assume the expenditure plan continues for n time periods.
The detailed analysis of the case where n= 2 has been thoroughly explored by
Fisher90. Hicks has however developed the ideas in more general terms.
The Stream
R E1; R2 E2; R3 E; etc.
may on balance after n time periods be positive, negative or zero, Assuming
that the individual does not plan to encroach on his initial stock of assets, we
need consider only cases where on balance

n t n Et
t =1 t=1
Where the inequality holds, the consumer unit plans to be on balance a "len-
der" over the entire n periods. His lendingg" might take the form of (a) hold-
ing money or (b) holding securities. If all his lending take the form of (b), at
the end of the n periods he will have a capital sum C, in securities for pass-
ing on to relatives or for adding to future resources. Alternatively, we can
assume the sum C is spent during the nth period. The stream R1- E1; R2 -
E2; etc. is then, on balance, zero.
It is nowpossible to analyse theplanning of expenditure in familiar terms.
If the price of some commodity, say X, rises, then with the elasticity of ex-
pectations equal to unity, current and expected prices of the commodity will
all rise in the same proportion. These prices, when discounted by the rate of
interest (here assumed constant), will also rise by the same percentage. The
relative prices of X at different dates will remain unchanged, and so we need
consider only the substitution effect against X in favourofother goods. This
effect will ,of course, be positive and, if X is not an inferior good, the income
effect against X will also be positive. But while these effects are clear, itis
not possible to say in what manner the current and future demand for X will be
reduced. Current demandfor X mayor may not be reduced by the priceincrease.
If, on the other hand, the price of X rises, but the elasticity of expecta-
tions iszero, therewill beno increase in the expected pricesof X. The current
price of X relative to future expectedprices of X will have increased and there
will be a substitution effect not only in favour of other commodities but in
favour of future purchases of X itself. The income effect, on the other hand,
will be small, since it derives solely from the current t increase in the price
of X. The net effect, then, may well be to cause the consumer unit to postpone
his expenditure.
88In fact individuals plan only part offuture want-satisfaction, cf. expenditure on durables.
89The treatment here follows closely that of Hicks, p.229 ftf.
90Also Friedman, Ch. 2.


If the price of X rises, and the elasticity of expectations is greater
than one "... then we have to deal with a rise in expected prices more than
proportionate to the rise in current price".91 The substitution in favour of
present purchase, deriving from the greater rise in expected prices, may
well be stronger than the substitution against current purchase of X. The
income effect against current purchases may also be overwhelmed by the
aggregate substitution effect, and the net result may serve to increase cur-
rent demand.
In a similar way we can consider the effects of changes in the interest
rate. A rise in the rate of interest lowers discountedprices of future purchases
relative to theprices of currentpurchases, thegeneral effect of this rise would
therefore seem to be to lower current consumption. The ultimate result, how-
ever, depends upon the income effect.
Now a change in the rate of interest affects the capitalised values of
the planned stream of expenditures as well as the capitalised value of
receipts. A rise (fall) in the rate of interest will reduce (increase) both these
capitalised streams. The crucial point is whether or not the effect will be
the same on both streams. In general, if the average period of the indivi-
dual's receipts stream is less than the average period of his planned ex-
penditures stream, the individual evidently plans to be a "lender" in the
sense that he plans to spend less than he receives in the present and near
future and to spend more than he receives in the distant future. A rise in
interest rates will therefore tend to increase the expenditure of such an
individual. In this situation substitution and income effects work in opposite
directions and current expenditure might increase, diminish or remain con-
stant. There is no way of telling.
If the individual's average period of receipts is greater than his average
period of expenditure, he evidently plans to be a "borrower'' in the short
and medium run and "lender" in the later periods. An increase in the in-
terest rate will therefore have the effect of decreasing his current expen-
ditures. In this situation both substitution and income effects operate un-
equivocally in the same direction. Over the market as a whole, if "bor-
rowers", as defined (a) tend to outnumber "lenders", or (b) react more
speedily to interest rate changes, or (c) are equal in strength in terms of
numbers or speed of response to interest-rate changes, a rise (fall) in the
interest rate will tend to increase (lower) current consumption expenditures.
We have so far been discussing the effect of changes in prices, the
rate of interest and expectations, on the consumption of a given individual,
with the stated assumption that the difference between receipts and ex-
penditure in any given "week" is reflected in a change in the individual's
holding of securities. B1ut what if the individual does not choose to balance
his expenditures and receipts in this way? Suppose instead he prefers to
balance his streams of receipts and expenditures by holding or partL ng with
money, would the argument be affected if we continued to assume stationary
conditions? If money is considered as some sort of durable good, it is easy
to see what will happen to the desire to hold money if the interest rate
changes, other things being equal. A rise in the rate of interest, as we have
seen, tends to diminish the demand for present commodities but to increase
Hicks, p. 332.


the demand for future commodities. The same would apply to the demand for
money. Likewise, a general price inflation would tend to make the individual
increase his holdings of money relative to his holdings of other commodities.
But supposing rates of interest and prices remain constant, will there
be any reason why an individual should choose to hold money rather than
securities? The answer to this question is definitely in the affirmative.
First of all money might be held in order to balance out the uneven
periodicity of receipts and expenditures. Secondly, money might be held in
order to make some unforseen expenditure. This second portion of money
held by the individual is likely to be affected by interest rate changes but
perhaps only mildly, "... it probably bears a fairly constant relation to the
aggregate volume of expenditure".92
The demand for money will, on the other hand, be nil if the individual is
confident of being able to execute a plan of spending all of a constant flow
of future receipts in each successive time period.
The analysis so far has been in terms of what happens in a particular
"week" or planning period. But for our purposes, the important question is
whether the ratio of consumption to income can reasonably be expected on
a priori grounds to be independent of the level of income over successive
"weeks" or planning periods.Both Friedman and Duesenberry argue that
it is. The arguments of both these economists have been discussed above,
though Friedman, relying basically on the Fisher diagrams, comes to the
same conclusion as Duesenberry but for slightly different reasons.93 The
Friedman argument is clear and appealing in its explanation of the consumer
unit's behaviour under conditions of certainty. Under uncertainty, however,
the situation becomes confused. There may or may not be reasons for feeling
that, when uncertainty prevails, the savings proportion will increase as
income increases. There does however seem to be reason for explicitly
including, where possible, certain kinds of non-human wealth, e.g. liquid
assets in the formulation of the consumption hypothesis.
An examination of traditional consumption theory is rather disappointing
in that there are no clear indications as to what we can expect, on a prior
grounds, from the behaviour of the ratio of consumption or savings to income
over time. We have, however, been able to distinguish the theoretical from
the measured or observed variables, and we have been led to consider the
importance of factors other than income in determining the behaviour of
aggregate consumption.
Hicks, p.241.
Friedman, p.13.






1832 1930

Among the first early estimates of national income in Jamaica is that
of the late Professor Frederic Benham. With the idea that "The National
Income of a country is the net total of all income received by individuals
and institutions within that country....the net value of all production within
the country, plus net income received from abroad", 1 Benham estimated that
in the calendar year 1942 the national income of Jamaica was 33.3 m. The
net value of all goods and services produced in Jamaica was estimated to be
31.7 m. and the net revenue received from abro ad was 1.6 m. The term
"net" is used to denote the total value of goods and services produced, less
the value of imported materials, fuel, etc., less excise duties, less repairs,
replacements and renewals, plus services rendered by government, care being
exercised throughout to avoid the double-counting of local intermediate pro-
Several sectors in the Jamaican economy were distinguished. Table I shows
the net contributions of these sectors.



Net Value of Production

(a) Exports 4,148,000
(b) Foodstuffs for local consumption 10,488,000
(c) Manufacture etc. for local consumption 3,432,000
(d) Services rendered by Government 2,300,000
(e) Other services 7,630,000
(f) Rental value of all houses 3,800,000
(g) New construction 250,000
(h) Repairs, replacements, renewals 350,000

Total (Excluding (h)) 31,698,000

Net Income From Abroad 1,600,000

National Income 33, 298, 000

aBenham, Ibid. p.24.
Benham also estimated that, very roughly, the contributions of the Agri-
cultural, Manufacturing and Services sectors of the economy were as follows:

Agriculture, forestry and fishing


IF. Benham, "The National Income of Jamaica, 1942". See Development and Welfare in
the West Indies, Bulletin, No. 5. p.1, Barbados


The average income per heac of population in 1942 was estimated to be 26
per year or 10 shillings per week The distribution of income was found to be
highly uneven. Of the total national income figure of 33 m. "some 5 million
went to the few thousand individuals paying income tax; the rest of the popu-
lation received on the average, the equivalent of 35 shillings a week per
family of four". 2
The disposal of the National Income of Jamaica in 1942 is shown in
Table 2.

Food and drink 16,000,000
Hotels, boarding houses etc. 750,000
Tobacco 800,000
Clothing etc. 3,500,000
Shoe repairs 500,000
House rents 3,800,000
Household goods (fuel, soap, crockery) 1,000,000
Domestic service 1,350,000
Passenger transport 1,250,000
Direct taxation 1,200,000
All other 3,150,000

Total 33,300,000


Since these items have been valued at market price it appears that what
Benham meant by national income is what we now know as net national pro-
ductat market prrces.But this appearance is difficult to reconcile with the
previous deduction of excise duties from manufactures etc. which enabled
him to arrive at the identical estimate of national incomeof3 3. 3m.3t is true
that excise duties amount to only 587,000, but duties on all imports in 1942
amounted to 1.3 m.4 In order then to estimate the net output of factors of pro-
duction resident in Jamaica or net national product at factor cost we should
subtract these amounts from the national income estimateof 33.3 m. Most of
these deductions would be made from consumption expenditures rather than
from investment. This is done in Table 3.


Consumption expenditures 28,950,000
Direct taxation 1,200,000
All other 3,150,000
Less indirect taxes 2,000,000
Net National Product at Factor Cost 31,300,000

2Ibid. p. 27.
3Ibid. p. 12
4Ibid. p. 20.


Disposable income in 1942 was thus of the order of 30.1 m. and the
ratio of consumption expenditure to disposable income was about 0.97 in
Mr. Allan Morals attempted similar estimates for Jamaica for 1943. He
probably obtained an estimate of the national income somewhat better than
Benham's because he was able to draw on the information from the island-
wide Census of 1943, a body of data which was not available to Benham.
Morals employs three methods of calculating the national income: (1) the net-
value-of-output approach on whichBenham relied; (2) the net-income approach;
(3) the net-expenditure approach.

This multiple approach to national income accounting, as we shall see
later, makes it possible to detect inconsistencies in the estimates and gives
a clearer picture of the interdependence of the various sectors within the
Using the net-value-of-output approach Morals estimates the sectoral con-
tributions to be as shown in Table 4.


Primary production 12,700
Secondary production 9,700
Trans ort and distribution 11,700
Government services 4,200
Miscellaneous services 6,200
Housing, etc. 4,000

Gross value of production 48.500
Less value of deductions 9,000

Net value of production 39,500
Plus net income received from abroad 1,000

Net National Output 40,500

The amount of repairs, replacements and renewals which ought to be deducted
from net national output in order to make it comparable with the concept of
national income used by Benham is 394,000.
The second method of calculation involved the assessment of personal
incomes from rental of property, undistributed profits, government income and
household subsistence production. Morais estimates incomes from these
sources to be as shown in Table 5.

S-"National Income 1943-46", Jamaica, Table 11, p. 63 (a).



Wages, salaries and other
personal services 9,000
Rent 9,300
Profits and interest 14,000
Government income 4,700
Subsistence production 2,500
Income from abroad 1,000
Net National Output 40,500

Finally a calculation of income based on estimated expenditure was
attempted. This latter method provides some means of cross-checking the
previous estimates. Using a market price valuation Morals presents the esti-
mates shown in Table 6.


Personal consumption expenditures 35,200
Government consumption expenditures 4,522
Investment and savings 2,868
Remittances abroad to foreign
property holders 485
National Income at Market Prices 43,075

When valued at factor cost personal consumption expenditure is reduced to
32,625,000, and the National Income at factor cost checks with the previous
The national income estimate for 1943 yields a per caput figure of 33
per year or 12/6 per week. Direct taxes in 1943 amounted to 1,599,000 so
that disposable income from all sources was about 38.9 m. With total con-
sumption expenditure at 37.1 m. the ratio of consumption expenditure to
disposable income in 1943 stood at 0.95. There has apparently occurred a
slight drop in the "average propensity to consume" between 1942 and 1943,
which would seem to bear out the proposition that as the level of income
increases the proportion of income consumed declines. But we ought in fact
to look at what happens to the ratio of personal consumption expenditure
to personal disposable income before coming to such a conclusion. Neither
of the two National Income studies so far summarised gives any direct esti-
mate of personal disposable income. From Benham's estimate it would appear
that, in 1942 personal income in Jamaica was about 31.8 m.6 Net personal
income tax (including stamp duties) was 1.2 m., so personal disposable
income in 1942 was about 30.6 m. Since government expenditure on goods
Benham, p 26.


and services was in the region of 2.3 m.7 personal consumption expenditures
were evidently 29.0m. less 2.3m. i.e. 26.7m. In 1943, as we have just seen,
personal consumption expenditures (at factor cost) had climbed to 32.6 m. and
personal income less direct taxes stood at 33.7 m. (See Table 7.)

1942 1943
Personal consumption expenditure ... ... 26.7 32.6

Personal disposable income ... ... ... 30.6 33.7

"Average propensity to consume" ... ... 0.87 0.97

The ratio of personal consumption expenditures to personal disposable income
therefore apparently increased between 1942 and 1943. This conclusion, how-
ever, should only be tentatively accepted, as the figures themselves (apart
from differences in definition) suffer from numerous errors. Personal consump-
tion expenditure in 1942 as shown in the table is most likelyunder-estimated,
but from the figures given by Benham it is not possible to obtain a closer
estimate. The presumption however seems to be in favour of a much higher
figure for 1942 and an estimate of the average propensity to consume of
0.90 or more.

The Thorne Study

The next official attempt to measure the national income of Jamaica was
undertaken by A.P. Thorne who, during 1953, had been invited by the Univer-
sity College of the West Indies, acting on behalf of the Jamaican Government,
to make an economic accounts study for the island. The aim of the study was
mainly "...to indicate approximately what were the relative dimensions of
the major sectors of the economy in 1952, and to throw some light, as far as
available data would permit, on the inter-relationships of these sectors...."8
In fact Thorne was able to provide similar estimates for 1950 and 1951 as well,
though admittedly the figures for these years are probably less accurate.
As we have seen in the case of Morais' study it is possible to estimate
national income by three independent methods. This is, of course, recognized
by Thorne, but Thorne more readily admits the crudeness of any methodother
than the net-value-of-output approach for Jamaica. For in Jamaica "The
income approach is ...barred...for want of a sufficient coverage of the population
by income tax statistics, and through the lack of unemployment insurance and
employment exchanges and their statistics."9 Expenditure and income accounts
are therefore off-shoots of the basic production accounts.
Benham, p.17.
A.P. Thorne, "Size, Structure and Growth of the Economyof Jamaica", Supplement to
Vol. 4 No. 4of Social and Economic Studies p.1.
9Thorne p.74.
Thome, p.74.


Another important feature of the Thorne scheme is that clear distinctions
are made between the various concepts of output: gross and net, national and
domestic etc. The U.N. system of classification is adopted. A series of sector
accounts, which show clearly the inter-relationships within the economy, and
which form a distinct improvement on the Benham and Morals estimates, are
presented. These sector accounts, shortly to be described, refer to the year
Eleven sectors within the Jamaican economy are distinguished. They are
as follows:
(1) Main Export Industries
(2) Other Industries (Primary)
(3) Other Industries (Secondary)
(4) Public Utilities
(5) Distribution, Transportation and Services
(6) Central Government and Agencies
(7) Local Governments
(8) Persons and Private Non-Profit Institutions
(9) Construction and Installation
(10) Savings and Investments
(11) Rest of the World

The money-flows between these sectors are summarised in Table V (Appendix).
The entries in rows indicate "Payments To" the sectors at the head of each
column: the entries in columns indicate "Receipts From' the various other
sectors at the head of each row. The entries in the leading diagonal of this
square matrix are of course all zero since we are here dealing only with inter-
sectoral flows. Thorne gives a fairly detailed description of all the entries
in this table and consequently xve shall not repe at his description here.10
Suffice it to say, however, that personal savings of the Jamaican community
in 1952 are revealed to be extremely low (969,000), both in absolute terms
as well as in relation to gross savings.
The contribution by various sectors to gross domestic product at factor
cost are shown in Table 8.


Main export industries
Other industries (primary)
Other industries (manufacturing)
Public utilities
Distribution, transportation and services
General government (central and local)

10Thr-rn-, p.49 foil.

L m. %
12.6 13.3
17. 4 18. 3
7.5 7.9
0.9 0.9
40. 3 42.4
6.0 6.4
10.3 10.8
95.0 100. 0


About 32% of Jamaica's net output in 1952 therefore derived from Agriculture,
forestry and fishing, as compared with 40% from distribution etc. and a mere
8% from manufacturing industries. Gross domestic product at market prices
stood at 103,258,000, and this amount, according to Thorne was allocated in
the manner described in Table 9.


Subsistence consumption 8, 575
Monetary consumption expenditure 80,645
General government consumption expenditure 9, 013
Gross domestic fixed capital formation 13,695
Increase in stocks 849
Exports of goods and services 24,451
Less imports of goods and services -33, 970
Total 103,258

Net factor income payments from the rest of the world amounted to 557.000
in 1952, depreciation allowances were estimated to be 4,890,000, so net
national product at factor cost (national income) was 90,629,000 in 1952.
This figure is more than double the 1943 estimate of 40.5 m. From Table 9
it can be seen that total personal consumption expenditure at market prices
in Jamaica amounted to 89.2 m. in 1952, a figure which is almost ten times
the Government's estimated consumption expenditure. Net fixed capital for-
mation in the same year amounted to 8,805,000 most of which amount
represents undistributed profits of corporations resident in the island.
Personal disposable income (at market prices) can be calculated from the
entries in the "Persons and private non-profit institutions account". Of a
total receipts figure of 91,409,000, 1,220,000 were paid out in the form of
direct taxes leaving 90,189,000 to be allocated to personal consumption or
to personal savings. The ratio of personal consumption expenditures to personal
disposable income is therefore given by 892 = 0.989. This estimate is slightly
reduced to 0.988 when we make the calculations in factor cost values.
Receipts of persons and private non-profit institutions in 1950 and 1951
were estimated to be 69,380,000 and 78,660,000 respectively. Direct taxes
rose from a gross amount of 910,000 in 1950 to 1,050,000 in 1951. So dis-
posable personal income (at market price) was 68,470,000 in 1952 and
77,610,000 in 1951. The ratio of personal consumption expenditure to personal
disposable income (at market prices) was 0.989 in 1950 and 0.990 in 1951.
Indirect taxes, less subsidies, amounted to 7,173,000 in 1950 and 8,212,000
in 1951. So the ratio of personal consumption expenditure to personal disposable
income (at factor cost) was 60.6 = 0.989 in 1950, and 687 0990 in951.
61.3 69.4

See Account 4 in Th e, p.
See Account 4 in Thorne, p.68.


Estimates of National Income (1953 ), Department of Statistics,

Following Thorne's study, a national income section was set up within
the Department of Statistics, with the purpose of continuing the annual es-
timates of output in Jamaica. It is this section which has been responsible
for the estimates made since 1953. The collection of information became
rationalised and the -nethods of estimation employed by Thorne were adopted
with very minor modificatiors.12 No substantial revisions were made in
Thorne's original estimates for 1950-1952. Those revisions that were under-
takenwere "... directed towards the basic size and structure of asubsector..."
rather than towards changing the entire estimates on the basis of new evi-
dence.13 For example, it was decided to revise many of the sub-items in
Thorne's private consumption expenditure estimates for 1950-1952, (in
particular food and clothing) but this revision was carried out without any
change in the total private consumption expenditures for those years.14
The estimated gross domestic product at factor cost for the years 1953
to 1960 is shown in Table 10.


1953 1954 1955 1956 1957 1958 1959 1960

106.7 119. 7 136.4 158.5 191.9 198.7 208.0 230.2
Estimates of the national income and other components of gross output are shown in Table 11.
(Current values: millions of pounds sterling)

1953 1954 1955 1956 1957 1958 1959 1960

National income
Personal consumption
Gross domestic fixed
capital formation
Net domestic fixed
capital formation
Personal savings
Transfers to Government
and rest of the world
Disposable personal
income e

99.0 109. 9 125. 1 143. 2 171. 0

177. 3 188.6 204.9

97. 3 106.6 121.0 131. 3 145.8 155. 6 169. 2 183. 2

14. 2 17. 3 21.5 38.8 52. 1 47.4 46. 1 51.7

7.4 8.8 12. 2 28.2 38.0 32.4 30.3 34.0
( 4 0.7 0.6 4. 5 11.2 6. 3 3.6 5. 1

0.4 0.5 0.5 0.6 0.6 0.7 1.3 1.3

98. 1 107.8 122. 1 136.5 157.6 162.6 174. 1 189.6

Source: National Income Accounts, 1958, 1959, 1960, Jamaica.

1See "Preliminary Report on the National Income of Jamaica, 1953-55; with Support-
ing Tables". National Income Section, Department of Statistics, Kingston, Jamaica.
13Ibid p. 29.
14Ibid Table 5. Cf. also p. 16 of Thorne.


The calculated ratio of personal consumption expenditures to disposable
personal income over these years was therefore as follows:


1953 1954 1955 1956 1957 1958 1959 1960
0.992 0.989 0.991 0.962 0.925 0.957 0.972 0.966

Except for the year 1957, the ratio has remained high and close to unity
in spite of large increases in national and per caput income over the decade
of the 50's.
Pre-war and nineteenth century estimates of national income in
Jamaica reveal a similar sort of picture of aggregate consumer spending out
of disposable income.

Estimates of National Income (Eisner, Deane), Jamaica 1830 1938

Mrs. G. Eisner in a recent publication has presented a wealth of data on
economic change in Jamaica during the century 1830 to 1930.1s Her interest is
mainly in long-term trends, and so national income estimates at twenty-year
intervals only are calculated. In order to eliminate short-run fluctuations in
output deriving from the extreme variability in the value of exports over these
years, the estimates for the years 1850, 1870, 1890, 1910 and 1930 each re-
presentthree-year averages with the stated years falling in the middle. (The
estimate for the year 1850, for example, is an average of estimates for the
years 1849, 1850 and 1851). The estimate for 1832 is the only single-year
estimate representing output conditions in Jamaica in the period just prior to
complete emancipation. The estimate for 1850 is supposed to give a picture of
the Jamaican economy after the short-term adjustment to emancipation. As the
nineteenth century wore on a series of laws detrimental to sugar interests, as
well as changing social and economic conditions, served to push the economy
away from dependence on sugar production and towards the cultivation of
bananas.16 The estimates for 1890 and 1910 show the effects of this tran-
sition in its early and final stages. Poverty and mal-administration led to the
famous peasant rebellion in 1865. The local Assembly was compelled to sur-
render its constitution: the Governor (Eyre) was recalled and what was once
"representative" government became Crown Colony government. The estimate
for 1870 represents the impact of these political changes on output.
Eisner's method of calculating national income, like Thorne's, relies
openly on the net-value-of-output approach.The income and expenditure tables
do not provide altogether independent checks on the output estimates since
they are themselves derived from those estimates. "Furthermore even in the
case of output it is not possible to state any margins of error".17 However, as
official publications were used throughout in preparing the estimates it is not

15G. Eisner, Jamaica 1830 1930. Manchester, 1961.
Eisner, Ch. 11.
Eisner, p.23.


likely that any major item produced in Jamaica was omitted. For the same
reason, goods and services entering external trade as well as the public
utilities and transport and communications sectors "... are fairly reliable;
their accuracy generally improving after the inception of Crown Colony govern-
ment in 1866".18 The main source of inaccuracy in the estimates is likely to
lie in the estimation of the innumerable items of peasant production, which
items do not appear in official publications but which together comprise the
fairly substantial subsistence sector. In the absence of much evidence on
this score, Eisner has probably devised the best possible method of estimating
this kind of output in Jamaica.19
A summary of the Eisner estimates of output is shown in Table VI(App.).
The ratio of total personal consumption expenditure to gross domestic
product20 for each of these "years" is shown below.

1832 1850 1870 1890 1910 1930

STotal personal expenditure 0.67 0.90 0.91 0.91 0.88 0.92
Total income (G.D.P.)

The ratio of consumption to gross output in Jamaica evidently remained high
and fairly stable over the century following emancipation.
The derivation of disposable personal income from G.D.P. involves the
following set of calculations:
Gross domestic product at market prices
(1) Plus Net factor incomes from abroad
Equals Gross national product at market prices
(2) Less Capital consumption allowances
(3) Less Indirect taxes
(4) Plus Subsidies
Equals National income (or net national product at factor cost)
(5) Less Undistributed corporate profits
(6) Less Corporate profits tax
(7) Less Government income from property and enterprises
(8) Plus Net government transfers to persons on current account
(9) Plus Net remittances from rest of the world to persons
Equals Personal income
(10)Less Personal income tax
Equals Disposable personal income21

Eisner, p.24.
Eisner, p.9-12.
2Eisner, Table 8. V.
21National Accounts, 1958, 1959, Dpmt. of Statistics, Jamaica, p.44.


The Eisner estimates of items (2) and (5) are extremely weak.22 Nor are
her estimates of the other items any but rough indications of the orders of
magnitude. We can, however, obtain some idea of personal income in Jamaica
during these years by consulting her summary income table reproduced
in Table 12.


Items 1832 1850 1890 1930

Income from agriculture and manu-
factured exports (excluding
planters' profits)

Other wages and salaries

Other incomes

Grand Total

Direct taxes

Disposable personal income

Personal consumption expenditure
aEisner, Table 8. IV.

1,953.3 1,869.6 3,948.4 8,806.9


673.6 1,708.9 4,788.5

1,120.4 976.8 2,406.8 5,024.5

3,738.7 3,520.0 8,064.1 18,619.9

3,738.7 3,520.0 8,064.1 18,521.7

3,350.2 3,348.3 7,945.1 18,322.5

The ratio of personal consumption expenditure to
income is therefore as follows:

disposable personal


1832 1850 1890 1930

0.92 0.94 0.98 0.99

In spite of vast increases in the aggregate and per caput level of income
between 1832 and 1930 the average propensity to consume remained over 0.90
and in fact showed some tendency to increase rather than to decline between
the immediate pre-emancipation years and 1890.
Phyllis Deane's study of national income in Jamaica in the years 1929-
1938 tells a similar story. Defining "taxable income" as the income generated
within a country inclus iv of income earned abroad by national factors of
production, and "national income" as income earned by all factors of pro-
duction exclusive of income accruing to foreign factors operating in
Jamaica,23 Miss Deane presents the estimates shown in Table 13.
Eisner, PP. 14-19.
23Phyllis Deane, The Measurement of Colonial National Incomes. Cambridge,
1948, p. 104 and pp. 14-19.



(1) (2) (3) (4) (5) (6) (7) (8)

Net Tax-
Year able National
Income Income

Personal Net Do-
Ccnsump- mestic
tion Invest-
Ex pendi- ment

19,187 18,735 16, 135 715 1,885 16,850 0.95 0.64

19,053 18,590 1I,886

18,339 17,901 15,201

17, 177 16,741 14,097

16,391 16,391 13,507

16,553 16,183 13,474

17,147 16,660 13,916

17,474 16,940 14 141

19,420 18,680 15,680

20,319 19,597 16,461

726 1,978 16,612 0.95

675 2,025 15,876 0.96

644 2,000 14,741 0.96 0.82

560 2,000 14,391 0.94 0.82

619 2,090 14,093 0.96 0.82

699 2,045 14,615 0.95 0.81

729 2,070 14,870 0.95 0.81

910 2,090 16,50D 0.95 0.81

865 2,271 17,326 0.95

Source: Deane, p. 137-141.

The ratio of personal consumption expenditure to net taxable income
(Column (8) ) did decline somewhat between 1929-1938, according to Deane,
but it is interesting to note that the ratio of personal consumption expenditure
to disposable personal income (estimated here as the sum of personal con-
sumption expenditure at factor cost plus net domestic investment) remains
very stable over the period at about 0.95



The various estimates of national income which we have briefly reviewed
admittedly suffer from several types of errors. One source of error arises
from incomplete coverage. This type of error affects in particular the estimates
for the years prior to 1943, in which year the most comprehensive Census of
Jai.iaica to date was undertaken. But apart from the lack of adequate overall





me nt



(3)+(4) (3)-(6)


surveys and source material, there is, in addition, error deriving from
the varied methods of estimating value added in the subsistence or non-
monetary exchange sector of the Jamaican economy. In the final analysis,
estimates of output added by this sector have had to rely on the subjective
valuations of the estimators. This feature of national income estimation in
Jamaica, as well as in most underdeveloped countries, affects, in particular,
the expenditure accounts of Eisner, Benham and Deane. 24 In so far as our
present task is concerned, however, the whole question of errors is important
only to the extent that it affects significantly the various estimates of the
average propensity to consume in Jamaica. We therefore wish to see if there
are any differences between the average consumption ratios in terms of monetary
expenditure alone, as against the ratios in terms of monetary and non-monetary
expenditures as variously estimated in Jamaica. These sets of calculations
are shown in Table VIII (App.). The basic information which enables these
calculations to be made is summarised in Table VII.
It is clear that whether we consider national income or personal dispos-
able income as the denominator, or whether we include or exclude subsistence
output, the ratio of consumption to income in Jamaica has shown no secular
tendencyto increase ordecrease over theperiod 1850 to 1960, and this in spite
of an increase in average annual(current) income per head of more than twelve-
fold over the period. It is this observed stability in the long-run "average pro-
pensity to consume" which led us to consider hypotheses alternative to the
well-known Keynesian hypothesis of the consumption function.
There are, however, a series of statistical problems which are sometimes
overlooked in discussion of theKeynesian hypothesis. 25 Keynes clearly stated
that his hypothesis referred to "real" or deflated rather than current money
values of consumption and income. 26 Individuals, in effect, do not suffer from
a money illusion but make consumption decisions on what they consider to be
their real-income position. Monetary consumption expenditure is thus a function
not only of money income but of the price level as well. This being so it is
quite possible that the relationship between current money values of consump-
tion and income may well fail to reveal the underlyingnon-proportional relation-
ship in "real" terms.
In most of the national income studies in Jamaica discussed above, data
on price changes have been presented. As usual the data suffer from many
imperfections, chief among which is perhaps inadequacy of coverage. The
price series, nevertheless, give a general picture of inflationary and deflation-
ary tendencies over these years. The deflated values of consumption ex-
penditure and income are shown in Table IX (App.). Again there seems to be
no tendency for the ratios of "real" consumption to "ieal" income to decline
consistently from year to year, although the ratios are on the whole lower
when expressed in "real" terms than in current money values. In Chart 2
we have fitted linear equations of the form
C = a+ bY
to the estimates of "real" personal consumption expenditures (C) and "real"
disposable personal income (Y) both including non-money expenditures. The
estimated equation for the period 1832- 1938 is
C = .840 + 0.838 Y (R2= 0.995)
25 Eisner, pp.9 12; Deane p. 108, Benham, op. ci t.
26G. Ackley, Macroeconomic Theory, p. 227, foll.
General Theory ... p.90.


For the years 1950-1958 the estimated equation is

C = -4.782 + 0.992 Y (R2 = 0.991)
(Both C and Y are here measured in m.
Judging from the high positive figure of R2,it appears that income is
a highly important variable in determining consumption in Jamaica. Further
there seems to have been some increase in the marginal propensity to consume
after the war years. The value of the correlation coefficient remains high,
even when undeflatec values of consumption and income are related. For
the period 1832-1938, a linear equation fitted to the undeflated values
(inclusive of subsistence) yields

C = -0.132 + 0.966 Y (R2 = 0.997)

and for the period 1950-1960 we have

C = 4.339 + 0.935 Y (R2 = 0.941)
The marginal propensity to consume is higher in current than in real values
for the pre-second World War period but lower for the post-war years. In the
pre-war years almost all increases in disposable personal income went to-
wards improvements in levels of living via consumption spending.
In real terms the average propensity to consume in Jamaica increased
from 0.918 for the period 1832-1938 as a whole to 0.949 for the period 1950--
1958. The elasticity of consumption with respect to income, as measured, is
therefore less than one for the pre-war years and greater than one for the
post-war period. Further, Jamaican consumers seemed to have suffered more
from a money illusion in the post-war years than in the period 1832-1938.
The relationship between current valued estimates of personal consump-
tion expenditures and disposable personal income for the entire period 1832-
1960 is shown in Chart 3. The scatter of points seem to fit closely a line
drawn through the origin with a slope of 0.97.
It is clear that theproportionalityof the relationship between consumption
and income is not impaired by allowing for changes in population. The slope
of the curve and the scatter of the points remain very much the same, although
the units are different.
Some considerable caution is, however, needed in interpreting consump-
tion-income relationships derived from a system of national accounts.27 The
first point to recall (and this holds with particular force in the Jamaican
estimates of national income) is that the estimates of expenditure out of the
national income or gross national product are not wholly independent estimates
but basically derived from the estimates of output. The budget studies which
have been made in Jamaica since 1939 have not, as a rule, been used to
provide estimates of annual consumption expenditure in the national accounts.
This might not however be a very serious problem, since the indications are
that the average propensity to consume of the typical Jamaican household is
itself high and has remained high over the past thirty years (see later dis-
cussion). But it is precisely this feature which raises some serious problems
as regards the direction of causation in the consumption-income relationship.
According to the Keynesian hypothesis, consumption is high or low
because income is high or low. [t is however easy to see, from the logic of
the national accounts, that the size of, and changes in, the national income
27 ey, it 233-
G. Ackley, op. CIt, pp.233-236


are themselves dependent on the size of, and changes in, aggregate consump-
tion. The relationship between consumption and income therefore works in
two opposite directions. A high observed correlation between consumption
and income can thus, in this case, be interpreted as representing simply a
high correlation between totals and their own largest components.28 The
observed high correlation coefficients do not therefore indicate a unique
dependence of consumption on income as against the reverse relationship.
How then can we demonstrate the existence of a consumption function?
Haavelmo argues2 9 that in estimating the consumption function we should
bear in mind the general economic model, of which the consumption function
is only a part. If, as is usually the case, our model says that investment
determines income, which in turn determines consumption, then the ultimate
variable on which consumption depends is not income but investment. Es-
timates of the functional relationship between consumption and income will
be biased if no attention is paid to the dependence of income on the exogenous
variable-investment. Ackleymakes the samepoint but in a different way. The
basic national income identity is
(1) Y = C + S
where Y is incomeC is consumption and Sis savings. The normal Keynesian
consumption function is given by
(2) C = a +.bY
Because of the identity (1), it is clear that Y will also be dependent on C.
Both relationships operate simultaneously and in opposite directions. Our
problem is to isolate from the data the dependence of C on Y, i.e. the con-
sumption function.
If the only relationship which exists between C and Y were that in which
C determines Y, then the linear form of the dependence is given by

(3) Y = c+ dC
and the variation in Y which is not explained by C is completely independent
of C, by assumption. From the identity we see that this unexplained variation
is due to S Hence, where C determines Y we should expect to find no correla-
tion between C and S or I(Investment). But if there exists in addition a
functional dependence of C on Y, then S will be positively correlated with C,
and this correlation will be perfect if the dependence of consumption on income
is perfect.30 The dependence of consumption on savings (or investment),
therefore, turns out to be a test of the existence of the consumption function.
It is easy to see what the parameters in the consumption-savings relation-
ship will be. Substituting eq n (1) above into eq n (2) we get
C= a+ b (C + S)

i.e. C (1- b)= a+ bS

C= a + b S.
1- b 1- b
G. Ackley p.233.
2T. Haavelmo "Methods of Measuring the Marginal Propensity to Consume". Journal of
the American Statistical Association, March. 1947, pp. 105-122.
G. Ackley, p.234.


Thus if we find the regression of C on S, the intercept on the C axis will
be a and the slope of the regression line will be b where ais the
F-b ~b
constant in the consumption-income relationship and b is the marginal propen-
sityto consume. We thus have here an indirect and probably more proper method
of estimating the parameters of the consumption function. Chart IV shows the
relationship between "real" personal savings and "real" personal consumption
expenditures in Jamaica for the years 1950-1958. It is not possible to derive
any sure information on personal savings from earliernational income estimates.
But the movements of personal consumption expenditures and net domestic in-
vestment for the years 1832-1938 (See Table VII (App.)) certainly suggest some
positive correlation.
The estimate of personal savings for 1950-1958 at constant prices is
based on the movement of "real" gross domestic investment as estimated by
Bethel.31 A straight line fitted by least squares to the scatter of points yields
the following equation:
C = 89.982 + 6.267 S(R2 = 0.590)
We therefore have,
b = 6.267
1- b
b(1 + 6.267) = 6.267
.*. b =0.862

This estimate of themarginal propensity to consume is less than our direct
estimate of 0.992, but it is still greater than the estimated marginal propensity
to consume for the period 1832-1938. Further, since,
f1"- 0.62 = 89.982
we have a= 12.418, which estimate is again different from our direct estimate
of 4.782.
The observed correlation coefficient (r= 0.768), although low comparedto
the correlation coefficients in our directly estimated consumption functions, is
significantly different from zero. We are thus inclined to feel that there exists
a sufficiently strongrelationship between consumption and income to enables
to speak of the existence of the consumption function in Jamaica. We may at
the same timenote that the "true" marginal propensity to consume in the post-
waryears was still higher than the estimatedmarginal propensity to consume for
the pre-war period.

JAMAICA, 1939 1958

Cross-sectional evidence on the consumption function in Jamaica has
hardly been explored. This lack has stemmed largely from the customary
incredulity attaching to the income information in budgetary surveys. Further,
budgetary information in Jamaica has, until recently, been confined mainly to

31J. Bethel, "Some National Income Aggregates ForJamaica At Constant prices", Social
and Economic Studies, Vol. 10, No. 2. June 1961, pp. 128- 155.


a rather restricted sample of households (usually from the "working-class" in
the Kingston Metropolitan Area). The results of these surveys could hardly
be deemed to reflect the pattern of consumption behaviour throughout the
But it would be wrong to dismiss entirely such survey information as is
available. These budget surveys were, after all, designed for the purpose of
calculating weights for the Retail Price Indices. If they do not reflect the
consumption patterns of all social and economic classes in Jamaica, they at
least represent, or were designed to represent, the consumption pattern of
the majority of Jamaican income-earners, viz. "working-class" people. They
can therefore provide us with information on the effects of differences in
income on consumption among this group of consumers.
The effect of differences in income on consumption is, of course, not the
same thing as the effect of changes in income on consumption, or the con-
sumption function proper. In order, then, to apply the concepts of marginal
and average propensities to consume to budget data the assumption has to
be made that information obtained from observing consumers in different
circumstances at the same moment in time is relevant in forecasting the
behaviour of any particular consumer or group of consumers as his or their
circumstances change through time. This assumption is, of course, likely
to be falsified for a number of reasons; but some such assumption has to
lie at the back of any analysis of cross-sectional data.
In analysing budgetary evidence on the consumption function we are
interested in seeing whether average and marginal propensities to consume
vary as between different surveys collected at different moments of time. We
are also interested in explaining the difference in average propensities which
we find between households interviewed at the same time.
In Jamaica, as previously indicated, there have been, since 1939, four
surveys of household expenditure. The last of these surveys taken in 1958
(which we shall be discussing in some detail later on) is the first survey in
which households in all parts of the island and at all levels of income were
interviewed "simultaneously".32 As the object of making these surveys was
mainly to obtain an average expenditure pattern for the survey group as a
whole, the method of presentation of the information in the various reports
has not been such as to enable us to calculate, in the way we would like,
average or marginal propensities or income elasticities.Any thorough analysis
of the survey data is therefore limited unless we consult the original schedules
from which the surveys are derived. Cumper has done this for the 1953-54
Kingston Metropolitan survey and the present writer has helped in the arrange-
ment of the 1958 All-Island budgetary survey.
The broad averages and frequency tables presented in the budget reports
are however not entirely useless.
The first official Cost of Living Enquiry in Jamaica (sponsored by the
newly formed Labour Department) was begun on August 7th, 1939 and lasted
until October 6th of the same year. The purposes of the enquiry were rather
restricted.33 In all, the schedules of 486 households were assessed. These
households were selected "...with reference (as far as known) to the dis-
tribution of the population..." from the Kingston area and its central districts
only, and each household interviewed earned not more than 2 per week.34
32seeHousehold Expenditure Survey, Janaica, 1958, Preface.
33See The Cost of Living Survey 1939, Janaica, pp. 1-3.
34bid. p. 2.


The weekly receipts and expenditures, as well as the frequency of employment
of adult members of the household were recorded. Table 14 shows averages
for the survey group as a whole.

Weekly Cash Receipts per Weekly Cash Expenditure per
Household Household
(Shillings) (Shillings)
Earnings 9.2 Food 8. 1
Allowances 1. 6 Fuel and cleaning 1. 2
Rent owing to household 0. 6 Rent, rates and taxes, etc. 2. 8
Clothing 1.7
All other 3.7
Total 11. 4 Total 17.5

The fairly substantial excess of weekly cash expenditure over weekly
cash receipts for the average household, as well as for many of the individual
households, is, according to the report, due to incomplete reporting of income;
an upward bias in expenditure deriving from credit, arrears in rent, etc.; and
the assessment of clothing expenditure as a weekly expense "...when in fact
it may not have been incurred ir the week under review".3s But this is not
all; no attempt was made to value the food gifts received by 45% of the house-
holds, nor the items of clothing received free by some 10% of the households
assessed. Further, no assessment of "yard"-produce received by 35% of the
households was made. The rate of unemployment among the households inter-
viewed was, however,high and this also helps to explain the excess of cash
expenditure over cash receipts.
Of the adults in the sample ("adults" being defined as those 16 and over)
18% were in regular employment as clerks, messengers, mechanics, bakers,
truck drivers, etc.; 24% worked on their own account as cabinet makers, shoe
makers, butchers, etc.; 10% were intermittently employed and 16% of those
who would normally be either regularly or intermittently employed were un-
employed during the week of interview. The remaining "adults" (some 32% of
the total number of adults) had no gainful occupation.
It is not possible from the figures presented in the report to obtain any
idea of the effect of differences in income levels on the ratio of average
household expenditure to average household receipts. The report merely shows
that the ratio of weekly cash expenditure to weekly cash income among
Kingston "working-class" households in 1939 was somewhere in the region
of 1.5/1.
In 1953, Mr. J.R. Goodman, United Nations statistical consultant, was
assigned to the Jamaican Central Bureau of Statistics his task being that
of designing a survey of Kingston households such as would provide new
weights for the urban Cost of Living Index. The methods used in making this
survey have been briefly discussed in Section 1. Again the aim of the survey
was "... to obtain a sample of households which would be representative of
working-class people in the Kingston Metropolitan Area".36 This over-

35Cost of Living Survey, p.2.
36Household Expenditure Survey 1953/54. Kingston, Jamaica, p.4.


riding objective led to the inclusion of all workers, unskilled, semi-skilled
and "white collar", earning less than 300 per annum. The total number of
households selected for interviewing was 1,500 (approximately 21/2% of all
households in the Kingston Metropolitan Area). Detailed figures of average
weekly expenditure on a number of items were, of course, calculated and a
summary table showing the change in expenditure pattern between 1939 and
1953/54 was compiled. (Some adjustments in the 1939 figures had to be made
in order to make the figures adhere to the definitions used in the 1953-54
survey (See Table 15).


Items 1939 1953/54

(%) (%)
Food 52.63 52.09

Fuel and cleaning 7.89 7.59

Clothing 10.52 6.96
Rent 18.42 12.29
All other 10.54 21.16

Total 100.00 100.00

The report contains no information on the demographic, occupational or
income characteristics of the households, but Cumper in a hand tabulation
of 1,180 of the original schedules was able to extract a considerable amount
of information on these matters.37
The average size of the households in Cumper's calculations is 3.47
which is almost indentical with the 1939 figure. There is also a strong
tendency for family size to increase with income38 (See Table 16).
The ratio of adults to children and males to females tends likewise to
increase as the average level of household income increases. These latter
correlations are clearly what one would expect, as the earning capacity of
the household is dependent on its age and sex composition. Most of the
female members of "working-class" households in Kingston are occupied as
domestics, higglers or subordinate office staff. Table 17 shows the percentage
distribution of households by occupational groups. (We may note that the
number of domestics and labourers shows a strong tendency to diminish as
the level of household income increases).

G.E. Cumper, "Expenditure Patterns, Kingston, Jamaica, 1954", Social and Economic
Studies, Vol. 7, No.2, June 1958, pp. 165-177.
Owing to the irregularityof weekly(reported) income inJamaica, Cumper estimates house-
hold "income" as the sum of recorded expenditure on all items (except household equipment and
clothing) plus net "savings". See Cumper, op. cit., p. 166.



Income No. of Average No. of Ratio of Adults Ratio of Males
Group House- Persons per (14 and over) to to Females
( per wk.) holds Household Children (0-14)

0- 346 2.47 1.05 C. 68

2- 349 3.19 1.92 0.71

4- 180 3.76 1.93 0.81

6- 90 3.88 2.26 0.80

8- 37 4.05 2.26 0.81

10+ 78 5.33 2.22 0.86

Total 1,080 3.47 2.10 0.76


Income Skilled and Domestics and
Group "White-Collar" Semi-Skilled Labourers
(Z per wk.)

aThis table is based on a "systematically" chosen sub-sample of 256 of the 1,180
schedules (See Cumper, op. cit.)

The problems surrounding the estimation of household income have been
mentioned above. Cumper, in an attempt to side-step these difficulties,
favours the use of total expenditure rather than reported income as the "true"
measure of household income. The difference which this makes to the income
distribution of the households in the sample is revealed in Table 18 as well
as by Chart 5.



Group Reported Income Assumed Income
( per wk.) (Cumper)

% %

0- 36.0 29.3
2- 26.8 29.6
4- 12.6 15.2
6- 8.2 7.6
8- 4.4 3.1
10+ 11.9 15.1

Total 100 100

In constructing Chart 5 we have assumed that most incomes in a given
range will cluster about the mid-point of the range (we have not plotted the
points for households earning above 10 per week).
In order to see how close is the agreement between the income dis-
tribution of the sample and the income distribution of all households in the
Kingston Metropolitan Area, we have also shown in Chart 5 the distribution
of income among wage-earning heads of households (see also Table 19).


Wage-Earning Heads of Households

Income Group
( per wk.)

Under 1
1 -
2 -
3 -
4 -
5 -
6 -
10 -
14 -
20 -
Not stated

11, 164


P ercentages
17. 4
18. 3

a "Report on a Sample Surveyof the Population of Jamaica. Oct./Nov. 1953" Kingston, 1957.
Table 61, p.105.


A few points must be noted. First, the numbers in Table 19 refer to wage-
earning heads of households oily. The total number of income-earning heads
of households in Kingston in 1953 was, according to the survey, distributedin
the way shown in Table 20.

All statuses 85,818
Employer 1,240
Own-aocount 18,473
Wage-earner 45,895
Unpaid family workers 44
Apprentice 177
Other 18,960
Not stated 399

aOp. cit, p.iii. Table 67.

Wage-earning households therefore accounted for the majority of all
households in Kingston and are most likely to have a major effect on the
average income and income distribution of all households in Kingston. We
may note further, however, that the income of subsidiary wage-earners is
not included in the income-grouping. The importance of this source of income
is, as we saw above, likely to depend on the size and structure of the urban
household; as yet, nothing firm can be said on this score. We mention these
points here in order to clarify the meaning of the comparison between figures
drawn from the budget surveys and figures from the population census. Both
sets of data are not identical in terms of what they purport to show. But we
do expect them to show at least some vague similarity.
From Chart 5 we can see that the distribution of households by reported
income is broadly similar in bott the budget and census surveys, though the
budget distribution seems to have given somewhat less weight to low-income
households and somewhat more importance to households in the upper-income
groups. In the middle-income range, i.e. between 3 and 7 per week, there
is fairly close correspondence between the two distributions. The distribution
of households by "assumed" income, however, tends to be more even than
either of the two distributions by "reported" income. Cumper does not give
the average "reported" income Ver household for each income group, but it
is possible to calculate from his tables the average and marginal propensities
to consume out of 'assumed" income by income groups. It is clear that the
overall average propensity so calculated will be lower than that which would
be obtained by using "reported" income, owing to the differences in income
distribution just mentioned. Further, this calculated propensity will, for the
same reasons, be lower than the average propensity found in the 1939 survey.
The calculations are shown in Table 21.
The overall average propensity to consume of 0.96 agrees fairly well
with the average propensity to consume calculated from the national income
data for the years 1950--1958. For the year 1953, in particular, the ratio of
personal consumption expenditure to personal disposable income stands at
0.992 in current prices and 0.912 at constant prices Had the income dis-



Total Assumed
Income % Expenditure a Income b Average
Propensity to
Group of Houeholds (shillings per (shillings per Consume
(F per wk.) week) week)

0 29. 3 26.7 21.9 1. 22
2 29.6 58. 1 57.2 1.02
4 15.3 96.0 97.3 0.99
6 7.6 131.5 135. 3 0.97
8 3. 1 173. 5 181. 3 0.96
10 4.6 229. 4 240.9 0.95
15 3. 1 344.3 359. 5 0.96
20 2. 5 408.3 453.0 0.94
25+ 4.8 832.0 909.8 0.91

Total expenditure = sum of all items of expenditure less expenditure on clothing and
durables, less credit, debts.
"Assumed" income = sum of items 1, 11, 111, IV, Vl, V111 and 1X in Table 4.,p.
168, Cumper. Op. cit.
tribution by "'reported" income been used to weight each group propensity
we would have arrived at an overall average propensity to consume of 1.07.
This estimate is, as anticipated, higher than the previous estimate and
closer to the estimate derived from the 1939 Cost of Living Enquiry. It is,
however, hard to conceive of a group of households maintaining such a high
propensity over long periods of time.
Of the total number of households in the 1953/54 Kingston sample, only
60% had an average propensity to consume out of "assumed" income of more
than 1.00. The result of plotting consumption against "assumed" income is
shown in Chart 6. The actual scatter of observations deviates but slightly
from the break-even line, and only among households in the higher income
groups. A straight line fitted by least squares (weighted) regression to the
scatter of points, has the following form:

C = 7.5 + 0.90 Y (R2 = 0.99)

The marginal propensity to consume is thus less than the average pro-
pensity: the income elasticity of consumption is less than one. Income and
consumption are equal at an average household income level of 75 shillings
per week. The correlation coefficient is again very high, but as in the case
of the national income data we should be on guard here against possible
bias. It must be recalled that our estimate of income is based on total house-
hold expenditure itself, and as we have argued above, in such cases it would
be better to correlate total expenditure with "savings" i.e. estimated income
less total expenditure. This is done in Chart 7. In the chart, consumption
expenditure is plotted against both recorded weekly net savings and against
net savings out of "assumed" income. Both types of savings are clearly
positively correlated with average household expenditure and fairly regularly
distributed about a straight line described by

C = 72.0 + 9.505 S (R2 = 0.99)


Our indirect estimates of the parameters in the consumption income
relationship are thus:

b = 0.90
and a = 7.2

These estimates agree closely with the direct estimates, and the size
of the correlation coefficientin the consumption savings relationship suggests
that the consumption-income relationship is distinguishable in the budget
data despite the use of the expenditure method of estimating household
The present writer and two statistical assistants organised the original
schedules of the 1958 Household Budget Survey in a form which makes a
similar analysis possible.
Tables X XII (App.) summarised the total inflow and outflow of cash
and credit per household for different income groups in the three sampling
areas. Like Cumper we adopted the expenditure-plus-savings approach in
classifying the data. Unlike Cumper, however, our total expenditure classi-
fication includes purchases of clothing and durables and includes as well
investment or business expenditure undertaken by the household in its role
as producer of goods and services for the market. It is extremely important to
bear in mind this latter point, especially when'reviewing the tables for the
Rural and Main Towns samples. For in these areas, as we have previously
emphasised farming and other "business" households predominate. Average
weekly expenditure strictly on household account will therefore tend to comprise
a smaller proportion of total expenditure undertaken by the household, the
result being that in some cases, especially among the upper income groups,
average "Total Household Expenditure" per week sometimes falls below
the lower bound of the income range.
The term "Total Current Outlay" is defined as the sum of average weekly
total household expenditure for current living, personal insurance payments
and positive net increases in savings or decreases in debts. These items
together represent the total money allocations of the household on current
account. These money allocations have been financed out of weekly income,
credits and/or dissavings. The sum of these latter items we have called
"Total Cash Supply".
If all items of cash receipts and disbursements were accurately reported,
and if households indulged only in consuming goods and services and in
saving, there should be no difference between the two totals. In fact as the
tables show there are substantial differences between them. (The "Balancing
Difference" in these tables shows the excess of Total Cash Supply over
Total Current Outlay). In the Kingston Metropolitan sample, the Balancing
Difference is relatively small and positive among households earning an
average weekly income of below 6 per week, but is a substantial negatively
increasing amount among households earning above 6 per week. In the Main
Towns sample, with the exception of households earning over 15 per week,
the Balancing Difference is positive and relatively small. In the rural sample
it is positive throughout, increases with income, and is more significant
relative to Total Current Outlay than in either the Kingston or Main Towns


The inadequate reporting of household indebtedness in the Kingston
sample probably explains the major proportion of that sample's Balancing
Difference. Respondents, especially among the upper income groups in Kingston,
are loath to reveal the degree of their dependence on hire-purchase facilities
or on grocery credit. Part of the urban consumer's ethic is that the extent of
his indebtedness should at all costs be kept a secret. On the other hand a
considerable proportion of the positive Balancing Differences in the Rural
sample would more properly appear as expenditure undertaken by the household
in its role as producer. Owing further to the fact that average weekly income
as recorded is greater than Total Current Outlay for all income groups in this
sample, a large portion of such investment or business account expenditures
seems to have been financed largely from current and past savings rather than
from borrowing. This is especially true of households earning above 8 per
The absolute and relative magnitudes of the Balancing Difference are
smallest in the Main Towns sample. As most of the discrepancies show Total
Cash Supply greater than Total Current Outlay, the evidence points to the
importance of business expenditures among these households as well. But
since the values of the Balancing Difference are so small the consumption
role of these households appears to be overwhelming and this feature makes
the Main Towns households more akin to the Kingston households.
We are thus presented with several alternative ways of looking at the
consumption function, each depending upon what we define "income" to be.
If we are interested in the relationship between consumption and income over
a short term, perhaps not exceeding a year, then it would appear to be more
appropriate to relate weekly household expenditure to weekly recorded income.
If on the other hand the propensity to consume over a longer time-period is
the main interest, then we ought to use either Total Current Outlay or Total
Cash Supply as our income variable, since either of these totals tends to
reflect more adequately the household's totalcommand over financial resources
for a longer time-period than is implied by the notion of average weekly
recorded income. Our problem here is in effect similar to that faced by Fried-
man (cf. discussion in Section I). We are ultimately interested in that part of
the household's receipts which, by virtue either of its regularityor assured-
ness, may be considered as "permanent".
It will, of course, be recognized that these various ways of looking at
the propensity to consume are intimately related. The connection derives not
only from the accounting method but from a priori reasoning as well. The
major source of Total Cash Supply is weekly income. The ability to command
credit from the grocer, the hire-purchase companies or the agricultural loan
banks also depends on weekly earnings. Where these earnings are stable, we
should also expect to find stability in cash supply and in current outlay.
Further than that,we should expect both short-term and long-term consumption
functions to be stable.
In order to see how well these expectations are fulfilled in the present
data, we have estimated by least-squares three types of consumption functions
for each sample. The forms of equations used are as follows:
1) C = a+ bY ; i= 1, 2, 3
2) LogC = a + b log Yi ; i= 1, 2, 3.
Where C refers to average weekly total household expenditure for current living
plus personal insurance payments,


'Y1 = Average Weekly Recorded Income.
1'2 = Total Current Outlay.
Y3 = Total Cash Supply.
We have also calculated the regressions of each of Total Current Outlay and
Total Cash Supply on recorded weekly earnings as well as the regression of
Total Current Outlay on Total Cash Supply. The results of all our calculations
are shown in Table X[II. The data from which these calculations were derived
exclude all households earning over 15 in each sample. The comparisons
between sampling areas are therefore free from any bias which could derive
from dissimilarity in income ranges.
The values of RP' in both the linear and log-linear consumption functions
are everywhere high. The values of the regression coefficients are, with one
exception in the rural sample, significant in the statistical sense. The re-
gression constants, however, especially in the log-linear equations, are not
all significant.
The marginal propensity to consume out of recorded income (as estimated
by the regression coefficient, b, in the equation C = a + b Y1) diminishes
rapidly as we move from the urban to the rural areas. The elasticity of con-
sumption with respect to recorded income (as estimated by the regression
coefficient b' in C = a' + b' Y') also diminishes as we move away from Kingston
to the Main Town and rural areas. The situation tends to be the reverse for
the marginal propensity to consume out of Total Current Outlay and its
corresponding consumption elasticity. Estimates of both these magnitudes are
higher in the rural than in the urban areas.
Since recorded income and Total Cash Supply are intimately related, it
is not surprising to find that consumption exhibits similar behaviour with
regard to both of these variables. But while the marginal propensity to consume
out of Total Cash Supply tends to be greater than the marginal propensity
to consume out of recorded income in Kingston and the Main Towns, in the
rural areas it is significantly less. A similar pattern of behaviour is displayed
by the corresponding consumption elasticities, except that the differences
tend to be less significant.
The lower half of Table XIII shows the estimated relationships between
the three income variables. The amount by which Total Outlay increases in
response to given increase in weekly earnings is greatest in the Kingston
sample and decreases monotonically as we move to the Main Towns and rural
areas. The relationship between Total Current Outlay and Total Cash Supply
behaves similarly. The increase in the Total Cash Supply, relative to given
increases in weekly earnings, is on the other hand least in Kingston and rises
constantly as we move out to the rural areas. The elasticities corresponding to
these various propensities are subject to a relatively high degree of variation,
and because of this the differences between the estimates for Kingston and
the Main Towns samples, especially, are quite likely insignificant.
We may note finally, before considering the economic implications of our
results, that in the Kingston and Main Towns areas some of the regression
constants in our consumption functions have negative signs. This would imply
that a range of positive income exists for which consumption is negative, a
situation which is not amenable to commonsense. The explanation for this
phenomenon lies in the nature of the linear function and in the features of the


income distribution in the different samples. As in the case of the Engel curve
(cf. Section III) the linear formulation tends to be invalid for very low income
levels whenever the range of variation of the original observations is wide. It
is true that in the present analysis we have restricted the range of weekly
household income to between 0 and 15. However, the sums of squares and
of cross-products in the regressions have all been weighted by the number of
households appearing in each income group. So that, in effect, the differing
patterns of income distribution which prevail in the various sampling areas
(and which we have previously mentioned) have been incorporated into our
estimates of the consumption function. Since an overwhelming proportion of
households in the rural sample come from the same very low income groups it
is not surprising to find that the linear formulation of the consumption function
is valid for that sample; whereas in the urban areas, where a wider dis-
tribution of households by income prevails, the relationship between con-
sumption and either weekly earnings or Total Cash Supply are valid only
for households (albeit the vast majority) earning above certain minimum
amounts, i.e. approximately one shilling in weekly earnings or three-shillings
in Total Cash Supply.
No similar problems arise in interpreting the negative intercepts in the
log-linear consumption functions, since in these formulations attention is
focused on the rates of change of income and of consumption rather than on
their absolute levels.


Our estimates show that over the same given income range the proportion
saved from week to week out of any given increases in weekly earnings per
household rises from 1% in Kingston to 9% in the Main Towns and 27% in the
rural areas. Out of a given increase in the Total Cash Supply available to the
household (and this, it will be recalled, includes net dissavings and/or
increases in debts) the increase in the proportion "saved" is even more
striking; from about zero in Kingston to 44% in the rural areas. Since debts
incurred have to be paid off eventually and savings consumed have to be
replenished if the household is to remain a going concern, it is not surprising
to find, further, that for a given increase in Total Current Outlay more is
generally devoted to these purposes in the Kingston and Main Towns samples
than in the rural areas. The estimated consumption functions are therefore
quite consistent with what we would expect.
Following the Friedman line of argument, however, the fact that the
marginal propensity to consume out of weekly earnings tends to be higher in
urban than in rural areas is to be interpreted to mean that transitory com-
ponents of income loom larger in income changes in the rural areas. This
explanation is supported by the higher income elasticities in the urban areas,
for on the "permanent income" hypothesis these elasticities estimate P,
or the proportion of the total variance of measured income contributed by the
permanent component of income (cf. discussion in Section I). Following the
same line of argument.changes in Total Outlay contain a greater proportion of
permanent elements in the rural than in the urban areas.


Since changes in weekly cash earnings determine crucially changes in
the Total Cash Supply (cf. the relationship between Y3 and Y1) it is not sur-
prising to find that the elasticities of consumption with respectto weekly earn-
ings and Total Cash Supply behave similarly. The decline in the latter elasti-
city between Kingston and the rural areas is, however, more striking than the
decline in the former and this (again following Friedman) means that the de-
cline in the contribution of permanent components of income between Kingston
and the rural areas is greater for changes in Total Cash Supply than for
changes in weekly earnings.
We can, however, obtain more precision in our discussion if we calculate,
for each sample, estimates of the permanent and transitory components of in-
come in each income variable. Now since

P (Y Y )2
y y( y- y)2
we have
P Y(y- )2 (i -Y p)2

that is to say, the variance of the permanent component of income alone is
equal to the product of the income elasticity of consumption (P ) and the vari-
ance of measured income ( X (Y Y.)2 ). The square-root of P (y )2 as
usual, estimates the standard deviationof this permanent component.
Similarly, since the permanent and transitory components of measured in-
come are assumed to be uncorrelated in the Friedman hypothesis,
(Y.-y)- 2 = (yp p)2 + (Y t) 2

or (y t-
o(Y-Y)2 ((y- y)2
= 1- P

This means that the product of the variance of measured income and the com-
plement of the income elasticity of consumption estimate the variance of the
transitory components of income. Table 22 shows these estimated magnitudes.
It is to be noted that since we have employedhere the log-linear consump-
tion function, the assumption that the mean transitory components of income
are zero is not needed. Further, the estimated standard deviations of income
and its components are each to be interpreted as measuring the dispersion re-
lative to the mean. In other worcs they are to be regarded as estimates of the
coefficients of variation of the original observations. (These measures of re-
lative dispersion have each beer converted to natural logarithms by multiply-
ing by log e102-2.3026.).
One difficulty in applying these notions of permanent and transitory in-
come to the Jamaican data is immediately apparent from the table: what mean-
ingcan we attach to the relative dispersion of the transitory components when
the elasticity of consumption with respect to income is greater than unity? In
the case of the weekly-earnings and Total Cash Supply concepts of income in
the Kingston sample where theproblem arises, it is more than likely that mean
transitory components are positive, with magnitudes depending greatly on the
extent of unreported indebtedness, which, as we have previously observed,
provides the most important explanation of the Balancing Difference in this


Proportion of Total Variance Relative Dispersion Relative Dispersion Relative Dispersion
of Income Contributed by of Measured Income b of Permanent Component c of Transitory Component
Permanent Component a
Income Concepts

Kingston Main Rural Kingston Main Rural Kingston Main Rural Kingston Main Rural
Met. Area Towns Met. Area Towns Met. Area Towns Met. Area Towns

Average weekly
recorded income 1. 024 0.971 0.902 0.598 0.618 0.624 0.605 0.609 0.593 0.105 0.192

Total current
outlay 0.969 0.968 0.992 0.631 0.620 0.568 0.621 0.610 0.566 0.110 0.110 0.045

Total cash
supply 1.037 0.994 0.867 0.588 0.603 0.648 0.598 0.602 0.603 0.045 0.237

aElasticity of consumption with respect to income as estimated by the regression coefficients in the log-linear consumption functions.
Standard deviation of the natural logarithms of income.
CSquare-root of the product of the consumption elasticity and the variance of the logarithms of income.
dSqure-root of the product of unity minus the consumption elasticity and the variance of the logarithms of income.


Theproblem does not and cannot arise in the current-outlay concept of in-
come, since (net) positive cash savings and/or decrease in debts occur as the
level of household income increases in all sampling areas.
We are inclined to feel, further, that there is some point ih arguing that,
in view of the social and psychological attitudes to high spending found es.-
pecially among urban Jamaican consumers (cf. discussion in Section I), tran-
sitory components of income are not entirely unrelated to permanent com-
ponents, and are often treated as if theyinvolved changes in permanent income
status. Such attitudes are more easily satisfied when credit is easily avail-
able, as tended to be the case in the mid 50's in Jamaica. Having said this,
however, we have not contradicted the essential features of the permanent in-
come hypothesis. We are merely suggesting an enlargement of the notion of
permanent income to include transitory, largely unearned income, the response
to which, on the part of consumers, is like their response to a change in per-
manent income. On this view it is not surprising to find that the proportion of
the total variance of weekly earnings contributed by the permanent component
of income sometimes exceeds unity in our present sample of Kingston house-
A remarkable feature of our results is the stability of the relative disper-
sion of thepermanent component of income, however defined,in all areas. All
the estimates of this magnitude lie between 56.6% and 62.1% of mean income.
The lowest estimate is found in the rural areas and the highest in theKingston
area. On the other hand, excluding the two ambiguous cases already mentioned,
the estimated coefficients of variation of the transitory component vary be-
tween 4.5% and 23.7% of mean income. Subject to our qualifications above,
much of the observed difference in the elasticities of consumption with respect
to income, as variously defined, would therefore appear to derive from differ-
ences in the relative dispersion of transitory components of income. The low-
est estimate of this elasticity coincides with the highest estimate of relative
dispersion of the transitory component and the highest estimates of the elasti-
city coincide with the lowest estimate for the coefficient of variation of the
transitory components.
It is these features of the income structure in the three sampling areas
which, according to the permanent income hypothesis, explain the differences
in average "savings" proportions. In Table 23 we show theratiosof arithmetic
mean consumption to arithmetic mean income (as variously defined) for house-
holds within the income range 0 and 15.


Ratio of Arithmetic Mean Consumption to
Arithmetic Mean Income in:
Income Concepts
Kingston Main Towns Rural Areas

1. Average weekly
recorded income 0. 983 0. 896 0.837
2. Total current
outlay 0. 972 0. 972 0.994
3. Total cash
supply 0.970 0.886 0.756


The average propensity to "save" is highest in the rural areas and for
the Total Cash Supply concept of income, the corresponding estimate of
relative dispersion of transitory components is also the highest. The average
propensity to "save" out of weekly earnings is relatively high in the rural
areas and this also corresponds with a relatively high estimate of the dis-
persion of transitory components. The average propensity to "save" out of
Total Current Outlay is lower in the rural than in the urban areas, and this
again corresponds with a relatively low estimate of dispersion of transitory
components. All these results are quite consistent with the permanent income
hypothesis as we understand it.
There are other implications which arise out of our calculations. The
relationship between Total Current Outlay and Total Cash Supply shows the
extent to which households increase their total weekly cash allocations'in
response to given increases in weekly recorded cash receipts from all sources.
The difference between these two income concepts reflects, as we have
previously suggested, the extent to which households are involved in production
for the market. Since each of these income variables is dependent upon weekly
earnings, it is possible to derive estimates of a household "investment"
function by calculating the difference Y3 Y2 in terms of the ultimately
determining variable, Y1. This is done in Table 24. In order to provide a
comparison we also show the estimated short-term savings function implied
by the relationship between weekly consumption expenditure and weekly
recorded income in each sample.


"Investment" Functions

Kingston Y3 Y2 = 8.660 0.083 Y1
Main Towns Y3 Y2 = 6.805 0.048 Y1
Rural Y3 Y2 = 3.368 + 0.317 Y1

Short-Term "Savings" Functions
Kingston Y1 C = -0.949 + 0.008 Y1
Main Towns Y1 C = -0. 873 + 0. 095 Y1
Rural Y1 C = 7. 106 + 0.274 Y1

The marginal short-term savings and investment propensities are much
higher in the rural than in either of the urban samples. In a sense none of the
households in the different areas shows any tendency to live within its means
or its command over financial resources. The urban household tends to consume


more than its income in the short-run and less than its income in the long-run.
The rural household tends to consume more out of income in the long-run and
less in the short-run. We thus have here another feature distinguishing the
urban from the rural consumer in Jamaica: the greater preference for present
as opposed to future consumption among urban households. This feature is
not only a result of the differing roles of the typical household in the urban
and rural areas but seems to have been due also to the different price ex-
periences of urban and rural consumers, especially for the period under
investigation. As we shall see shortly theprice index has tended to rise faster
in the urban areas. This has quite probably had the effect of making urban
consumers desire to hold less money and more goods in the present' since the
value of their money was depreciating at a faster rate. The result is upward
pressure on the already high average propensity to consume out of current
earnings in these areas.


A study of the changing pattern of distribution of total expenditure as
between food, clothing, rent etc. is one way of obtaining a clearer picture of
changes in the level of living of the population. From such a study it is also
possible to obtain insights into the effects of changes in prices, "real"
income etc. on the quantitative intake of consumer goods and services.
We may notice at the outset, however, that since the aggregate con-
sumption-income relationship in Jamaica appears to be proportional, the
implication is that the summed effects of all variables other than income
upon the individual components of aggregate expenditure have, over the long-
run, been negligible. With respect to prices in particular, the proportionality
of the aggregate consumption function justifies rather strongly the ise of the
homogeneity assumption often made in empirical studies of demand.39 It is
nevertheless quite conceivable for sudden changes in relative prices, changes
in the demographic features of the population, introduction of new and superior
consumer goods etc. to upset the proportional relationship;40 but there seems
to be no a priori reason for believing that the effect of these changes on the
consumption-function can be anything but transitory.
According to Eisner's national income estimates, total personal con-
sumption expenditure at current prices increased by more than five-fold between
1832 and 1930. Expenditure on food increased, over the same period, some
six-fold; expenditure on clothing tripled, while expenditure on other consumer
goods increased by almost ten-fold.
The proportion spent on food increased from 44.3% in the immediate pre-
emancipation period to 49.3% in 1950. It remained steady at 53.7% during the
entire period of Crown Colony government, and even after banana cultivation
superseded sugar as the main staple. By 1930 the proportion spent on food
had returned to the 1850 level (See Table 25). The proportion spent on clothing

39See article by Stone, "The Analysis of Market Demand", op. ci t.
40See G.A. Ackley and D.B. Suits, "Relative Price Changes and Aggregate Consumer De-
mand", American Economic Review, Vol. XL, Dec. 1950, p.785 foil.; also Arthur Smithies,
"Forecasting Postwar Demand: I", Econometrica, Vol. 13, 1945, pp. 1-14.


1832 1850 1870 1890 1910 1930


000 % 00 % 000 000% 000 % 000 %

1. Food, drink and
tobacco 1,486.2 44.3 1,655.4 49.3 2,771.5 53.7 4,271.6 53.7 5,382.2 53.7 9,044.0 49.4

2. Clothing 648.9 19.3 603.1 18.0 792.1 15.4 1,228.4 15.4 1,605.3 16.0 2,225.1 12.1

3. Other consumer
goods 299.0 8.9 280.2 8.3 517.2 10.1 689.9 8.7 787.4 7.9 2,355.0 12.9

4. Personal services 363.2 10.8 393.4 11.8 444.8 8.6 646.8 8.1 859.3 8.6 1,765.4 9.6

5. Passenger trans-
port 23.7 0.5 89.7 1.2 165.3 1.7 386.5 2.1

6. Rent 552.9 16.7 416.2 12.6 595.7 11.7 1,018.7 12.9 1,215.7 12.1 2,546.5 13.9


Source: Eisner, op.

3,350.2 100

cit., Table 8.V.



5,145.0 100

7,945.1 100

10,015.2 100




on the other hand, fell from 19% of the total personal consumption expenditure
in 1832 to 12% in 1930. The large absolute increase in money expenditure on
the group of items under the title "Other Consumer Goods" meant also a large
relative increase in expenditure on these items. The proportions devoted to
personal services and rent fell modestly over the period.
The changing pattern of consumption in "real" terms over the period is
somewhat different from that just described. The only item of consumers'
expenditure which appeared not to have experienced some considerable price
inflation over the century beginning with 1830 is "Home Produced Clothing"
i.e. "'...services rendered by dressmakers, tailors, shoemakers, etc. in making
up imported materials".41
The prices of imported food and clothing increased by almost 50%, while
the prices of home-produced food and other consumers' goods more than
doubled. (The proportion of total personal consumption expenditure spent on
imports was estimated to be 28%, half of which was devoted to food.42)
When personal expenditures at current prices are deflated by the price
indices, the pattern of consumption shown in Table 26 emerges.

JAMAICA, 1832-1930

Items 18 32 1850 1870 1890 1910 1930

1. Food, etc. 55.1 56.6 56. 3 53.9 53.9 52. 3

2. Clothing 15. 2 16.7 15.7 18. 1 16.0 12. 9
3. Other consumer
goods 5.9 5.7 7. 1 7.8 7.7 11.7
4. Personal services 11 7 9,7 7.8 7.0 8.6 8.0

5. Passenger trans-
port 0.5 1.5 1.7 2.8

6. Rent 12.1 11.3 12.6 11.7 12.1 12.3

Total 100 100 100 100 100 100

The decline in the proportion spent on clothing is now much less spec-
tacular. The proportion spent on food, etc. shows a moderate decline after
emancipation. The proportion spent on other consumer goods is lower in
"real" terms, but over the entire period it behaves rather similarly in "real"
as in monetary terms. The rent proportion remains relatively constant in
"real" terms, whereas in money terms it had tended to decline.
From the evidence of Eisner's data there appears to have been no "real"
startling increase in the level of living of the Jamaican population between

Eisner, pp. 39-40.
Eisner, Appendix, I and II.


1832 and 1930. Personal consumption expenditures per caput (at 1910 prices)
rose modestly from 11.2 per year in 1832 to 13.5 per year in 1929, 1930 and
1931.43 (Aggregate expenditures at constant prices (1910 = 100) rose from
4.1 m. in 1832 to 13.7 m. by the end of the decade of the 20's). However,
during the period significant changes were occurring in the pattern of income
distribution of the Jamaican populace. Small-settlers' income after emancipation
increased relatively to planters' profits. This was a direct result of the decline
of the plantation economy and the rise of peasant-farming, which were
dominant trends during the nineteenth century. There was some significant
improvement in levels of living, especially between the years 1870 and 1930.
Between these years the proportion of total expenditure devoted to necessaries
declined markedly as expenditure on passenger transport, personal services
and other consumers' goods increased.
By 1938, total personal consumption expenditures was estimated by Deane
to have risen to 18.4 m. (at current prices). This total was distributed as
shown in Table 27.
The proportion devoted to food is still in the region of 50%, but the
proportion spent on clothing has fallen significantly below the average pro-
portion (12.1%) for the years 1929, 1930 and 1931. Rent has remained fairly
steady compared with the Eisner estimates.
The general price index fell by about 30 points between 1930 and 1938,
while the index of value of imports for home consumption increased by 3
points.44 At 1910 prices, aggregate personal consumption expenditures in
1938 stood at approximately 17.6 m. The mean population of Jamaica in 1938
was 1,137,000, so per caput consumption expenditure at constant (1910 = 100)
prices in that year was about 15.5 as compared with the estimated 13.5 in
1930. Some increases in the level of living of Jamaican consumers therefore
did occur during the inter-war years and during the heyday of banana culti-
The changes in the pattern of income distribution over these years were,
however, not as pleasing. In 1930, the total income of agricultural labourers,
workers in construction and manufacturing, and domestic servants was estimat-
ed at 3,925,000 or about 19% of the national income for that year.46 In 1938
the income of the same group of workers was somewhere in the region of

43Eisner, p. 319, foll.
Deane op. cit., p.141.
The price index of clothing and many food items follows closely the overall import price in-
dex, which index for the years 1930-38 moved as follows:

1929 110 1932 75 1935 79
1930 97 1933 69 1936 80
1931 79 1934 76 1937 95
1938 100
The overall general price index, on the other hand, moved as follows:
1929 134 1932 109 1935 97
1930 128 1933 106 1936 90
1931 117 1934 101 1937 102
1938 100
In "real" terms therefore the proportion spent on food and clothing fell by more than the
money proportions would indicate.
4The net taxable income from banana cultivation was estimated at 1,839,000 as compared
with 900,000 from sugar and rum in 1938 (Deane, Table 96, p.125).
Eisner, p.99.



'000 %

Food, drink, tobacco 9, 200 49.9
Clothing 1,814 9.8
Rent 2, 239 12. 1
All other 5, 183 28. 2

Total 18, 436 100
Source: P. Deane op. ci t. p. 125.

5,487,000 or 20% of total taxable income. Other wages and salaries over the
same period moved from 2,322,600 (11.5% of national income) to 1,580,000
(7.8% of national income); small-settlers profits from 6,945,200 (35% of
national income) to 1,346,000 (6.5% of national income); other profits and
salaries from 5,447,000 (27% of national income) to 8,726,000 (more than
40% of national income in 1938).
Thus while the share of national income going to labourers and small-
settlers at the best remained steady, the share going to other profits and
salaries increased over the decade of the '30's. The distribution of income
then tended to become more uneven, granted even the assumption that the rate
of entry to the labour force is equal for all occupational groups.47
The world depression of the later '30's and increased unemployment did
not improve the position.. It was not surprising, therefore, that serious labour
disturbances broke out in 1938, with the result a Royal Commission of
Fnquiry was set up to investigate the social and economic conditions existing
in the West Indian islands 48
A similar combination of increasing average levels of living, high un-
employment rates but this time coupled with external migration and some
amount of overt discontent (though not on the scale of the '30's) has occurred
within recent times. In terms of the political aspects of economic growth, the
growth in relative shares of output is evidently at least as important, as the
growth in the "average level of living".
Morals' national income estimate for the year 1943 shows total personal
consumption expenditure at current (market) prices to have risen to 35.2 m.
(see Table 28).
The proportion of total expenditure (at current prices) devoted to food,
drink and tobacco increased from 50% in 1938 to over 60% in 1943. Expenditure
on clothing, footwear, etc., also increased its relative share; rent however
fell relatively from 12.1% in 1938 to 9.1% of total expenditure in 1943. This
The employment statistics for these years, although very poor, seem to show, if anything,
a greater percentage increase in the numbers of labourers, small-settlers and domestics. The
Census estimates(1921) of thepercentage in these occupations is 76%; and according to Deane's
estimate, the proportion similarly engaged in 1938 was about 80% of the total estimated labour
force (excluding the estimated numbers of unemployed). See Eisner, p.162 and Deane, p.97.
4Report on the West India Royal Commission, 1945, C.M.D. 6607.



Items Imports Local Products Total

m. % m. % Em. %

Food 2. 2 6. 2 16. 5 46. 9 18.7 53. 1

Drink and tobacco 1.2 3.4 1.4 4.0 2.6 7.4

Clothing, footwear, etc. 3.7 10. 5 2. 1 6. 0 5.8 16. 5

Housing and Household goods 0. 3 0.8 5. 6 16.0 5. 9 16.8
All other 0.5 1.4 1.7 4.8 2.2 6.2

Total 7.9 22. 3 27. 3 77.7 35. 2 100

Source: Morais, op. cit., Table 16.

apparent reversal in the trend away from "necessaries" was no doubt due in
large part to the numerous price and import controls imposed on several
articles of consumption during the war years. The movements in the Cost of
Living Index (urban) between August 1939 and September 1943 are shown in
Table 29.


Year Food Rent Clothing Flel and Ohr Its

Aug. 1939 100 100 100 100 100 100

Sept. 1940 122. 5 102. 5 158.7 124.6 115. 1 122.0
1941 139.5 106.9 225.6 146.5 123.4 141.4

"1942 145.7 112.0 281.0 157.4 149.6 155.0

1943 145.5 117.0 287.8 173.9 158.7 158.7

Source: Department of Statistics, Annual Digests (1939-1943).

The largest price increases occurred among items of clothing, fuel and
cleaning and other imports. The average price of food increased by almost
50% between 1939 and 1943. If we make allowance for the distortionofprice-
pattern during the war years we arrive at the following pattern of consumption
expenditures for 1943 at constant (August 1939 = 100) prices (Table 30).


(AUG. 1939 = 100) PRICES, JAMAICA 1943

Items m. %

Food, drink and tobacco 14.7 65. 0
Clothing 2.0 8.8
Rent 2.8 12.4
All other 3. 1 13 8

Total 22.6 100.0

The effect of removing the price inflation has been to push up the "real"
proportionate expenditure on food, and to halve the share spent on clothing.
"Real" per caput income is now 10.8 (at 1939 prices) as compared with
approximately 16.1 in 1938. The average level of living thus fell slightly
during the war and this fall is to some extent confirmed by the increase in
proportionate expenditure on "necessaries." (The proportionate expenditure
on food, beverages and tobacco at constant (1939 = 100) prices increased
from 50% in 1938 to 65% in 1943; clothing declined modestly from 9.8% to 8.8%;
rent remained steady at 12% but all other items fell absolutely from 5.2 m.
(28.2%) in 1938 to 3.1 m. (13.8%) in 1943).49
After the war the growth in living standards in Jamaica was resumed.
There is no immediate post-war estimates of aggregate consumption spending,
but Thorne estimates that by 1950 total personal consumption expenditures at
current market prices had reached 71.8 m., i.e. almost double the 1943
estimate and more than three times the estimated value for 1938 (Table 31)



1950 1951 1952
(Em.) % (mn.) % (tm.) %

Food, beverages and to-
bacco 35.8 (49.8) 40.4 (49.3) 46.8 (49.3)

Clothing, etc. 8-6 (12.0) 9.9 (12. 1) 9. 6 (10. 1)

Rent and water 5.6 ( 7.8) 5.9 ( 7.2) 6.3 ( 6.7)
Household durables,etc. 5.5 ( 7.7) 6.9 ( 8.6) 9.8 (10.4)
Transportation and com-
munications 8.0 (11. 1) 8.8 (10.7) 9.9 (10.4)

All other 8.3 (11.6) 10.0 (12.1) 12.4 (13.1)

Total 71.8 (100.0) 81.9 (100.0) 94.8 (100.0)

At current market prices "All other items" fell from 5.2m. in 1938 to 4.9m. in 1943.


The actual outlay on food, beverages and tobacco in 1950 is greater
than the 1938 and 1943 figures by 26.7 m. and 14.5 m. respectively. Ex-
penditure on food, etc. thus increased on the average by some 2.2 m. per
year between 1938 and 1950. Expenditure on clothing, etc. increased from
1.8 m. in 1938 to 8.6 m. in 1950; expenditure on rent rose from 2.2 m. in
1938 to 3.2 m. in 1943 and 5.3 m. in 1952. Second only to food, etc. the
largest absolute increases in expenditure since the pre-war and mid-war days
occurred among household durables and household equipment, purchases of
automobiles, travelling expenses and miscellaneous services. Expenditure
on these "luxury" type goods and services increased by 17 m. over the 1938
estimate and by 14 m. over the mid-war figure.
The pattern of consumption expenditures in the early '50's is therefore
somewhat different from the pattern revealed in the earlier national income
data. Food, beverages and tobacco in 1950, 1951 and 1952 still comprise
about 50% of total consumption expenditure, but the proportion spent on clothing
and footwear declined sharply from 16% in 1943 to 10% in 1952. Rent also
declined from 12.1% in 1938 and 9.2% in 1943 to 8% in 1950 and 6% in 1952.
The large absolute increases in expenditure on "luxury" type items(noted in
the previous paragraph) also meant an increase in the proportion spent on
these items from 28.2% of total expenditure in 1938 to 52% in 1952.
Price increases over the period were, as is to be expected, considerable
and uneven. The (urban) Cost of Living Index more than tripled between 1939
and 1952. The prices of clothing items almost doubled between the end of the
war and 1950 and more than quadrupled since the immediate pre-war years. The
index of rent rose modestly by 60% above the base year while food, fuel and
cleaning and other items more than doubled.

Cost of Living Index (Urban), Jamaica, (August 1939 = 100)
July 1950
Food Rent Clothing Fuel and Cleaning All Other Total
256.1 158.6 451.5 244.4 239.4 255.6
When the expenditure figures for 1950 are expressed in constant (1939= 100)
prices, we get the following pattern:
Personal Consumption Expenditures at Constant(August 1939= 100)Prices,
Jamaica, 1950.
m. %
Food 14.0 50.0
Rent 3.6 12.8
Clothing 1.9 6.8
All other 8.6 30.4

Total 28.1 100
Total "real" expenditure per head is estimated to be 19.8 in 1950 as
compared to approximately 16 in 1938 and 10.8 in 1943. Proportionate
expenditure on food in "real" terms remained at 50% of total "real" ex-'
penditures, but clothing dropped from 9.8% in 1938 to 6.8% in 1950 while the
"luxury" items increased from 28% to 30%.


The considerable changes which occurred in the pattern of "working-
class" expenditures, as well as in the range of consumer- goods available to
the post-war Jamaican consumer, of course, diminishes considerably the
usefulness of the comparison between pre-war and post-war "constant price"
data. It does, however appear from the "current" price estimates ofexpenditures
that the distribution of expenditures between "luxury" items and "necessaries"
remained (except for the war years) very much the same between 1938 and
1950, in spite of fairly substantial increases in aggregate and per caput
income. The decline in the proportion spent on rent is due largely to the house-
purchasing boom in the post-war years. In spite of considerable increases in
the prices of itemsof clothing etc. the proportion of "real" expenditure devoted
to these items remained roughly the same in 1950 as in 1938. (The price
inflation occurring among these items probably reflects some improvement in
"quality" as well as increased profiteering by drygoods' merchants).
The need for a new weighting pattern for the Cost of Living Index, as we
have previously indicated, led to the urban and rural Household Budget Surveys
of 1953/54 and 1956. The new weights, now based on -a new definition of the
"working-class" household in Jamaica and an extension of the old commodity
groups are shown in Table 32.

Urban Rural

Food 51.5 58.0
Rent etc. 12.0 4.0
Clothing 8.0 16.0
Fuel and cleaning 7.5 6.0
Other personal expenses 9. 0 8. 0
Other household expenses 12.0 8.0
100. 0 100. 0

We have already touched on the differences in expenditure patterns between
urban and rural households in Jamaica. We now wish to use this expenditure
pattern to measure the changes in "real" levels of consumption expenditures
in the '50's. In order to do this, however, it seems advisable to obtain an
All-Island Retail Price Index by combining the expenditure patterns and prices
of the urban and rural areas.
According to the 1953 Sample Survey of Population the total number of
households in Jamaicainthatyear was 377,757, of which total 85,188 resided
in the Kingston Metropolitan Area.' 0 Thus approximately 78% of all households
in Jamaica resided in country parishes in the mid '50's. Combining the rural
and urban expenditure patterns, according to the proportion of households in
each of these areas, we arrive at an all-island weighting pattern for retail
prices. These weights are shown in Table 33.
For the years 1957 1961, we can now calculate All-Island Retail Price
Indices which make some allowance for locational differences in household
expenditure patterns and in prices (see Table 34).

500p. cit. Table 60, p. 104.


JAMAICA, (DEC. 1955)

Food 56.5
Rent, etc. 5. 7
Clothing 14.3
Fuel and cleaning 6. 4
Other personal expenses 8.2

Other household expenses 8.9

Total 100.0


Items 1957 1958 1959 1960 1961

Food 103 110 111 113 122
Rent, etc. 103 105 107 112 123
Clothing 101 103 104 107 112
Fuel and cleaning 102 105 106 108 116
Other personal expenses 99 104 110 111 121
Other household expenses 103 105 105 106 108

Total 102 107 109 111 119

aEconomic Survey, 1960 P.51, and Economic Survey, 1961 P-67.
It should be obvious that this index would move closer to the rural Retail
Price Index than to the urban Index owing to the numerical importance of rural
households in Jamaica. The average level of prices in the urban areas tends
to be higher than in the rural areas. For the years shown this was due largely
to the higher prices paid for food items, and rent. The rate of change of urban
prices tended also to exceed the rate of change of prices in the country districts
over theperiod. Urbanprices increased by about 5% per year, rural prices by
3.8% per yearand our All-IslandIndex by 4.2% per year between 1957 and 196L
For the period 1950-56 we have had- to rely solely on the urban price
index in order to get a picture of general price changes throughout the island.
We thus expect our All-Island Price Index to overestimate both the level and
rate of change of prices throughout the island for the early '50's. But we do
not expect the overestimate of the rate of change of (all-island) prices to be
as great as the overestimate of the level of (all-island) prices. The estimated
movements of all-island prices for 1950-61 are shown in Table XIV(App).


Between 1950 and 1956 prices of all consumer goods rose by an estimated
25%, this being due largely to an increase of 31% in the prices of food items
and 25% in the price indices of rent and clothing. The estimated "real"
changes in consumption expenditures between 1950 and 1956 are shown in
Table 35. These estimates differ only slightly from similar estimates made
by Bethel.51

PRICES,(DEC. 1955:= 100), JAMAICA, 1950 56

Items 1950 1951 1952 1953 1954 1955 1956

Food and beverages 48.8 49.2 45 1 44.6 50.9 53.7 59.3
Clothing 10.5 9.1 9.0 10.6 13.2 14.2 14.9
Rent, etc. 7.3 7.6 7.8 7.7 8.3 8.2 8.1
Fuel and cleaning 7.4 8.3 11.1 10. 2 11.0 12.2 11 8
All other 15.8 16.8 19.9 28.3 34.8 39.6 46.5
Total 89.8 91.0 92.9 101.4 118.2 127.9 140.6

The difference between "real" and current price expenditures is rather
striking. Total personal consumption expenditures at current market prices
increased from 71.8 m. in 1950 to 142.0 m. in 1956, i.e. by 70 m. or almost
100%. In "real" terms the increase was 50.8 m., representing 57% of the
1950 constant price figure.
Almost 50%, then, of the observed increase in current consumption ex-
penditures during the first half of the decade of the '50 s was due to price
Between 1950 and 1952 increases in current consumption expenditures
were almost entirely due to price inflation. The greatest increases in "real"
consumption expenditures occurred between 1953 and 1955. Inper caput terms
"real" consumption increased from roughly 66.0 in 1950 to 94.0 in 1956
that is by approximately 4.7 (or 7%) per head per year. This represents a
considerable and unusual improvement in the average level of living of Jamaican
consumers. Real outlay on food and beverages increased by 11 m. over the
entire period, after falling in 1952 and 1953. The proportion spent on food and
beverages moved from 54% in 1950 to 42% of total "real" consumption expen-
ditures in 1956. The proportions spent on rent, etc. and clothing in 1950 were
8.1%and 11.7% respectively. By 1956 the proportion of "real" total consumption
expenditures devoted to rent, etc. had fallen to 5.8% and the clothing pro-
portion had moved to 10.6%. The "real" expenditure on "All other" items
more than doubled between 1950 and 1956, moving from 18% of total expen-
diture in 1950 to 33% of total expenditure in 1956. By 1956, then, Jamaican
consumers were consuming in "real" terms roughly the same amounts of food
and beverages as of "luxury" items.
Between 1957 and 1959 total consumption expenditures at current market
prices increased by 24.9 m. The estimated increase at constant (Dec. 1955 =
100) prices was 13.6 m. (Table 36).

J. Bethel, op. ci t.


PRICES (DEC. 1955= 100), JAMAICA, 1957-1960.

Items 1957 1958 1959 1960

Food and beverages 62. 1 62. 7 66. 3 67. 8
Clothing 17.7 17.0 17.9 21.3
Rent, etc. 8.4 8.8 8. 5 8.8
Fuel and cleaning 12.5 13. 2 13.9 14.6
All other 53.7 56.0 61.4 66. 1

'Ttal 154. 4 157. 7 168.0 178. 6

In per caput terms total "real" consumption expenditures increased
from 98 in 1957 to 111 in 1960 or by approximately 4.3% per year over the
three-year period. There was therefore some slowing-down in the growth of
average levels of living towards the end of the decade. The upward trend in
"real" expenditure on "luxury" type goods and services, however, continued,
and by the end of the decade expenditure on these items comprised the largest
proportion of total "real" consumption expenditures. Real expenditure on food
and beverages rose modestly by 6 m, between 1957 and 1960; expenditure on
clothing and rent remained relatively steady in "real" terms.
The retardation in "real" growth of average levels of living which we
have just noticed was due partly to unusual price inflation of certain locally-
produced foodstuffs during the latter years of the decade (especially in urban
areas) and was certainly affected by the sharp rise in rents (including electri-
city, rates, taxes, etc).52 The distribution of personal consumption expen-
ditures in 1959 nevertheless shows that by the beginning of the '60's, Jamaican
consumers had fairly satisfied their demand for the basic "necessities" of
life and were now devoting larger proportions of their income to the purchase
of more sophisticated goods and services, especially within the groups Trans-
portation and Communication, and Household Durables.

Tourist Expenditure

The expenditure figures which we have been discussing so far refer to the
expenditures of all persons permanently or temporarily resident in Jamaica.
They thus include expenditures by tourists, which amounts have grown rapidly
in the post-war years. They also include the value of gifts sent abroad by
residents, but these -amounts are less substantial (See Table 37).
"National" consumption expenditure is arrived at by adding together
"Expenditure on all items" and "Expenditure of residents abroad" and sub-
tracting from the total "Tourist expenditure in Jamaica" and "Value of gifts
sent abroad." The last row of figures in the table therefore represents the
expenditure by Jamaicans in Jamaica in each year of the decade 1950-1959.
52Economic Survey 1960 pp. 5152.
Economic Survey 1960, pp. 51-52.


Items 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959

on all items 71.8 B1.9 94.B 104.0 115.8 130.5 142.0 157.5 168.7 182.3
of residents
abroad 0.4 0.6 0. 6 0.7 0.8 0.9 1. 0 1. 1 1. 2 1. 2
Tourist expendi-
ture inJamaica 4.4 4.6 6. 1 7.8 9.9 10.3 11.6 12.6 13.2 15.7
Value of gifts
sent abroad 0.04 0.03 0.04 0.07 0. 08 0. 09 0. 10 0. 12 0. 14 0. 15
Expenditure 67.8 77.9 89. 2 97. 3 106. 6 121.0 131. 3 145.8 156.6 167.7

It is clear that during the '50's tourist expenditure in Jamaica increased
at a faster rate than national expenditure itself. From 6.2% of total expenditure
in 1950 it fell slightly to 5.6% in 1951, thereafter increasing continuously up
to 1954, in which year it stood at 8.3%. There was some slackening off in the
rate of growthin the second half of the decade, and in 1959 tourist expenditure
comprised 8.3% of total expenditure.
The effect of tourist spending on the pattern of aggregate consumers'
spending in Jamaica is not difficult to see. Most of the spending takes the
form of payment of bills at hotels, clubs and restaurants. The second largest
allocation is in the form of taxi-cab fares. Purchases of native straw goods
and wood products come next in order of importance in the tourist's budget.
The tourist's pattern of expenditure is therefore vastly different from the
expenditure pattern of the resident Jamaican household, as is to be expected.

We might notice also that whereas tourist expenditure accounted for
almost 10% of total expenditure in Jamaica in 1959 the contribution of the
tourist industry as a whole to the national income of Jamaica was a mere 2.4%
in 1959 (See Table 38). Its contribution to national income has however
increased since 1954 (when it stood at 1.6% of national income). The increas-
ing contribution has been largely due to a greater reliance on local supplies
by hotels catering to tourists. (Hotel expenses comprise more than 50% of the
tourist's budget).
The pattern of expenditure as displayed by the tourist has had some
effect on the trend towards "luxury"' consumption in Jamaica, which we
discussed above. Almost all expenditure in the national accounts under the
sub-headings straw goods, wood products, hotels, etc., and a large proportion
of liquors, clubs and restaurants and taxi-cab services reflect expenditure
by tourists.
The effect on the overall pattern is however slight and the "multiplier"
effect of tourist expenditure on national income and national consumption
expenditure must be very small indeed judging from the national income
generated by the industry.


1954-1959. (Em.)

1954 1955 1956 1957 1958 1959

National income from
tourism 1.82 2.52 3.23 3.93 4.23 4.44
Tourist expenditure 9. 96 10.28 11.61 12.57 13.16 15.66
Total national income 109.94 125.11 143.23 170.99 177.27 187.67

Source: National Income Section, Department of Statistics, Jamaica.


We have so far been looking at aggregate consumers' expenditure and its
components as shown in the national income accounts and budget surveys of
Jamaica. We have seen that there have been fairly substantial increases in
aggregate "real" and money consumption expenditures. The increases in
average "real" per caput expenditures have however been rather sluggish
over most of the period covered. In the century following emancipation "real"
annual per caput expenditure remained relatively constant; but there were
significant movements in the direction of a more equal distribution of income
consequent on the change from plantation economy to peasant cultivation.
Between 1930 and the beginning of the second World War, i,e, during the
heyday of banana cultivation, "real" expenditures increased by about 2 per
head per year. There was also a modest change in the pattern of consumption
awayfrom "necessaries" and toward "luxuries". The trend in the distribution
of income, however, seemed to have reversed with a larger proportion of
national income going to profits.
The world-wide depression of the late '30's served only to aggravate the
frustration of the labouring classes, and a series of disturbances occurred
during 1938. Out of these violent disturbances was born the political-trade
union movement of modern Jamaica.
The second World War had the effect of lowering the average "real" level
of living of Jamaican consumers; but after the war the tempo of growth in the
level of living accelerated. The increasing industrialisation of the Jamaican
economy, the discovery of bauxite and other post-war developments had the
effect, in spite of rapid price inflation, ofincreasingthe "real" average per
caput level of consumption by about 4% per year. By the end of the decade
of the '50's the demand for "necessaries" seems to have levelled off and the
expansion of "luxury" items of consumption, e.g. motor-cars, refrigerators and
other household durables, fed by easy credit facilities, had begun to accelerate.
Part of this acceleration of "luxury" purchases was perhaps due to the rapid
increase in tourist arrivals and expenditures in the island.
The trend of the resident Jamaican household towards "luxury" purchases
is nevertheless unmistakably clear.








In this section we discuss in somewhat more detail the components of
total household expenditure. The discussion is centred on two aspects:
1. The differences in absolute and relative disbursements between income
groups and between the three sampling areas.
2. Types of Engel curves appropriate for describing the observed variation
in expenditures by Jamaican consumers.


In each sample, the changes in expenditures as the average level of
household income increases are fairly consistent with what we would
normally expect. For households within the given income range 0-15 per
week the decline in proportionate expenditure in the category Food is well
manifested. All the items within the Food group, with the notable exception
of (Alcholic and Non-Alcholic) Beverages, and the possible exception of
Milk Products and Eggs, tend to behave like "necessaries". But among the
non-food items the changes in expenditure patterns, as income increases, are
not as unmistakably clear. Proportionate outlay in the category Clothing tends
to increase modestly throughout; proportionate outlay on Housing (including
Fuel and Light)decreases at first and then increases. Expenditure on Durables
behaves differently in each of the three samples. In the Rural sample the
Durables proportion increases monotonically with income, in the Main Towns'
sample its behaviour is U-shaped and in the Kingston sample the proportion
increases at first and then seems to decline slowly.
In all income groups the tendency is, among Food items, to concentrate
expenditure on Fruit and Vegetables, Meat and Fish and Cereals and Bakery
Products; there is also a tendency for proportionate outlay on these items to
increase as we move away from Kingston to the Rural areas. Among non-food
purchases proportionate expenditures on Housing and Clothing figure pro-
minently in all income groups and in all samples. But while the Housing
proportion is almost twice the size of the Clothing proportion in the Kingston
and Main Towns' samples, in the Rural sample the proportions are approxim-
ately on a par for most income groups. Most of the other commodities in the
non-food group behave rather irregularly as household income increases. In
general, however, these proportions tend to rise more rapidly with income in
the Kingston and Main Towns' samples and tend also to be, on the average,
higher in these areas.
It is therefore obvious that marked differences in expenditure patterns
exist between different localities and different income groups in Jamaica, and
some kind of index to show up more clearly and more simply these changes
in expenditure patterns is needed.
As a first approximation to such an index we may, by comparing adjacent
income groups, obtain estimates of income elasticities of demand for the
various commodities. We may refer to these estimates as arc-elasticities, to
distinguish them from the estimates of elasticities at a point on a smooth
curve fitted to the data.
An arc-elasticity of demand which increases as income increases will
indicate an increasing propensity to purchase the commodity in question, and
one which decreases as income increases will indicate a decreasing propensity.
Table XVII shows the estimates for each sample. The estimates are each the
product of the expenditure elasticity of demand for the commodity group and
the elasticity of total expenditure with respect to estimated long-term income
in each sampling area' (cf. Table XIII).
1S.J. Prais and H.S. Houthakker p.101.


The arc income-elasticity of demand for Food items tends generally to be
less than one and to decline as the level of household income increases, which
is in keeping with expectations. The estimated elasticities for non-foods are
less well-behaved but they generally tend to be greater than one. Both these
features are in keeping with the classic predictions of Engel's law and bring
out in sharp relief the contrasting behaviour of expenditure on "necessaries"
and on "luxuries".
Within the food group, expenditure on Beverages, Food consumed away
from home, Meat and Fish and Milk Products and Eggs are "luxury" or "semi-
luxury" purchases among households earning less than 4 per week arc-
elasticities varying between 0.8 and 1.5 The arc-elasticity of demand for
Edible Oils, on the other hand, seldom rises above 0.5 for most of the groups
compared. The estimated elasticities for the other Food items generally fall
between 0.5 and 1.0.
Expenditures on Durables and Transportation are "luxury" type expenditures
among most of the income groups compared. The propensity to spend on Hous-
ing tends to increase with the level of household income while the elasticity
of demand for Clothing seems to remain invariant and above unity.
The arc-elasticity of demand for expenditure on Personal and Medical
Care falls with increases in income in the Kingston and Rural areas and
behaves rather erratically in the Main Towns' sample. The arc-elasticities of
demand for the other non-food items are highly irregular in their behaviour
but their values are generally larger for households moving between the two
lowest income groups in each sample.
The differences in expenditure priorities between the three sampling
areas can be discovered by looking closely at a given category of between-
group comparisons. For this purpose we have chosen category III in Table
XVII i.e. households moving between groups 4-6 and 6-8 per week. For
this category of comparisons the arc-elasticities of demand for Fruit and
Vegetables and Oils and Fats are lower in the Rural than in either of the
Urban samples. Proportionate expenditure on these items is at the same time
higher among the Rural households. The Rural household is in effect, relative
to its Urban counterpart, saturated with Fruit, Vegetables and Edible Oils,
and consequently places less importance on purchases of these items as
income increases. The same applies to a less extent in the case of Sugar,
Sweets and Condiments. The situation is different in the case of expenditure
on Milk Products and Eggs, Cereals and Bakery Products, Both these items
are "luxuries" among middle-income Rural households, in spite of the fact
that both absolute and proportionate outlay on these items are highest in the
Rural areas. The explanation here is quite probably to be found in genuine
differences in "tastes" between the different localities.
Neither the arc-elasticity of demand for, nor proportionate outlay on, Meat
and Fish varies significantly between the three sampling areas for households
in category III.
Among the non-Food items, Main Towns' middle-income households place
greater importance on purchases of Durables, Clothing, Personal and Medical
Care than either of their Rural or Kingston counterparts. This feature is
accompanied by, and partly derives from, generally lower absolute and pro-
portionate expenditures on these items in the Main Towns. Quite the reverse
applies to expenditure on Tobacco and Housing: expenditures on both these
items are more important among the Kingston and Rural households, although
only in the case of Housing does the arc-elasticity appear to be significantly
affected by differences in proportionate outlay. (For the income groups in

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs