Preview

INTERNATIONAL LABOUR OFFICE

STUDIES AND REPORTS
Series N (Statistics) No. 24

INTERNATIONAL COMPARISONS
OF FOOD COSTS
by
ROBERT MORSE WOODBURY

MONTREAL
19 4 1

PREFACE

The comparison of real wages in different countries to set forth
the relative economic positions of wage earners under different conditions of production and consumption is a fascinating study but
one beset with difficult and intricate problems. On the one hand,
the problems of securing data on wages, taking into account the
many differences in methods of collection and the differences in the
significance of wage rates, earnings, minimum and normal wages
must be faced and the mass of data for different occupations and
industries converted into manageable and useful figures. On the
other hand, a means of comparing the cost of living of wage earners
in different countries must be found in order to give a valid picture
of relative standards of comfort. Even this latter problem, relatively the simpler of the two, is itself beset with difficulties. Analysing
the cost of living into its component parts, food, rent, fuel and light,
clothing, and miscellaneous items, and studying each part separately, has seemed to offer the best means of reaching a satisfactory
solution. The International Labour Office has already published
special reports on methods of international comparisons of food and
of rent, and has made annual surveys of retail prices of certain foodstuffs, etc., as well as occasional surveys of rents in selected cities.
The present report is a further study of the specific problem of
international comparisons of food costs. It was prepared by Robert
Morse Woodbury, assisted by Walter Kull of the Statistical Section
of the International Labour Office. Prior to the emergency situation
caused by the war, the work was carried on under the general direction of J. W. Nixon, Chief of the Statistical Section; a memorandum embodying the principal elements of the report was sent to the
members of the Committee of Statistical Experts of the International Labour Office for their comments and suggestions. A
meeting of the Committee planned for the spring of 1940 had to be
abandoned. As a result of changes produced by the war it has been
decided to issue the document as a research report in order that the
2

PREFACE

various countries and interested persons may have an opportunity
to consider and test the proposed method. It is believed that the
method presented herewith represents a definite advance over
previous methods of comparisons and that with progress in obtaining valuable results for other elements of the problem, including
that of combining the ratios obtained for these different elements,
the goal of international comparisons of real wages may be brought
nearer.
Acknowledgments are due for valuable suggestions and criticisms to members of the Committee of Statistical Experts, to
E. F. Penrose and Robert Guye of the International Labour Office,
and to Faith M. Williams and Alice C. Hanson of the United States
Bureau of Labor Statistics. Other acknowledgments to authors and
writers on the subject of international comparisons of food costs
whose works have been consulted are made in the report itself.

CONTENTS

PREFACE

INTBODUCTION

Price Ratio Formulae
Method of Regional Baskets and the Problem of Regions
CHAPTER I. The Group Basket Method
Formation of Groups
T h e Group Baskets
Method of Calculating Ratios
Results
,
Correction Factors
Errors of the Group Basket Method
CHAPTER II.

Applications

27
29
30
35
38
38
50
55

Substitutions

Types of Substitutions
Criterion for Substitutions
Basis for Equivalence
Mathematical Methods
Practical Techniques
Substitutions and t h e Group Basket Method
Results
Scope of Substitutions
Conclusions
C H A P T E R V.

Conclusions

58
58
59
61
63
64
65
68
69
71

APPENDIX I. Résolutions adopted by International Conferences, the Committee of Statistical Experts and the Governing Body of the International
Labour Oßce
APPENDIX II.

15
15
20
21
22
23
26
27

Cities
Population Groups
Eponomic Groups :
Climatic Differences
CHAPTER I I I . Errors in Basic Data
Quantities
Prices
Exchange Rates
C H A P T E R IV.

3
7

Tables

73
79

1. Correction factors t o P 0 ' and P i ' by countries
2. Values of P„' after application of correction factors, calculated by
the Group Basket Method, based upon 18 commodities in 19 countries,
prices as of October 1938
3. Values of P i ' after application of correction factors, calculated by
the Group Basket Method, based upon 18 commodities in 19 countries,
prices as of October 1938
4. Values of the substitution ratio for P 0 ', with substitutions between
rye and wheaten bread
5. Values of the substitution ratio for P i ' , with substitutions between
rye and wheaten bread
2 4

79
80
81
82
83

INTERNATIONAL COMPARISONS
OF FOOD COSTS

INTRODUCTION

This Report presents the results of studies made by the Statistical Section of the International Labour Office in continuation of
its work on the problem of international comparisons of cost of
living. 1 Preliminary results were submitted to the members of the
Committee of Statistical Experts of the International Labour Office
by mail for criticisms and suggestions ; it was originally planned to
call a meeting of the Committee to discuss them in the spring of
1940. Unfortunately, owing to war conditions, this project of a
meeting of the Committee had to be abandoned. However, various
members of the Committee sent in comments and suggestions, and
these have been utilised in preparing this report for publication.
The report is therefore presented as a research report on the methodology of international comparisons of costs of living, in continuation
of the work of the Office on this topic.
The present publication is limited to the subject of the international comparison of costs of food.
The International Labour Office has already published a number of studies on the problem of comparative food costs. Its first
work on the subject was undertaken following a recommendation
by the First International Conference of Labour Statisticians in
1923 that the Office continue and develop statistics of relative wage
rates and cost of living in different countries, begun by the British
Ministry of Labour. In the British compilation the British worker's
food consumption was adopted as the basis for calculating relative
food costs in different cities in selected countries; for purposes of
1

INTERNATIONAL LABOUR OFFICE: International Comparisons of Cost of

Living. A Study of Certain Problems connected mth the Making of Index Numbers
of Food Costs and of Rents, Studies and Reports, Series N, Statistics, No. 20,1934.
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

international comparison this basis was replaced in the Office study,
published in 1924, by a series of food budgets for different groups of
countries or regions, seven in number; these were priced in each of
a series of cities, the ratios of the cost of each regional budget in
each city to that in London were computed, and finally these relative ratios were averaged to secure a single figure for each city.1
The basis for the regional grouping was stated as "the rough correspondence of quantities consumed" or "an approximate similarity
in the quantities of the chief articles consumed".2 In general, the
quantities were based on the budgets used in cost-of-living statistics
of the different countries. Subsequently, the Office set up a single
international food basket representing average quantities consumed
per consumption unit in workers' families in the different countries
and priced this basket in each country in order to show relative
costs. The international basket as originally set up in October 1929
contained 25 articles3; in 1933 the number of articles was reduced
to 14—omitting 11 relatively unimportant items covering only
about 15 per cent, of the value of the 25 articles—in order to avoid
problems arising where the price series did not cover all articles4; and
the costs of this revised basket have been published each year.
In addition, special mention should be made of two studies,
one embodying the results of the enquiry undertaken to ascertain
the cost of living in various European cities on the same standards
as that of employees of the Ford factories in Detroit5; and the
second, embodying a study of the theoretical problem of measuring
the relative costs of food and rent.6 In the first of these the geometrical average of the Laspeyres and Paasche indices, the so-called
Fisher "ideal" formula, as explained in the next paragraph, was used
between Detroit and each other city ; the second included a proposal
to set up "regions of good comparability" in order that accurate
price ratios might be calculated between countries or cities within
each such region. In connection with these various reports and
1
International Labour Review, "A Comparison of t h e Levels of Real Wages in
Certain Capital Cities", Volume X , No. 4, Oct. 1924, p p . 630-652, especially
638 et seq.
' Ibid., p. 639.
' International Labour Review, "Comparison of Real Wages in Various Countries", Vol. X X , No. 4, Oct. 1929, pp. 580-588.
4
International Labour Review, An International Comparison of t h e Retail
Prices of Certain Important Foodstuffs, July 1929 t o October 1932", Vol. X X V I I ,
April 1933, pp. 530-538.

' INTERNATION AI. LABOUR O F F I C E : A Contribution

to the Study of

Internal

tional Comparisons of Costs of Living; on Enquiry into the Cost of Living of Certain
Groups of Workers in Detroit, U.S.A. and 14 European Towns, Studies and Reports,
Series N , Statistics, No. 17, 1932.
•INTERNATIONAL LABOUR O F F I C E :

International

Comparisons

of Cost of

Living. A Study of Certain Problems connected with the Making of Index Numbers
of Food Costs and of Rents, Studies and Reports, Series N , Statistics, No. 20, 1934.

INTRODUCTION

studies, furthermore, the recommendations of the International
Conferences of Labour Statisticians should be mentioned 1 , as well
as the work of the International Labour Office Committee of
Statistical Experts. 2
PRICE RATIO

FORMULAE

The formula for an index number of prices starts from the concept that if the cost of identical quantities of a series of commodities
is ascertained in two countries, prices being expressed in terms of a
common currency, the ratio between the two costs is the index or
average price-ratio desired. If the quantities are represented by
q's and the prices by p's, one country designated as the "base"
country, being represented by subscript 0, and the second country
by the subscript 1, and the symbol 2 represents the sum, this
formula, known as Laspeyres', may be given as:
p „
0

P!g°

2 PoQo

Using the quantities of the second country as weights, another
formula, known as "Paasche's formula", may be set up,
p
1

^Ptfi
2 p0qi

The classic formula for the price index ratio is the geometric
average between those two ratios, the "ideal" formula of Professor
Fisher:

V 2 poÇo S PoÇx
This formula gives the price ratio between two countries only.
Obviously, if price ratios are to be ascertained for each pair of
countries in a series of many countries, a considerable amount of
calculation is required, depending on the number of countries. 3
'INTERNATIONAL LABOUR OFFICE:

The International Standardisation of

Labour Statistics, Studies and Reports, Series N, Statistics, No. 19, 1934, pp. 2123, 32,
57-60.
2
A detailed review of these reports and recommendations, though of great
interest in a study of the whole problem of international comparisons of cost of
living and of real wages, is not attempted here. See Appendix.
• For n countries the number of possible combinations is ••••" . For 19
countries—the figures given later cover 19 countries — there are 171 combinations of countries taken two at a time, or 171 pairs of countries between which
price ratios are desired.
24

INTERNATIONAL COMPARISONS OF FOOD COSTS

The device of averaging 2 indices, each of which has equal
claim to be considered accurate, in order to obtain the best single
value is obvious enough.
The procedure of averaging has this further justification.
Since the base for P 0 is the set of average quantities consumed in
country 0, this value may itself be considered an average: and if
detailed quantity baskets are available for each income or social
class or even for each family, a different value of the price ratio
would be obtained for each such quantity basket; in other words,
P 0 expresses merely an average result. Similarly Pi represents an
average price ratio based upon average quantities consumed in the
second country. An average of these average results for P 0 and Pl
would appear to be a logical extension to obtain a best single value
for the price ratio between the two countries.
In the usual interpretation, however, these values are considered to be limiting values. The price ratio, P 0 , allows the average
dweller of country 0 a sufficient sum to enable him to buy in the
second country the identical quantities of the articles consumed by
him in the base country. But he may be able to obtain in the second
country, for less money than he spent in the base country, a budget
of goods that gives him his standard of food consumption—perhaps
not precisely the individual quantities of the identical items—which
will give him satisfactions equal to those obtained from his customary budget of goods in the base country. If an equal standard of
food consumption or equal satisfactions are used as a test for budget
equivalence rather than identity of quantities in the two budgets,
then P 0 is a maximum for the ratio based on the quantities of
country 0, since a budget costing less money, and giving therefore a
smaller ratio, may give equal satisfaction in country 1 to that
obtained by the budget of the base country. Similarly, if the budget
of country 1 be used as the base, 2 p0qÍ7 (the denominator in the
fraction P t ) is likewise a maximum and a smaller sum may give
equal satisfaction in country 0 to that given by the commodities in
country 1. In this case, Pi must be regarded as a minimum limiting
value, since the denominator may be replaced by a different and
smaller sum that still yields equal satisfactions in both numerator
and denominator.
If this reasoning is adopted, the further step may be taken.
Letting the smallest value of P 0 and the largest value of Pi that
give equal satisfaction in numerators and denominators in the price
ratio fractions respectively be regarded as the true values—in other
words, assuming for example that the true price ratio from the point
of view of country 0 is represented by the smallest ratio which will
satisfy the test of equal satisfaction in the costs of the commodities

INTRODUCTION

consumed—and letting the difference between P 0 and the corresponding true value be represented by d0 and between P t and the
corresponding true value by di, then the two true values are P0—do
and Pi+di. With these two true values, the procedure of averaging
to obtain the best single value appears equally appropriate. In
order to obtain usable results, in the absence of direct evidence
some assumptions as to the relative value of d0 and di may be
made. Since neither can be negative, they tend to offset each other;
if do equals d t —the simplest hypothesis—the average of the two
true values is the same as the average of P 0 and Pi. 1
Another type of average ratio, which uses as weights the
arithmetical average of the quantities, is
2 Pi (go + gi)
~ 2 Po (go + gi)
In practice this does not differ materially from the result
obtained by the formula P = \/p0pt at the range of values commonly
found in price ratio comparisons. In a series of tests, for example,
the formula using the arithmetical average of the quantities between two countries were found to give results differing only very
slightly from those of the Fisher formula; in 5 cases, for example, the
largest divergence was only 0.5 per cent, the average difference
being only 0.2 per cent.
For the practical purpose of computing the price ratio between
the two countries, therefore, there is little to choose between the two
formulae, and the choice may be influenced by such practical considerations as ease of computation, ease of manipulation, and other
points. 2
This formula is of special interest, since if rates are to be calculated for a number of countries, a simple extension gives, for the
price ratio between country 1 and country 0:
2 Vl (go + gl + gü + - gn-l)
/oi =

2 Po (go + gl + g2 + ... gn-l)

If d is the average percentage correction and if d—d*¡=d¡,
(1-ri)
2 p. g.
/ (1-d) 2 Pig. Zp.g", _ . . / I T T r
=
v
\
2 R Î ,
S p. 9, (1-d)
\ ( l - d ) 2 p „ ? „ SPog>
^ ^ '
The calculation of the effect of substitutions may be regarded as an attempt to
determine do and di. See Chapter IV.
/=

A/2Pig.

Bowley offers mathematical proof that the formula •„ ' / ' .—r is the
¿j Po wo "T gl/

best one to take account of the differences in the quantities consumed in the two
countries. See A. L. BOWLEY, "Notes on Index Numbers", in Economic Journal,
Vol. XXXVIII, 1928, pp. 216-237.
2u

INTERNATIONAL COMPARISONS OF FOOD COSTS

the formula using as weights for the prices the quantities of the
international basket if the q'a cover all the countries in the comparison, or of the regional basket if the q'a cover the countries in a
particular region or group.
This formula with the quantities of an international basket,
i.e., the average of the quantities in the food baskets of all the countries in the comparison, has often been used in making international
comparisons. It is analogous to the formula commonly used for
price index changes over a period of time, the average quantities
over the period being used as fixed weights. It has the great advantage of simplicity, since the ratio of the costs of the international
basket in any two countries (in terms of a common price unit, e.g.,
the Swiss franc, the U.S. dollar) yields the desired price ratio.
The formula just described, however, has the obvious defect
that in a comparison between two countries, e.g., country 0 and
country 1, the quantities taken into account include not only the
quantities q0 and qi, for the two countries concerned, but also
?2. Î3 up to çn.i, quantities for other countries. The advantage
of having quantities that do not change from country to country as
the countries directly compared change, is countered by the disadvantage that the quantities used as weights do not correspond
exactly to those of any pair of countries, but differ from these by
larger or smaller amounts.
The importance of this disadvantage depends in practice, other
things being equal, upon the divergences between the international
basket and the local baskets. The question may be raised how much
of a difference is to be found in practice between the final ratios
obtained by means of the international basket method and those
reached by a direct calculation of P = y/pQpi between each pair of
countries ? To test this point, average quantities for a new international basket, including 18 commodities, were calculated for the
19 countries for which satisfactory price and quantity (family
budget) data were available.1 The ratios of food costs thus obtained
were then compared with the ratios calculated by using the Fisher
formula. The differences between the results of the two series, based
on some 70 comparisons—about 2/5ths of the possible combinations
—averaged 5 per cent., while in the extreme case, the difference rose
to 20 per cent. If the "direct" method—the geometric average of
1
Quantity data were taken from the tables presented in "An International
Survey of Recent Family Living Studies: II. Food Expenditure and Consumption Habits", in International Labour Review, Vol. XXXIX, June 1939, p. 81.
All the countries covered by that article, except Colombia, are included in the
international food basket, and France, Great Britain, and Italy are added. See
also Robert M. WOODBURY, "International Comparisons of Food Costs", in
International Labour Review, Vol. XLI, Feb. 1941, p. 160, footnote 14.

INTRODUCTION

P 0 a nd Pi—between each pair of countries, gives a close approximation to the true cost-of-food ratios between them, the international
basket method is thus subject to considerable error.
Of these two methods, then, the one gives satisfactory results
but involves extensive calculations where comparisons between
many countries are involved, while the other, though requiring
little work, shows large divergences from the true results.

METHOD OF REGIONAL BASKETS AND THE PROBLEM OF REGIONS

The possibility of an intermediate method having the advantages of both and giving substantially accurate results is suggested
by previous work on regional baskets.1
So far as the formula is concerned it is evident that a regional
basket formula
T

Pi (g» + gl + - g»)
~ 2 Po (go + gì + - qù

will approach closer to
Z pi (go + gi)
~ 2 po (go + gi)
the more nearly the average quantities consumed in countries included in the region approach the average of the two countries concerned in the price ratio between them. It is equally clear that the
regional basket formula will give better results than the international basket formula if the region includes only countries with
closely similar food consumption.
In the historical development of the subject it was suggested
that a satisfactory price ratio between any two countries of the same
region could be obtained if regions were established on the basis of
"good comparability". Though the term "regions of good comparability" presents an alluring prospect, the problem of criteria for
determining the regions offered difficulties.2 Furthermore, the
problem of comparisons between the cities and countries in different
regions or outside regions was left in an unsatisfactory state; the
very term used to describe the regions tended to suggest that all
other ratios were vitiated by "bad comparability", and the specific
1

See INTERNATIONAL LABOUR OFFICE: International Comparisons of Cost of

Living. A Study of Certain Problems Connected with the Making of Index Numbers
of Food Costs and of Rents, Studies and Reports, Series N, No. 20.
' Ibid., pp. 31-59.

INTERNATIONAL COMPARISONS OF FOOD COSTS

suggestion that chain relatives be used to construct ratios between
cities in different regions1 was subject to the serious disadvantage
that the results varied according to what cities were used as intermediate steps in the chain.
The first task in the further analysis of the problem, therefore,
was to examine the possible criteria for establishing regions. These
include similarity in geographical elements and various tests involving similarity of food consumption.
Similarity in Geographical Elements.
Similarity in geographical elements including population,
climate, language, etc., though an obvious and natural test, for
present purposes seems vague and unsatisfactory. On this basis,
for example, the' Scandinavian countries of Northern Europe might
form a single group, those of Western Europe another, including, e.g.,
France, Belgium, and the Netherlands, those of Central Europe a
third, including Germany, Austria, Czecho-Slovakia, and Switzerland, those of East Central Europe a fourth, including, e.g., Poland
and Hungary, and those of Southern Europe a fifth, including
Italy, Bulgaria, etc. Unfortunately the allocation of countries like
the Netherlands, Switzerland, or Czecho-Slovakia appears arbitrary
as between two contiguous groups, since the test does not specify
which of the many elements, for example, nationality, language,
traditions, income, food habits, is to be accepted as decisive in
assigning a country to one or another group.
Similarity in Food Consumption.
For purposes of establishing regions for comparing food costs
a criterion based on similarities of food consumption seems both
logical and appropriate. Four tests were examined.
A simple test of the level of food consumption can be found by
comparing the quantities consumed in each country with a set of
standard quantities. The quantities of the international basket
give a suitable standard of reference. A simple method of combining
the results for different commodities is to use their local prices.
Following this plan, table I shows, an index of the level of food
consumption as the percentage which the local consumption forms
of the average consumption of 18 commodities of the international
' Ibid., pp. 6-7, 9-10, 69-71.

INTRODUCTION

basket priced in each country at local prices. This shows how representative the international basket is of the local food consumption.
On the basis of this index, Italy and Bulgaria appear at the
lower end of the scale, not far from Poland and Hungary. Germany,
Czecho-Slovakia, Austria, and Switzerland occupy a position in the
middle of the range. Finland, Norway, and Sweden are close together well towards the upper end of the scale, with Denmark somewhat higher but not far distant. Countries with unusual positions
(as compared with other indices) include Great Britain and the
United States, which are both very low, and Belgium, which is
unusually high.
TABLE I. PERCENTAGE THAT THE COST AT LOCAL PRICES OF THE
QUANTITIES CONSUMED LOCALLY BEARS TO THE COST AT LOCAL
PRICES OF THE QUANTITIES IN THE INTERNATIONAL BASKET
(19 COUNTRIES) OF THE COMMODITIES INCLUDED IN THE
INTERNATIONAL BASKET
Country
Italy

Per cent.

Country

Per cent.

50.8
68.6
74.8
76.5
78.0

101.8
107.2
108.4
111.0
111.9

78.7
80.8
93.6
95.5
96.8

112.1
123.1
123.6
142.7

Union of South Africa

An index of the quality of food consumption may be made by
comparing the relative consumption of the so-called "protective"
foods and the energy-giving foods. Table II illustrates three indices
of this type. In the first column the countries are ranked according
to the percentage which the "protective foods", together with
meats and fish, bear to the sum of these plus the staple foods (all
measured in kilograms); in the second column the countries are
ranked according to the kilograms consumed per consumption
unit of the protective foods, together with meats and fish; while the
third gives an average ranking in which double weight is assigned
to the first index. In support of these types of index it may be
pointed out that the staple foods tend to predominate in budgets
where the strictest economy prevails, while relatively much larger
proportions of the protective foods, together with meats and fish,
are consumed in the upper income classes. The increase in the proportion of the protective foods may therefore be considered to
correspond roughly with an increase in the level of living. Likewise
2 ¿

INTERNATIONAL

COMPARISONS OF POOD

COSTS

the larger quantities of these protective foods consumed per unit
would correspond to higher levels. A combination of the two indices in which the first criterion is given double weight attempts to
utilise both these indications.
TABLE II. RANK OF COUNTRIES ACCORDING TO THREE CRITERIA
OF LEVEL OF FOOD CONSUMPTION
•

• • '

I. Percentage of protective
foods plus meats and fish
to total including
staple
foods1

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.

Switzerland
United States
Finland
Sweden
Austria
France
Norway
Union of S.Africa
Denmark
Great Britain
Netherlands
Czecho-Slovakia
Germany
Hungary
Estonia
Belgium
Italy
Poland
Bulgaria

68
61
59
59
58
56
54
53
51
50
48
48
47
44
40
39
37
31
29

II. Kilograms per consumption unit of protective foods
and meat and fish

i.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.

Switzerland
Finland
Sweden
Netherlands
Norway
Austria
Estonia
Germany
Belgium
South Africa
Czecho-Slovakia
Denmark
France
United States
Hungary
Poland
Great Britain
Bulgaria
Italy

483
416
410
363
357
343
318
303
298
297
286
276
274
267
221
205
192
137
126

II
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.

. Weighted average
rank of I and II*

Switzerland
Finland
Sweden
Austria
United States
Norway
France
Union of S. Africa
Netherlands
Denmark
Belgium
Germany
Czecho-Slovakia
Estonia
Hungary
Great Britain
Poland
Italy
Bulgaria

The group of "protective foods, etc." includes milk, milk products, eggs, fresh vegetables
and fruits, and meats and fish; the group of staple foods comprises bread and cereals,
peas and beans, potatoes, sugar and fats.
s The rank of the first column is given twice the weight of that in the second column.

In this series of indices Finland and Sweden, together with
Switzerland, have the three highest ranks, followed after an interval by Norway and Denmark. The United States has a high
position. The countries of Central Europe, Germany, Czechoslovakia, and Hungary are close together in a middle position considerably below the positions of Austria and Switzerland. Italy,
Bulgaria, and Poland occupy positions towards the bottom of the
scale, somewhat below Estonia. Great Britain is well down on the
list.
Similarity in Quantities Consumed.
An index of quantities consumed can be formed in a similar
way to the price ratio P, namely, as the square root of Q0 and Qit as
follows:
Q =

Pogl

V 2 po9o

Piqi

S piqo

INTRODUCTION

If Q is equal or nearly equal to 1, the quantities in the two
countries compared may be considered roughly equal or at least
similar. Table I I I shows the value of Q for a number of pairs
of countries.
TABLE III.

VALUES OF THE QUANTITY RATIO, Q, FOR SELECTED
COMPARISONS1

Region and country

I. Western Europe
Belgium/France
Netherlands/France
Netherlands/Belgium

1.61
1.23
.73

II. Central Europe
1.02
Austria/Germany
Switzerland/Germany, ... 1.06
Czecho-Slovakia/Germany .93
1.04
Switzerland/Austria
Czecho-Slovakia/Austria..
.93
Czecho-Slovakia/Switzer.89
Q = -.¡2 V» g'

Region and country
III. East Central Europe
Estonia/Poland.
Hungary /Poland
Hungary/Estonia. ..
IV. Scandinavian countries
Nor way/Sweden....
Denmark/Sweden...
Denmark/Norway...
Finland/Sweden
Finland/Norway....
Finland/Denmark...

Q
1.05
.81
.76
.96
1.00
1.07
.92
1.00
.97

P. Q,

v 2 Po 9o S Pi Î«

Quantity indices equal or nearly equal to 1 are found between
Denmark and Sweden, Finland and Norway, Finland and Denmark.
For Norway/Sweden the index is .96; for Denmark/Norway 1.07;
and for Finland/Sweden .92. If regions were formed on the basis of
this criterion these Scandinavian countries would appear to constitute a fairly satisfactory region. The index for Estonia/Poland
is 1.05, for Hungary-Poland .81, for Hungary/Estonia .76. If this
criterion were to be adopted it would obviously be necessary to
establish the appropriate limits within which departures from unity
could be permitted for countries in the same region.
Similarity in Proportionate Quantities Consumed.
Another index based upon differences in proportionate quantities consumed is Staehle's index, which varies from zero for perfect similarity to 2 for complete dissimilarity. 1
K = 2

PoÇi

Po9o

Spoîi

SpoÇo

1
See Hans STAEHLE, "A General Method for the Comparison of the Prices of
Living", in Review of Economic Studies, Vol. IV, No. 3, June 1937, p. 206.

2 U

INTERNATIONAL COMPARISONS OF POOD COSTS

This index is the sum (taken without regard to sign) of the differences in value weights (proportions) applicable to each commodity
in the two countries compared, using however the same prices, i.e.,
those of the base country, to calculate the value weights.
This index differs from Q in that it takes into account not the
differences in the quantities consumed but the differences in proportionate consumption. Thus, if in one country twice as much of
each item is consumed as in the other, the index Q will give a value
of 2 corresponding to the fact that the consumption of one country
is twice that of the other, while Staehle's index would give an index
of zero, i.e., perfect similarity.
The use of Staehle's index would require a considerable volume
of calculations, since the value would have to be determined between each pair of countries—with 19 countries, 171 combinations
of countries taken two at a time. Also, some decision would be
needed on the criterion or limit for excluding a country from a
region. The index furthermore varies between income classes.
Thus, Staehle gives figures for Poland and Estonia comparing different income groups, showing a variation from .22 to .82 for different combinations of income.1
A general remark is applicable to all these specific criteria of
regions, namely, that all the indices assume that the fundamental
data on quantities are accurate. In certain cases, however, as suggested perhaps by the differences in rank of a particular country
according to different indices, this assumption may be questioned.
Thus, for example, the low position of Great Britain in the first
test illustrated may be due to the character of the source data. A
test based upon proportions, however, is not subject to the same
errors as a test based upon absolute quantities.
1
Two other tests are directed to the question of t h e applicability or appropriateness of a comparison of price ratios. These include t h e size of the differences
between P» and P i and the percentage of residual articles. I n regard to the first,
while it is true t h a t if the quantities consumed in two countries are identical Po
and P i coincide, it does not follow that if Po and P i coincide or are close together
the quantities of t h e two countries are closely similar. T h e criterion D =
(Pi — Po) / Po was developed at length in the study made by the INTERNATIONAL
LABOUR O F F I C E : International Comparisons of Costs of Living, Studies and R e ports, Series N , No. 20, pp. 18-19. Among t h e conditions of good comparability
it is stated t h a t "D should have relatively low value". I t is pointed out, however,
t h a t even though D may be very small in certain circumstances comparability
may be bad. This criterion is therefore not further developed here.
The percentage of residual articles is a test of the applicability and significance of the price ratio technique. If none of the articles in the two countries are
alike, t h e percentage of the residual articles is 100 and any price comparison is
impossible. The lower the percentage of residual articles t h e better t h e price
ratio reflects actual conditions and the less scope is allowed for t h e possibility
t h a t other prices not covered in the ratio are different from those included, as
well as for the possibility t h a t some commodities are entirely different, with t h e
result t h a t no price ratios covering them can be calculated.

INTRODUCTION

A survey of the results of these different tests shows that some
method for testing which is the best grouping of countries is necessary. For this purpose the value of P = y/pQpx calculated by the
V TI

( rt

Vl Kq

rt \

° **"g , may be used. Though it is not
2 po (ïo + Ci)
yet proved that either of these test values is a correct final result, at
least they do not favour one grouping as against another. It may
be said further that, if the test of the grouping is confined to intragroup ratios, a logical test would be to compare them with ratios
based upon the data for two countries only and, as already noted,
direct method, or

the alternative values, namely P and —
— are substan2 po (îo + ?i)
tially equal. For the most part objections to these test values as the
true value of the price ratio between two countries would apply
equally to the regional basket formulae. Hence the use of P as a
test for the regions based upon regional basket formulae seems
justified.
The results of certain of the groupings tentatively made show
errors in intra-group ratios of a relatively small amount as compared
with the international basket figures. Without proceeding further
at this point to present the results of all possible groupings of countries to form regions, certain generalisations may be drawn from
this discussion, and the method finally adopted to determine the
grouping, namely, the group basket method, together with its
results, will be discussed in detail in the following chapter. The
following general conclusions may be drawn. First, many of the
criteria under review appear to suggest the same or closely similar
groupings of countries. In some cases there appears to be little
to choose between alternative allocations of certain countries. For
purposes of grouping countries to yield food price ratios a grouping
based upon the data of food consumption appears logically preferable to one based upon other criteria. Proportional quantities offer
a better basis of grouping than absolute quantities, since it can be
shown that proportional rather than absolute quantities affect food
price ratios and since the use of proportions tends to avoid the consequences of certain defects in the original data.
Finally, as a consequence of the assumption that P can be
used as a test of the regional grouping, the whole problem can be
viewed from a different angle. From this point of view the use of
regions may be envisaged as a short-cut method of calculating
the value of P in n (n-l)/2 ratios in a series of n countries, by interposing a series of groups with the objective of diminishing the
volume of calculations. The problem then becomes, how can the
2 ¿

INTERNATIONAL COMPARISONS OF FOOD COSTS

countries be grouped in such a way as to reduce errors in the approximate values, P', to the minimum, and the final question is then
raised, are the results of the approximation sufficiently close to the
true results to justify the use of the approximation? With this
approach, the true criterion of grouping becomes that which reduces
the differences between the P"s and the P's to a minimum and the
problem of making comparisons between countries in different
groups is immediately resolved by using the geometric average of
the ratios formed with the respective group baskets instead of the
local baskets as the weights.1
The group basket method is described in the next chapter.
1
A proposal for comparisons within regions by means of the relative costs
of the regional basket and for comparisons between regions by means of the Fisher
formula was outlined in a brief memorandum by John Lindberg, then a member of
the Statistical Section of the Office, prepared for submission to the Fourth Conference of Labour Statisticians in May 1931.

CHAPTER I

THE GROUP BASKET METHOD
The basic formula for the food price ratio, as already discussed,
is the geometric average of P 0 and F 1 ( P 0 representing the average
food price ratio weighted by the quantities of country 0, that is,
using the food basket of country 0, and P t the average food price
ratio weighted by the quantities of country 1, that is, using the
food basket of country 1. The essential feature of the Group Basket
Method is the use of the basket of the group to which each country
is allocated as a substitute for the local basket, the group basket
being sufficiently similar to that of each of the countries in the
group to permit of its use for each local basket without impairing
the substantial accuracy of the results.
The group basket formula for the food price ratio between
country 0 and country 1 is found by substituting the group basket
for the local basket in the formula P - . Is pl q° S P l g l
\ S Po ço S pogi
If country 0 belongs to Group A and country 1 to Group B, and if
P' is used to denote the price ratio obtained by means of the group
basket method, then
p, _ p Pi QA S pi qB
V S po çA S pò ÇB
It remains to show by what method satisfactory groups can be
formed, and within what limits in practice errors are introduced by
substituting the group baskets for the local baskets.
FORMATION OF GROUPS

The basic problem then is the formation of groups the baskets
of which can be used without substantial loss of accuracy for the
respective local baskets. Obviously, if two countries, or regions or
cities have identical food consumption, the basket of one can be
substituted for that of the other without affecting the food price
ratio. Further, if the quantities in the basket of one country are
equal to those in the basket of the other times a constant factor, the
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

food consumption not being identical but having "similar proportions", the food price ratio will not be altered by the substitution of
one basket for the other. 1 Even if the baskets are not identical, if
the differences between them are not great, no large error will be
introduced into the calculation by substituting one for the other,
or by substituting for the local basket a group basket with similar
food consumption.
The principle to be adopted in establishing groups, therefore,
is similarity of proportionate food consumption. Proportions rather
than actual amounts consumed are significant, since the baskets of
two countries which have the same proportions of the different foods
though differing absolute quantities can be interchanged without
affecting food price ratios. In order to determine proportions, however, some method must be devised to eliminate the influence of
varying prices upon the proportions that different foods form of the
food budget in different countries.
In order to test practicable methods for grouping countries,
the basic data available for 18 commodities in 19 countries were
utilised. 2
Three methods for determining the grouping of countries were
tried. In the first two, a set of standard price factors was applied
to the quantities of the different foods to afford a basis for calculating proportions. In the first and simplest method, median prices
were utilised as the standard prices. All prices (October 1938 prices)
were converted into Swiss francs, at the exchange rates as of October
1938. For each food, the median price—among the 19 countries in
the series—was taken as the standard. These prices were multiplied by the quantities in the different countries, and the resulting
values were reduced to percentages devoted to the different foods
and food groups.
In the second method, the price of each food in each country
was expressed in terms of the price of bread in the same country,
and the median ratio—among the 19 countries—of the price of each
food in terms of the price of bread was taken as the standard price
factor for that food. These standard price factors were multiplied
by the actual consumption of the different foods in each country
and the result summed and reduced to a percentage basis; the
1

Let ci = aq¡, and po = pi

Then /., = J2_£l£° S ? " ? '
V S po Co S Po gJ
^ S Pi Ci S pi q>
' For basic data see chapter I I I .

a Z P» go S p « g «
V a S po go S Po g«

Itt

THE GBOTJP BASKET METHOD

result giving figures of "proportionate consumption" into which
varying prices in the different countries do not enter.
In the third method—which in fact was the first one actually
carried through—the consumption of each food in each country was
noted in terms of the range: the interval between minimum and
maximum consumption in any country in the series was divided
into scale units from 0 to 100, and the consumption of a given food
in a given country was rated to the nearest scale unit: for example,
50, if it fell half way between the lowest and highest points. These
scale ratings were then averaged for four groups of foods: bread
and cereals, meats, milk and milk products, and all other: in averaging, the ratings for the different foods were weighted in proportion
to average values. These weighted average indices for the four
groups were then summed and reduced to proportions on a scale
of 10.
In all these methods, as a final result, the food consumption of
each country is epitomised in a series of key figures adding up to
10 which show the proportionate consumption of the different foods.
In these proportionate consumption figures, the influence of price
variations in different countries is eliminated.
All these methods led to the same grouping of countries,
although the percentages themselves varied according to the particular method used. The method of allocating the countries to
groups is illustrated in table IV, which gives thefiguresof proportionate food consumption in the four food groups as calculated by the
first method—that of using median prices.
The table also gives a column showing the sum of deviations,
in tenths of a point, of the key figures for each country from the key
figures of the group in which it is placed. This furnishes a convenient
test of the allocation of countries to groups; the best allocation
should yield a minimum sum of deviations.1 With 19 groups—one
country to each group—the deviations are of course zero, and the
P"s coincide with the P's. With the number of groups fixed at less
than the number of countries, however, allocations, in principle,
can be made to groups in such a way as to yield a minimum sum of
deviations; in other words, the allocation is not dependent upon
arbitrary judgments.
So far as the number of groups is concerned, the decision depends, first upon how closely the different countries can be grouped
along definite lines: if all countries follow closely parallel lines of
food consumption, a single group—the international basket—might
1
Without regard to sign; substantially the same result is found if the "sum of
the squared deviations a minimum" is used as the test.

2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

give a satisfactory result. Secondly, the decision depends on the
balance between the accuracy of the result and the volume of calculations : the more groups the less error but the greater is the volume
of calculation. In final analysis, it depends upon the degree of error
which can be accepted as not invalidating or too greatly impairing
the value of the results.
TABLE IV. CLASSIFICATION OF COUNTRIES INTO GROUPS,
ACCORDING TO SIMILARITY OF PROPORTIONATE
CONSUMPTION
Group and
country

Bread and
cereals

Meat

Milk
products

Miscellaneous

S Idi1

Group I
Bulgaria
Italy
Group II
Estonia
Hungary
Poland
Group III
Belgium
Czecho-Slovakia....
Germany
Netherlands
Norway
Group IV
Austria
Denmark
France
Great Britain
Union of South Africa
Group V
United States
Group VI
Finland
Sweden
Switzerland

5.0
4.9
5.1

2.0
2.5
1.6

2.0
1.5
2.5

2.5
1.3
1.2

3.6
3.4
3.6
3.8

1.9
1.7
1.7
2.2

2.5
2.5
2.0

1.0
1.2
.8
2.2
2.3
2.2
2.0

2.1
.7
A
1.0

2.2
1.8
2.5
2.4
3.3

4.1
.5
1.3
.5
.1
1.7

2.0
1.7
3.2
1.7
1.8
1.6

5.8
1.1
2.4
.8
.3
1.2

3.8
2.3
2.5

3.8

1.1

3.7

2.0
2.3
1.4
2.4
2.0
1.9
1.3

2.3

4.2

2.2

5.1
5.0
4.8
5.4

2.1
2.1
2.6
1.6

Sum of the deviations from the group average taken without regard to sign.

In practice the test used, namely, the sum of the deviations a
minimum, does not correlate perfectly with the true criterion,
which is that the sum of the differences between the P"s and the
P's must be a minimum. In the true criterion, the correlation
between the deviations of the q's and the deviations of the p's from
their respective averages is a vital element in determining how
closely the P"s approach the P's, whereas in the simple test adopted

THE GBOUP BASKET METHOD

here for determining the grouping of countries this correlation finds
no expression. Furthermore, as developed above, the test has been
simplified to show deviations from the key figures for four groups of
foods only, whereas a more detailed test would take into account
each of the 18 food items. On the other hand, in so far as two or
more items have the same or nearly the same price ratio, grouping
them together gives a better test than if each one is taken separately,
since the final price ratio is affected by the variation of this total
quantity associated with a given price ratio and variations in the
components of this total quantity cancel out.
The allocation to groups as made in table IV seems reasonable
enough, with the exception of Denmark, for which the sum of deviations from the average of Group IV is unusually high. In fact, the
sum of deviations would be slightly smaller if Denmark were
assigned to Group V with the United States, or if it were placed
alone in Group V, the United States being assigned to Group IV.
The allocation to Group V was made on the basis of method III,
(which was the method first tried) which indicated a much better
result with Denmark assigned to Group IV. The method of allocation is perhaps not sufficiently precise in any case to warrant too
great attention to small differences in the criterion. An examination
in more detail of the deviations when all the items instead of the
four groups only were used indicated that the allocation of Denmark
to Group IV was slightly better. A final test by comparing P"a
with P'a showed that the United States alone in Group V gave
slightly better results than with both Denmark and the United
States in Group V, perhaps partly because with a single country
alone in a group, all the P 0 ' s involving that country are correct,
and the deviations are limited to the Pi ratios alone.
Attention should be called to the fact that the allocations are
made on the basis of the 18 commodities. If additional commodities
are included the proportions, for example of fruits and vegetables,
might be increased more in some countries than in others, and a
new grouping based on more complete data might well differ in some
respects from that adopted here. In view of this possibility, too much
weight should not be given to the present allocation, since it is
bound up with the limitation to 18 commodities and cannot be
considered necessarily as a grouping that would be obtained if full
data were available. The conclusion is nevertheless justified that
with a given set of quantities, a definite allocation of countries to
their groups can be made and for practical purposes the simple test
here used, based on four food groups with the sum of deviations
from the group averages a minimum, appears to yield a satisfactory
result.
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

Once the grouping has been established, the method of diagnosis
adopted has no effect whatever upon the price ratios. The ratios
depend solely upon the quantities of the group baskets and the
prices in the different countries. This point deserves special attention, since objections to and criticisms of the details of the method
of diagnosis do not affect the final results.
On the basis of the figures of table IV, the characterisation of
the groups by the method of diagnosis in terms of the key figures
for the four food groups is as follows :
Group

Bread and
cereals

Meat

Milk, etc.

I.
II.
III.
IV.
V.
VI.

5.0
3.6
2.2
2.0
1.3
1.6

2.0
1.9
1.5
2.5
2.3
1.3

2.0
2.3
3.8
3.6
4.2
5.1

Miscellaneous
1.0
2.2
2.4
2.0
2.2
2.1

The groups are arranged in general in order of descending
importance for the bread and cereals group and in order of ascending
importance for milk and milk products. Group III, however, has a
larger weighting of milk and milk products than IV. Group VI
has a slightly higher consumption of bread and cereals than Group
V, but is chiefly characterised by the very large consumption of
milk and milk products. Group IV has the largest weighting of
meats.
The difficulty of allocating certain countries to the appropriate
group is, in essence, due to the fact that in some cases they occupy
intermediate positions in the relative consumption of different
foods.

T H E GROUP BASKETS

The grouping of the countries having been established, a group
basket is calculated for each group by averaging the quantities
consumed per unit in the different countries in the group. 1
These quantities are given in table V for the six group baskets ;
covering the 18 commodities for which adequate price and consumption data are available. 2
* An alternative procedure of averaging not the quantities but the percentages
of total value (as found by multiplying quantities by the standard prices) did not
prove on the basis of a test to give superior final results to those reached by use of
simple average of the quantities.
1
Details of sources, etc., given in chapter III.

THE GROUP BASKET METHOD

T A B L E V.

. . . . ..... _.. ._
Item

THE S I X GKOUP B A S K E T S 1

Group

Group
II

Group
III

Group
IV

Group
V

Group
VI

170.8
33.6
5.1
3.2

42.9
115.3
31.5
4.3
0.6

5.4
8.6
5.0
1.3

10.5
8.0
2.9
1.3

56.0
31.6
26.6
5.2
1.1
15.5
9.9
4.2
5.3

40.2
7.5
16.3
6.7

Beef
Pork
Mutton
Veal

76.5
55.0
20.1
2.4
1.1
12.9
7.4
1.0
1.8

24.3
6.9
6.4
4.4

35.0
20.7
50.3
3.8
3.4
8.2
9.6
0.2
3.0

Lard
Margarine

2.8

3.7
1.1

2.0
10.7

2.3
4.9

6.5
0.0

0.9
5.6

Milk (litres)
Cheese
Butter
Eggs (no.)

28.9
5.1
0.8
88.5

172.6
5.6
7.0
166.4

110.8
4.1
8.0
169.2

Peas, beans
Potatoes
Sugar

7.6
23.7
7.4
0.4

115.9
0.7
2.7
48.0
3.7
159.9
23.2
1.1

2.7
149.4
21.4
3.6

3.6
68.6
27.3
2.9

152.1
4.4
11.9
326.0
7.4
88.9
31.6
5.5

284.5
4.8
11.7
158.0
1.5
98.4
33.2
5.1

Wheaten bread
Rye bread
Flour
Rice
Macaroni

1
Group I: Bulgaria and Italy; group II: Estonia, Hungary, and Poland; group III: Belgium,
Csecho-Slovakia, Germany, Netherlands, and Norway; group IV: Austria, Denmark, France,
Great Britain, and Union of South Africa; group V: United States; group VI: Finland, Sweden,
and Switzerland.
Quantities in kilograms, except milk (litres) and eggs (number).

METHOD OF CALCULATING RATIOS

Having the group baskets for the six groups and the price data
for the 19 countries for October 1938, the calculation of ratios according to the group basket method involves five operations.
First, the prices for each country are multiplied by each of the
group baskets, and the result summed to give for each country the
total cost, Spg, for each group basket at local prices.
Secondly, these sums are converted to a common currency
(Swiss francs) by using exchange ratios for the period to which the
prices relate.
Thirdly, P 0 ' and i Y are calculated for each pair of countries
using the group baskets appropriate to each case.1
1
Here and subsequently Po and Pi are used to denote ratios with quantity
weights of the denominator country and of the numerator country respectively;
similarly c<¡ and ct are used for correction factors to P, and P, ratios with weights
of the denominator country and of the numerator country respectively.

2 6

INTERNATIONAL COMPARISONS OF FOOD COSTS

Fourthly, P 0 ' and Pi' are averaged geometrically to give P'.
Finally, as will be explained later (pp. 23-24), correction factors
are calculated and applied to P ' to give the final corrected ratios
Though the primary concern of the present exposition is to discuss methods, the fact remains that the simplest and clearest way
to describe the method is to illustrate its application to specific
data. Accordingly, results of applying the method are presented
for the 18 commodities and 19 countries with prices as of October
1938. Incidentally, the question of the size of errors due to the
group basket method can be presented in equally concrete terms
by comparison with the results obtained by the direct calculation
of P using the same data. In this procedure, the quantities and
prices are accepted as given, and the question of errors in either
the basic quantities or prices is waived for present purposes (they
are discussed later in chapter III).
RESULTS

The results of the first two operations can be illustrated in
table VI, which gives the cost of each of the six baskets in each
country expressed as a percentage of the cost in Great Britain.
This calculation, though not required in the process of obtaining
the ratio is of rather more significance than the actual values as
expressed in Swiss francs, since the actual values are influenced
by the arbitrary levels of the quantities of the different baskets. By
expressing the results in the form of a ratio of the cost of each basket
in each country to the corresponding cost in Great Britain, the
general relationship existing between the several baskets can be
presented.
These results give also the necessary data for calculating the
desired price ratios (uncorrected data) between each pair of countries, as well as the actual costs of the baskets themselves. For any
given ratio, for example, Italy/Great Britain, P 0 ' is found from the
table as the ratio for Group IV of Italy and Great Britain (119.8),
Pi' as the ratio for Group I of Italy and Great Britain (123.6)
and P ' is then found by taking the geometric average of these two
ratios.

THE GROUP BASKET METHOD
TABLE VI. RELATIVE COST OF STANDARD BASKETS IN
NINETEEN COUNTRIES, OCTOBER 1938 1
{Great Britain = 100)
Country

Basket
I

Basket
II

Basket
III

285.3
Austria
178.9
176.1
102.6
Belgium
89.9
85.0
78.7
Bulgaria
91.9
90.9
94.7
Czecho-Slovakia... 137.8
88.5
86.2
90.8
Denmark
86.7
103.8
63.3
143.0
Estonia
63.3
91.9
83.1
Finland
84.5
217.6
75.3
79.7
France
100.0
153.1
152.6
Germany
130.0
100.0
100.0
Great Britain
123.6
109.5
124.7
122.4
Hungary
157.5
107.8
114.4
Italy
94.0
97.6
101.2
Netherlands
149.8
109.2
110.7
Norway
128.5
75.5
72.8
122.9
Poland
148.3
104.1
116.4
Sweden
105.1
111.3
Switzerland
1
Unionof
Africa
116.8
128.0 by exchange
These South
ratios are
in many cases affected
United States....
Czecho-Slovakia,
Denmark, Estonia, Germany,
and
126.9
143.5 Italy,

Basket
IV

Basket
V

183.1
99.1
91.2
92.1

167.3

92.7

64.9
84.5
81.6
164.0
100.0
132.0
119.8
101.9
113.5
79.1
106.6
120.1
112.6
restrictions,
Poland.
126.9

Basket
VI

153.8
81.2
91.4
87.7
90.0
80.3
88.8
81.4
58.6
55.2
73.7
67.3
75.5
71.7
154.3
136.4
100.0
100.0
.1
128.1
118.0
.3
111.8
101.2
.6
97.2
89..4
104.6
95..3
74.1
69..5
113.0
94.3
85.
117.0
98.
as103.9
in Austria,105.
Bulgaria,
113.7
94.3

CORRECTION FACTORS

A comparison of these group basket P0"s for a considerable
sample of ratios with the P 0 's as found by direct calculation reveals
that the latter are in a great majority of cases in excess of the
former. A reasonable explanation appears to be that when the
ratios—using any two countries designated as 0 and 1, country 0
being in Group A—

and — ^ - ^ are compared, the com2 Po 9A
2 po Ço
bination of local quantities with local prices gives the best economic
adjustment, that is the lowest relative cost, with the result that the
ratio in which the best adjustment figures, e.g., always the fVs> is
higher than its substitute ratio. Whether or not this explanation of
the differences is applicable, correction factors to correct for the
differences can be calculated and the final results will be somewhat
better than the uncorrected ratios.
The correction factor to be applied to a ratio using a particular
country, say country 0, as base is obviously equal to the average
ratio of all the group basket P0"s and the direct calculation PQ'a
using that country as base. This average ratio of all the group
basket P 0 "s and the direct calculation P0's can be computed
2 ¿

INTERNATIONAL COMPARISONS OF FOOD COSTS

without recourse to the lengthy process of first deriving all the
P 0 's, by the use of the product sum of prices and quantities using
average prices over all the countries in the comparison (the average
taken after conversion to a common currency).
Let Co be the correction factor to be applied to all the P 0 "s
using country 0 as the base
S Po'
Co =
S Po
The numerator is S P„' = (^^

+ * £ » £ + ï a L * +... ^ 2 ± l A

VSpoÎA

2po£

(S Pi

ÇA

2 Po ÇA

S Po ÇA

S Po ÇA /

+ S P2 ?A + 2 V» 5* + - 2 Pn-1 ÎA)

But (2 pi qA + Sp 2 «A +••• 2 pa.t qA) + 2p0qA=> n2 p a v q A where p .
is the average price.
Likewise 2 P 0 -

£ * * + * * * + * J Ü * + ... ^ ± A
\ 2 po ?o
2 po <Zo 2 po Co
2 p 0 go /

(2 pi go + 2 j>¿ q0 + 2 ps 9o +— 2 Pn-i ?o)2 po go
(S pi g0 + 2 p 2 g0 + . . . 2 p n .i g0) + 2 p 0 g0 = n 2 p 8V g0
Hence, substituting these values
=

(w 2 p a v gA - 2 po gA) (2 po g0)
(2 po qA) (n 2 p a v g0 - 2 p 0 go)

The corresponding correction factors for the P / ' s for the base
country may be derived from those for the P 0 "s of the numerator
1
2 pi Ci

country, because they are reciprocal : —
= 2 p 0 Ci
2 Po gì
2 pi gì

To calculate these correction factors requires, besides the data
already described, the following additional elements: (1) the set of
average prices (after conversion to a common currency); (2) the
product sum of each group basket by these average prices; and (3)
the product sum of each local basket by local prices.
The correction factors for P' are represented by the formula
c
= VcoTiTable VII gives the final ratios obtained by the use of the
group basket method, including the correction factors. The
results for P 0 ' and P\ and the correction factors are given in the
appendix.

TABLE VII. FOOD PRICE RATIOS CALCULATED BY GROUP B A S K E T METHOD, AFTER
BASED ON BASKETS OF 1 8 FOOD COMMODITIES I N N I N E T E E N COUNTRIES

N u m e r a t o r country
Denominator
country

Austria

Belgium

Bul- Czecho Dengaria Slomark
vakia

Estonia

Austria
Belgium
Bulgaria
Czecho-Slovakia
Denmark..

1.00
1.94
2.64
1.91
2.00

.52
1.00
1.09
.94
1.02

.38
.92
1.00
.91
.78

.52
1.06
1.10
1.00
1.04

.50
.98
1.29
.97
1.00

Estonia
Finland
France
Germany
Great Britain

2.79
2.23
2.27
1.12
1.84

1.38
1.11
1.20
.55
.93

1.06
.86
1.01
.47
.87

1.44
1.16
1-21
.59
.94

1.36
1.13
1.14
.56
.92

Hungary
Italy
Netherlands
Norway
Poland

1.40
1.85
1.78
1.61
2.38

.70
.81
.87
.79
1.18

.68
.64
.82
.66
1.05

.73
.81
.92
.83
i:23

.68
.90
.90
.81
1.16

.51
.68
.64
.58
.88

Sweden
1.71
Switzerland
1.53
Union of South Africa 1.61
1.41
United States

.82
.81
.80
.73

.73
.75
.73
.65

.90
.85
.81
.75

.91
.78
.80
.72

.58
.56
.53
.47

Ger- Great
Finland France many Britain

Hungary

Ita

.45
.90
1.16
.86
.89

.44
.84
.99
.83
.88

.89
1.81
2.14
1.71
1.77

.54
1.07
1.15
1.06
1.09

.71
1.43
1.48
1.37
1.46

.5
1.2
1.5
1.2
1.1

1.00 1.27
.79 1.00
.81 1.00
.41
.51
.65
.76

1.23
1.00
1.00
.48
.81

2.42
1.94
2.07
1.00
1.61

1.53
1.31
1.23
.62
1.00

1.94
1.61
1.62
.81
1.30

1.4
1.1
1.4
.6
1.2

.62
.87
.80
.74
1.00

.62
.69
.77
.70
1.05

1.23
1.58
1.57
1.42
2.08

.77
.81
.98
.89
1.30

1.00
1.07
1.25
1.14
1.70

.9
1.0
1.1
.9
1.4

.77
.68
.69
.60

.77
.69
.71
.62

1.50
1.42
1.38
1.25

1.01
.90
.87
.80

1.19
1.15
1.07
.96

.9
1.0
1.0
.8

.36
.73
.94
.70
.73

1
The differences between these price ratios and those published in the article in the February 1941 Internationa
factor; in addition revised quantities (wheat and rye bread) were used for Groups I I and VI baskets and revised

INTERNATIONAL COMPARISONS OF FOOD COSTS

ERRORS OF THE GROUP BASKET METHOD

What are the errors attaching to these results? These are
given full discussion in subsequent chapters so far as concerns the
basic data (chapter III); in the present chapter the error due
solely to the use of the group basket in place of the local basket is
considered, as shown by comparing the result obtained with the
group basket formula for P' with that obtained by using the
formula for P.
The comparison shows that the (corrected) group basket ratios
averaged within 1.4 (1.39) per cent, of the ratios calculated by means
of the formula for P. The average error, on the other hand, of the
P's without the use of the correction factor is 1.7 (1.74) per cent.
(These calculations were made using figures in which bread was
considered a single item with a single price.)
The extreme errors are also of interest. Using the corrected
ratios, the largest deviation found was 5 per cent, in two cases out of
342 ratios.
For countries in the same group the errors are somewhat less.
For the 54 comparisons the average error is only 1.28 per cent., as
compared with 1.41 per cent, for those outside the same groups.
The uncorrected figures show slightly greater divergencies:
1.40 per cent, for countries in the same group as compared with
1.80 per cent, for countries in différent groups.
These results throw light upon the average errors associated
with the method itself. These average errors are well within the
margin of error, as will be shown in later chapters, associated with
the prices and quantities and other sources of inexactness in the
price ratios. The group basket method thus appears to offer an
acceptable technique for international or interregional comparison
of food prices.
The somewhat artificial limitations imposed for purposes of
testing the method as method will be discarded in the subsequent
chapters. Chapter II will discuss the significance of the results obtained in terms of the actual cities and countries compared and will
consider methods of application to other cities, geographical areas,
national groups, and specific economic classes. Chapter III will be
devoted to problems of prices and quantities and their attending
errors. Chapter IV, finally, treats of the problem of substitutions
and the differences between the cost-of-food ratio obtained by substitutions and the food price ratio found by the Fisher formula and
the group basket method.

CHAPTER II

APPLICATIONS
The discussion in the preceding chapter has been artificially
simplified to throw into relief the problems and procedure of the
group basket method, and especially the errors associated with the
method itself. The present chapter will treat of the applications of
the method to the various specific cities and countries, and to different economic and other groups.
The problem is to find an average ratio between prices in two
different markets which expresses the difference in the cost of food,
having regard to the differing habits and standards of food consumption in the two markets. There are thus two elements in the
comparison: the prices, which are characteristic of a market, and
the habits of food consumption, which may be considered to be
characteristic of a population group. In the foregoing discussion of
method, the term "country" has been used for both these concepts.
In considering the significance of the results, these concepts must
now be clarified.

CITIES

The prices for a given "country" are in fact actually the average
prices for one or more cities in that country. In many cases, they
are averages for a number of the larger cities (omitting metropolitan
cities). The cities used as a basis foT the price calculations in each
country are given in the accompanying list.
LIST OF CITIKS

Country
Austria
- Belgium - Bulgaria - Czecho-Slovakia
Denmark - Estonia
- Finland
- France - - Germany - -

Cities
Vienna
Brussels
Sofia, Plovdiv
Brno, Bratislava
Copenhagen
Tartu, Tallinn
Helsinki
Bordeaux, Lille, Lyons, Nancy, Marseilles
Hamburg, Breslau, Leipzig, Cologne, Munich
2 i,

INTERNATIONAL COMPARISONS OF FOOD COSTS

Country
Great Britain -

Cities
- Birmingham, Bristol, Glasgow, Leeds, Manchester,
Newcastle
Hungary . . .
- Budapest
Italy - - - - - Florence, Genoa, Milan, Trieste, Turin
Netherlands - - - Amsterdam, The Hague, Rotterdam, Utrecht
Norway - - - - Oslo
Poland - - - - - Catowice, Lodz, Poznan
Sweden
- - - - Göteborg, Malmö, Stockholm
Switzerland . . . Geneva, Basle, Berne, Zurich
Union of South Africa Cape Town, Witwatersrand
United States - - - Baltimore, Boston, Denver, Los Angeles, New Orleans,
Philadelphia, St. Louis, San Francisco
-

In fact, therefore, price ratios are not between two countries,
but between prices in the cities in the two countries respectively
which are used as the basis of calculating average prices.
In principle, each city has its own price data. Grouping of cities
is subject to the condition that the prices of the group of cities
present a significant average : this is reasonably true for cities of a
given size range within a country or currency area where conditions
of food supply are not too different.

Applications

to Other Cities.

The results obtained for these cities or groups of cities may be
extended to other cities by applying suitable conversion factors.
If the price ratio is sought between a particular city in one country
and the group of cities in another country used as the basis of the
calculations in the table, the conversion factor to be used is the
price ratio between the particular city and the group of cities in
the same country which was used for the table. By multiplying by
these appropriate conversion factors the price ratios in the table
can be converted into price ratios for the particular city. 1 Thus if
i For instance, suppose the price ratio Berlin/Vienna is sought. The table
gives the price ratio Germany/Austria, where Germany is represented by 5
German cities and Austria by Vienna. The ratio Berlin/Vienna can be obtained
by means of the price ratios Berlin/5 German cities, used as a conversion factor.
The price ratio Berlin/5 German cities is found for each of the six group
baskets, by multiplying each basket by the prices of Berlin and of the 5 German
cities. Conversion factors for the six groups are then calculated by taking the
geometric average of the ratio of Berlin/5 German cities for Group III (Germany)
on the one hand and for each of the other groups on the other. The particular
conversion factor required is that which uses weights corresponding to the groups
covered in the particular ratio. Then, for the ratio Berlin/Vienna, the conversion
factor for use is the geometric average of the price ratio Berlin/5 German cities
with Group III and Group IV weights. In practice the conversion factors differ
but slightly; thus the six conversion factors for Berlin/5 German cities are 0.981,
1.008, 1.003, 1.001, 1.000 and 1.006. In this case the prices of Berlin are substantially similar to those of the 5 German cities. A similar calculation of the price
ratios, New York/8 American cities gives the six conversion factors: 1.078.
1.070, 1.075, 1.075, 1.074 and 1.083.

APPLICATIONS

price ratios for metropolitan cities or for small cities or for rural
areas are desired, the technique just indicated can be used to develop price ratios between these and cities in other countries a8
presented in the table.

Applications

to Cities in Other Countries.

Cities in a country not included in the original list can be added
to the scheme of food price ratios under the following conditions.
If the new country has food consumption habits similar to those of
one of the groups already established, price ratios can be calculated
by using this group basket in conjunction with the prices of the
additional cities. In practice, they can be calculated by using conversion factors in the manner just described, to substitute the new
cities for those of one of the countries in the group with which the
new country is assimilated.
If the cities in the new country have consumption habits that
do not accord with those of any of the groups already established,
a new group must be set up and the necessary additions of the new
prices and the new quantities made to the several basic tables.

II.

POPULATION GROUPS

The application of the group basket method to population
groups implies that the habits of food consumption represented by
the basket are characteristic of the particular group. In the discussion of the preceding chapter, the term "country" has been used
as if habits of food consumption were characteristic of the country.
Types of food consumption within a country are doubtless largely
determined by national habits and characteristics, or by the conditions of demand and supply. The validity of such a generalisation
or the limits to its validity in any specific case must be established
by evidence. One principal exception is where two nationalities or
race groups, with differing food consumption habits, occupy the
same area, in which case the national or racial food consumption
habits may conceivably exert a sufficiently important influence to
suggest the need for more than one local basket or basket type for
that area; on the other hand, local conditions of supply may in fact
be of sufficient importance to mask any effect of differing race
customs. Another influence is that of economic or social class; if
these different economic classes have markedly differing food habits
2 ¿

INTERNATIONAL COMPARISONS OF FOOD COSTS

—not merely differing absolute amounts but a differing composition of the diets—more than one local basket or basket type for a
single area may be needed.
In the next section, the effect of differences in habits of food
consumption in differing economic classes upon food price ratios is
examined in detail.

III.

ECONOMIC GROUPS

The problem of economic groups is important only if consumption varies with economic position, and only to the extent that such
changes in consumption produce an effect upon the price ratio. If
consumption varies with economic position in a community, then
when prices are weighted successively with quantities consumed
in the different economic groups, a series of Po's are obtained which
differ in accordance with the effect of these changes in consumption:
hence the price ratios between these economic groups and any given
group in another region or country will also vary.
(a) Changes in Consumption in Different Economic Groups.
To what extent is consumption affected by changes in economic
position ?
An analysis of changes in consumption with changes in income is presented for certain foodstuffs in an article on Food
Expenditure and Consumption Habits in the International Labour
Review for June 1939. * This shows changes in absolute quantities
consumed; for purposes of the present problem, differential changes
in quantities consumed affecting different foodstuffs differently are
important, while changes affecting all items uniformly have no
effect upon price ratios. A table showing the indices of relative
change in quantities consumed, with changes in income per consumption unit, is reproduced below: the varying importance of these
changes is indicated by the change of sign as affecting the different
foodstuffs, and the change in degree as indicated by the varying
size of the indices. In general, the change in income—100 per cent.—
used as the unit is relatively large—in a number of countries, in the
range of incomes shown in wage earning families, the average income in the highest income group is only about double that of the
1

Part II of An International Survey of Recent Family Living Studies.

APPLICATIONS

lowest group, so that in such cases the percentage change in consumption shown in the table corresponds to the change from the
lowest to the highest income groups among the wage-earning families.
(b) Diagnosis of Consumption in Different Economic Groups.
In terms of the method of diagnosis used in chapter I for
classifying countries in groups, however, differences in consumption
in different income groups are not especially great. Data were available for eight countries, including Estonia and Poland of Group
II, Belgium, Czecho-Slovakia, Germany, and Norway of Group
III, Denmark of Group IV and Switzerland of Group VI. The
results show in general that while in most cases all economic groups
are classified in the same group as the average for the country as a
whole, a slight tendency appears in some cases for the lowest
economic group to approach halfway towards a lower consumption
group, and the upper economic groups to approach halfway towards
one of the higher consumption groups.
In detail, for the two countries of Group II, all the economic
groups of Estonia are classed in Group II, and likewise all in Poland,
except that in the last mentioned country the lowest income group
occupied an intermediate position between I and II and might be
assigned to either. In respect of the bread and cereals consumption
alone, the lowest income group is fairly close to Group I, while the
highest income group is halfway between Groups II and III; but
when all food groups are considered the highest income group is
very much closer to Group II than to any of the others.
For the countries of Group III, in the case of Norway all income groups are best allocated to Group III. In the case of Belgium, the two lowest income groups are best allocated to Group
III, while the two highest can be equally well assigned to Group
III or Group IV. The four lowest income groups of Germany are
definitely allocated to Group III, while the highest income group
might be assigned to Group VI, Group V or Group III, the sum of
the deviations being respectively 12, 13 and 14. In the case of
Czecho-Slovakia, the two lowest income classes fall to Group III,
while the two highest income classes appear to belong to either
IV or V, although the advantage of such an assignment over an
allocation to Group III is not great.
For Denmark, the only country of Group IV for which the test
was made, the original allocation to Group IV was in doubt, since
it appeared to be assignable equally well to Group V. Of the five
income classes, one, the highest, would be Group IV; two, the lowest
2 u

TABLE VIII. INDICES OF RELATIVE CHANGES IN QUANTITIES OF SPECIFIC FOODS C
WITH CHANGES IN INCOME PER CONSUMPTION UNIT IN

Item

Bread

Meat and fish

Belgium

Colombia

Czechoslovakia

Finland

Germ

-10.9

104.4

-15.4

43.2

-1

-10.1

149.8

- 6.0

-0.1

75.9

182.4

56 6

57.2

- 9.6

-4.1

-2

30.3

41.0

119.7

Fats

-^7.7

Milk

29.7

227.4

Cheese

90.6

197.7

104.3

218.3

60.8

40.4

106.8

410.1

58.7

114.6

- 3.6

121.8

-15.8

20.3

Butter

Eggs.
Potatoes

17.5
Sugar

45.8

122.1

- 2.3

42.0

1
With 100 per cent, increase in income per consumption unit the index shows the percentage of change in quan
* Wage earners and lower officials' families.

APPLICATIONS

and the next highest, are equally good with either IV or V, and the
other two appear to go better with Group V.
Finally, with Switzerland, the only country in Group VI,
analysed by income classes, all the income groups are assignable to
Group VI.

(c) Group Basket Method as Adapted to Economic Groups.
In cases where the several economic groups are assigned to the
same group as the average for the country as a whole, the price
ratio obtained for each economic group by the group basket method
will be the same as that for the country as a whole: in other words,
the difference in consumption as income changes is ignored in calculating ratios. This does not prove that in calculating the P 0 and
Pi by the direct method, differences in results would not appear if
the actual quantities consumed in the different economic classes
were used in calculations. These differences, however, are merely
too small to be revealed by the group basket method.
Where, for a particular economic class the diagnosis reveals a
position, e.g., halfway between Groups III and IV, a close approach
to P might be reached by using an average of these two baskets for
the group basket corresponding to the base country. Thus, for
the upper income levels of Czechoslovakia, a better result might
be obtained by using average ratios calculated from group-baskets
III and IV for P 0 , and for Pi the ratio indicated by the basket for
the group in which the other country or the particular economic
class of the other country falls. This point needs to be tested in practice to determine whether such a device gives any substantial increase in accuracy.
Two final considerations are suggested. First, the negative
conclusion just reached that, in general, changes in consumption as
income changes do not produce substantial differences in the price
ratio, does not prove that this conclusion is true for all countries,
since in some countries such changes in consumption may be -more
substantial. The question must then be kept open and when economic position is shown to be important, the method of calculating
price ratios should take it into account. And in any case, for detailed analysis, a method may have to be utilised that permits of
taking into account changes in consumption with the changes in
income.
Secondly, the possibility is suggested of introducing intermediate baskets, in such a way as to minimise the amount of
additional calculation and to improve the fit between consumption
2 i

INTERNATIONAL COMPARISONS OP FOOD COSTS

in the particular case and the commodities in the basket chosen.
Including in the scope of the problem not merely each country but
each economic group in each country greatly increases the totaî
volume of calculation in the direct method and would correspondingly permit of increasing the number of group baskets while
still maintaining the relative advantage of the group basket method.
(d) General Problem of Comparison between Different Economic Levels.
A brief consideration should be given to the problem of price
comparisons at different economic levels.
Within a market, comparisons of cost of living at different
levels reveal differences solely in the costs of differing standards,
since the prices of identical articles are the same for all economic
levels—the price ratio being 1. In different markets, the objective
sought in the price ratio technique is to measure the difference in
the price ratio irrespective of any differences in the costs of the
varying standards themselves.
The Fisher formula seems to offer the best means for obtaining
this result. The slightly different weights associated with varying
quantities consumed at different economic levels often have, it is
true, a slight effect upon the price ratios, which may, in fact, show
a slight trend as income changes. These differences are due to the
different weights used, and not to differences in the costs of differing
standards themselves.
An alternative formula for determining the cost-of-food ratio is
/ = —P}_2lt where the level of food consumption in the two coun2 Po ?o
tries compared is the same. If this formula, which merely compares
the costs of the standards in the two countries, is used, the condition
that the standards must be equal is obviously essential for present
purposes. Where this condition is satisfied, the ratio of the costs of
food in the two countries gives the desired cost-of-food ratio.1
In practice, this condition appears to be closely related to the
concept of equal economic levels.
Three methods of determining equal standards of food consumption may be considered: the double expenditure test, or Q
equals 1 ; the equivalence of nutritive values ; and the equivalence
of satisfactions or utilities.
The double expenditure test of Frisch asserts that where
"double expenditures are equal, standards of consumption are
the same". The test is
2 Pi go 2 pi qi = S po Ço 2 p0 Ci
1

In this case t h e term "cost-of-food ratio" is preferred to food price ratio.

APPLICATIONS

By dividing through by the right hand side of the equation, and
extracting the square root, this reduces to \ / Q 0 Q , = Q = 1.
Where this condition holds, the true ratio is given by the Fisher
formula. 1 In other words, the condition for using the more general
formula leads to the Fisher formula.
To establish the equivalence of nutritive values or the cost of a
new basket with equal nutritive content would require an elaborate
technique of general substitutions, a problem which is considered in
chapter IV.
The equivalence of satisfactions or utilities seems to require
substantial equivalence of economic level as a prerequisite to determining the ratio.
The upshot of the matter appears to be, therefore, that the
alternative formula for a cost-of-food ratio can be applied for practical purposes only between groups at approximately the same
economic level.
It is possible that the price ratio is of special significance and
subject to less error or modification by substitutions or otherwise
where calculated between two equivalent levels in different countries.
If so, in order to obtain such a specially significant ratio, approximate methods of determining equivalence, such as equal real incomes, or equal food values, etc., might be utilised and the Fisher
formula applied to these equivalent groups.
This is not to say, however, that comparisons between different
income levels in different countries are invalid or inappropriate.
Comparisons may be made between any two income levels. As
suggested at the beginning of the present section, comparisons may
be made between two income levels in the same market: the fact
that the price ratio is 1 means that the difference in costs corresponds to the differences in the standards. Between two cities, a
similar price ratio may be needed if only to compare differences in
costs of two different standards after eliminating price differentials,
the latter being an essential step in the determination of the former.
In any case, the present enquiry is concerned only with the
price ratio, that is, the differences in costs due to price differences,
and not with differences in the costs of differing standards.
IV.

CLIMATIC DIFFERENCES

Application of the price-ratio technique to cities subject to
large differences in climate raises the question whether adjustments
1
See Robert M. WOODBURY, "Quantity Adjustment Factors in Cost of Living
Ratios" in Econometrica, Vol. 8, No. 4, Oct. 1940, pp. 323, 328.

2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

should be made to take account of the differences in quantity requirements caused by climatic differences. This raises three important points.
In the first place, in theory the price-ratio.technique may have
to be supplemented by quantity adjustments where climatic differences produce differences in quantity requirements. This is perhaps
best illustrated by the case of differing fuel requirements in warm
and cold climates : where the amount of fuel required for a given
degree of comfort varies, a fuel-cost ratio must adjust the quantities
of the numerator country in comparison with those of the denominator to provide for the differences in requirements in the two regions.
This whole question with relation to the cost of living index is discussed in an article "Quantity Adjustment Factors in Cost of
Living Indices". 1
Secondly, with regard to food, there is as yet insufficient evidence in comparisons between countries in the temperate zone to
show that more calories or other food elements are required in one
country than in another, or,if more calories are required,to showhow
much the allowance must be increased to yield the same standard.
Family budget studies have shown, it is true, that the average consumption in calories per consumption unit varies from country to
country, but this cannot be taken as valid evidence to show that
these differences correspond accurately to differences in requirements.
A primary obstacle to any attempt to take climatic differences
into account is the relative crudeness of the methods available to
judge calorie requirements of different individuals and the number
of important factors, age, sex, degree of bodily exertion, etc.,
which have to be taken into account in any such study.
The question of quantity adjustments in respect of food may
therefore be dismissed from further consideration here.
Thirdly, the reservation must be made that the whole procedure
of comparing costs of food can be applied only where conditions are
not too different. There is no significance in a price ratio between
communities as different as the Eskimo and the Sahara desert
tribesmen. The exact degree of difference which invalidates the
procedure is difficult to determine, but in practice, where differences
are great and striking, the mappropriateness of comparison is
obvious. Specifically, an index of similarity and dissimilarity is
afforded by the percentage of the food budget for one community
which is not available in another with which a comparison is sought :
the higher the percentage the less reliance is to be placed upon the
» Ibid., pp. 322-332.

APPLICATIONS

calculation of a ratio. Where all the items are different in the two
markets, the price-ratio technique breaks down completely; and
the alternative cost-of-food technique with substitutions required
for each and every item would easily evoke sufficient doubts to
emphasise the reservations that such a comparison must carry.
In practice, the techniques have been applied principally to
comparisons between countries of the temperate zone under Western
modes of living. For some purposes, comparisons outside these
restrictions may prove useful. For example, comparisons between
the costs of living in the Philippine Islands and the United States
are useful and of interest—for missionaries, for representatives of
business firms seeking to establish trade relations or branch establishments, for discussions of problems such as migration, tariff
adjustments, etc., etc. Such applications of the techniques discussed here between countries differing widely in climate, customs,
and habits are obviously subject to much greater reservations as to
margin of error, etc., than, for example, comparisons between the
United States and Great Britain.

CHAPTER III

ERRORS IN BASIC DATA
Errors in the basic data of quantities and prices affect price
ratios calculated by any method. In describing and testing the
Group Basket Method in chapter I, errors in the basic data on
quantities and prices were expressly left out of account. They must
obviously be considered, however, together with the means available for reducing them, in any complete discussion of the problem
of international comparisons of food prices.
In general, such errors depend upon the methods of collecting
the basic data. Since sources of data, methods of collection, and
techniques differ for quantities and prices, errors in quantities and
prices must be considered separately.
QUANTITIES

/. Sources of Data.
The principal source for data on quantities is the family budget
study. For all except three of the countries included in the tables
shown above, the quantity data were drawn from family budget
studies as listed in "An International Survey of Recent Family
Living Studies" in the May and June 1939 numbers of the International Labour Review.
In order to place the data on a comparable basis the quantities
of specific foods consumed are stated per (adult male) consumption
unit of standard type, the results as given in the sources being
converted where necessary by the Office to this form.1
In the three cases, France, Great Britain, and Italy, where
family budget studies were not used, the data were derived from
estimates of consumption supplemented by more or less detailed
investigations approaching the family budget study type. The
figures for France are the average quantities consumed per capita
in the Paris region used by the Comité technique de l'Alimentation
i For the method of conversion see International Labour Review, Vol. XXIX,
No. 6, June 1939, p. 821.

ERBOBS IN BASIC DATA

as weights for cost-of-living indices for this region, converted by the
Office to quantities consumed per consumption unit. The figures
for Great Britain are estimated by the Office on the basis of figures
in Crawford's The People's Food and Sir John Orr's Food,
Health and Income. Crawford's figures for population groups
C and D and Orr's figures for income groups I to IV were averaged
and the mean of these compared with the data published by the
Advisory Committee on Nutrition : for each foodstuff for which data
were available the lower of the two quantities, based on Crawford and Orr data on the one hand, and the Advisory Committee
on Nutrition data on the other, have been used. This series was
then converted into quantities consumed per consumption unit.1
The figures for Italy are quantities of food available for national
consumption during 1937 as published in the Italian Statistical
Yearbook. The per capita figures of the original were converted
by the Office into quantities per consumption unit.
2. Errors in Data.
(a) Methods of collection. The errors in data are dependent
on the specific methods used in collection. In the case of family
budget studies the usual method followed is a day-by-day accounting of items and quantities purchased in conjunction with changes
in inventory at the beginning and the end of the period. For the
most part the family budget data cover the period of a year. In
some cases, however, the food studies are limited to four periods
of one week each in each of the four seasons. Details of methods
followed in each country are given in the article already referred
to in the June 1939 number of the International Labour Review. An
alternative method which is sometimes used in food studies, especially those dealing with nutrition, involves the weighing of foods
actually eaten. This method is much superior in accuracy of final
results. Unfortunately, where this method is utilised the scope
of the study is usually restricted.
Estimates based upon general figures for consumption are
usually subject to considerable error. In the first place, such figures
often are based upon production plus imports less exports, a procedure which disregards the diversion of supplies available to other
uses than human consumption. The quantities of grain available
for human consumption, for example, must be converted into flour
or bread. At best, this method is unsatisfactory for estimates of
consumption among the wage-earning population.
i See footnote 2, p. 56.
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

(6) Errors of sampling and representativeness of data. Little
evidence is available on the question of the representativeness of
family budget data for wage-earners' families. The question whether
the data obtained for the sample of wage-earners' families chosen
for study are representative of the entire group of wage-earners'
families in the country as a whole depends upon the methods used
to select the families. Until recently, comparatively little attention
has been paid in family budget studies to methods of sampling.
An important question also is the geographical limitations of the
study; in many cases the studies are limited to the chief city or
cities of the country. An analysis of the variation of food consumption in different cities or income classes in the same country indicates comparatively minor differences, suggesting that food consumption is to a large extent determined by national customs and
habits, as influenced or governed by conditions of supply or price.
(c) Variability from year to year in quantities consumed. An
important question relates to the problem of sampling in time, or,
in other words, whether the quantities consumed vary significantly
from year to year. This variability might be due to differences in
the sample, i.e., if different families are chosen at different times,
or to changes in quantities consumed arising from any of a number
of causes, including specifically the effect of changes in prices.
This question is important with reference to the comparison of
family budget data in different countries at different periods. Is
it permissible, for example, to compare food consumption in Germany as in 1926-7 with food consumption in Sweden as in 1933 ?
Evidence on this question can be obtained for countries which
have taken annual family budget studies. Among them is Austria
for the period 1926-34. To test the influence of changes in quantities
upon the price ratio, P0, between Germany and Austria, calculations have been made using successively the quantities as found
in the Austrian enquiries for 1927, 1929, 1931, 1933 and 1934. The
variations in quantities consumed are due to all factors including
not only fluctuations in prices but also changing habits and changes
in the sample. The enquiry covered about 70 families in Vienna,
and as far as possible the same families were covered each year, so
that changes in the sample are of comparatively little importance.
The results showed that the value of P0 changed very little, the
figures being respectively .924, .927, .928, .923 and .922. The low
net effect is explained in large part by the fact that the changes
tend to offset one another, the decreases in the consumption of
certain articles being offset by increases in the consumption of
others. The details of these changes are shown in table IX.

ERRORS IN BASIC DATA

TABLE IX. CHANGES IN QUANTITIES CONSUMED PER CONSUMPTION
UNIT IN AUSTRIA, 1 9 2 7 - 1 9 3 4 , AND THE EFFECT ON P0, GERMANY/
AUSTRIA, OF USING AUSTRIAN QUANTITIES FOR DIFFERENT YEARS

Foodstuffs

Quantities (kg.) consumed per consumption unit
in Austria
1927

1929

1931

1933

1934

Bread
Flour
Rice
Beef
Pork

103.04
27.82
5.02
18.83
20.71

98.56
25.92
5.48
13.53
23.29

92.71
27.01
5.29
9.36
23.82

95.64
27.38
5.33
8.17
21.52

95.97
27.10
6.22
7.10
22.23

Veal
Lard
Margarine... .'
Milk (litres)
Cheese

4.22
4.70
1.81
194.00
2.92

4.62
5.37
2.34
195.64
3.43

6.05
3.29
2.34
179.22
2.77

3.88
2.70
2.63
169.76
3.07

2.99
2.27
2.82
148.17
3.21

Butter
Eggs (number)

4.44
199
2.09
54.19
26.33
2.00

5.07
219
1.79
53.29
27.01
2.00

4.56
212
2.48
52.20
25.19
2.00

3.90
189
2.33
55.67
23.82
2.00

0.924

0.927

4.49
226
2.23
50.74
27.01
2.00
0.928

0.923

0.922

Potatoes
Sugar
Coffeei
Po, Germany / A u s t r i a . .
1

Quantity estimated.

The fact that such studies in successive years reveal approximately the same general pattern indicates that the results may be
accepted as sufficiently reliable for present purposes. The variability of quantities over a period of time appears to have little
influence upon the ratios.
S.

The Problem of Coverage.

An important aspect of inaccurate or incomplete quantity
data is the lack of complete coverage. In part this is due to insufficient detail in family budget studies but for the most part to
gaps in price data.
Table X presents the percentages of coverage formed by comparing the cost of the eighteen commodities used in the calculations
of chapter I to the total food budget in each country. These percentages ranged from 54.4 per cent, for Denmark to 83 per cent, for
Finland. The table shows also the percentages of coverage within
each food group. These ranged from an average of 96.3 per cent.
for the group of milk products and eggs, 84.4 per cent, for margarine and fats, and 79.8 per cent, for bread and cereals, to 52 per
cent, for meat and fish and only 33.6 per cent, for vegetables and
fruits.
2 ¿

INTERNATIONAL COMPARISONS OF FOOD COSTS

TABLE X. PERCENTAGE OF TOTAL EXPENDITURE IN EACH FOOD
GROUP COVERED BY FOODS INCLUDED IN THE LIST OF 18 COMMODITIES,
IN FOURTEEN COUNTRIES
Food Group
Country

Bread
and
cereals

Meat
and
fish

Marg.
and
fats

Milk,
etc.

Belgium
Bulgaria
Czecho-Slovakia...
Denmark
Estonia
Finland
Germany
Hungary
Netherlands
Norway
Sweden
Switzerland
Union of South
Africa
United States
Average
(14 countries). ..

87.7
89.8
79.6
49.6
91.5
92.9
82.0
85.6
72.7
81.6
96.3
64.9

49.7
68.2
54.5
38.8
43.5
49.7
33.3
60.5
63.0
30.6
59.6
51.0

85.6
58.0
100.0
98.0
100.0
100.0
88.5
99.9
88.6
97.1
100.0
5.5

100.0
100.0
100.0
87
99
93
99
96
100.0
89.1
90.3
96.9

76.9
66.2

71.0
54.1

100.0
60.6

79.8

51.9

84.4

Vegetables
and
fruits

Miscel- Total
laneous

51.7
59.9
63.7
56.5
80.3
88.6
71.1
67.4
69.8
79.7
82.0
28.0

72.0
72.1
72.4
54.4
75.4
83.0
66.2
74.8
70.6
66.6
76.0
55.9

100.0
95.9

45.5
22.9
27.
22
56.
60
39
35
27.
43
24.8
13.0
18.8
14.8

61.9
73.5

70.4
60.9

96.3

33.6

68.0

69.3

The principal shortcoming of family budget studies in respect
of food items is lack of specific detail. An item of small importance
is placed under a more general head or is lumped in with a group of
"all other items", for example, "lard" may be returned as "animal
fats", "margarine" under "vegetable fats and oils", rice and macaroni placed under "other cereals", etc. In such a case, for example,
the omission of macaroni in the family budget study does not
indicate whether consumption is zero or some (small) quantity less
than the amount assigned to "other cereals". This point is of little
significance from the viewpoint of the family budget study; but
from that of cost of food comparisons, the absence of detail requires
estimates of the missing quantities if the commodity is to be included at all. Obviously, no useful price quotations can be obtained
for "all other" items. The different vegetables and fruit consumed
are often not reported in detail separately. The different qualities
of meat or the different cuts of beef, veal, lamb, etc., are often
ignored. Sausage is returned as a single item rather than with
details indicating food value or price. Difficulties are found also
in comparing certain fish items often specified as merely fresh fish
or salted fish, and lacking characterisation of the particular kind
of fish.

ERROBS IN BASIC DATA

These gaps in the specific details of foods consumed lead
directly to gaps in the data on prices and raise the questions (1)
what is to be done in regard to these gaps? and (2) how much
effect do they have on the final price ratios ? The first question is
usually answered by omitting the commodities for which either
quantity or price data are wanting. Other expedients such as
estimating the missing data are sometimes adopted where the items
are unimportant or where a satisfactory method of estimate is
available, and such expedients as pricing the particular kind of the
food item, e.g., fish, principally consumed in each country—which
raises the whole problem of substitutions (see next chapter)—are
sometimes resorted to. But by far the simplest and easiest way is to
omit the items altogether.
Assuming that this method (of omission) is adopted, an important question is how much difference does it make in the final
result. In other words is a P calculated on the basis of 18 commodities acceptable as approximating within a reasonable percentage of
error a true P calculated on the basis of the full food budget ?
No definite estimate of error can be given, since obviously the
difference between the two results depends on the number and
importance of the items omitted, in conjunction with the deviation
of the price ratios for the omitted items from the final price ratio.
In general, however, the smaller the omissions the better the results
should be.
In two cases the result of the computation based upon 18 commodities can be compared with the result based on full data. For
the ratio Germany/Norway, a test can be made on the basis not of
prices of October 1938 but of prices at the time the family budget
enquiries were conducted during 1927-29. The average price ratio
based upon all commodities in the food budget was .90, while that
based on the 18 items included in the present calculations was .88,
an error of - 2 per cent. In this case the coverage of meat and fish,
as well as of vegetables in the 18 items was relatively low both in
Germany and Norway. In the second case—the ratio of Germany/Switzerland—a similar test shows that the computation
based upon 18 commodities was 3.2 per cent, in excess of the ratio
based upon the detailed and comparable list of items in the full
budget.
Other evidence may be adduced in the form of a difference in
the price ratio based on a small number of items and one based on
a larger, though not the complete, number. Thus the price ratio
Sweden/Norway based on 23 items was found to be .94, one based
upon 15 principal items .94, and one based upon 10 principal items
.95. In these figures, the items characterised as principal items were
2 i,

INTERNATIONAL COMPARISONS OF FOOD COSTS

determined by the values of the Swedish budget. Where the
"principal" items were determined by the values of the Norwegian
budget, on the other hand, the result based on the 10 principal items
was .92 and on the 15 principal items .92. One might infer from
this that if the quantities included did actually cover the principal
items, the final results should be fairly close to the true ratio.
The unequal distribution of the gaps, in particular, the low
coverage of meats and fish, and vegetables and fruits suggests a
possible method of reducing the effect of the missing items. In one
case, for example, comparing the United States with Sweden, an
analysis of P 0 for the different food groups shows considerable
differences between the different food groups. Thus, with an average
price ratio of 1.37, the ratio for cereals and bread was 1.14, for meats
and fish 1.41, margarine and lard 1.2, milk products and eggs 1.65,
vegetables and fruits .89, and miscellaneous .93. Obviously, if
similar and parallel differences appeared in the ratios for the omitted
items, the final result would be affected by the relatively low coverage in the vegetables and fruit group.
If it could be established that the price ratios for the various
items within each food group were relatively constant, the price
ratio for the items included of each group might be extended to the
other items of the same group for which data on prices were lacking. 1
Sufficient evidence for this procedure is, however, lacking. But
even if the inference is not justified that a price ratio for a few items
in a group can be extended to cover all items in the group, it may
well be true that a price ratio for a particular item may fairly be
extended to cover closely similar items. For example, a price ratio
for one kind of native cheese might be extended to cover other
native cheeses, or in general, the price ratio for one quality of a
particular kind of meat might be extended to other qualities. This
means in practice, that, when this inference is justified, the weights
given for a particular price ratio may be extended to cover not
simply the specified articles but also the articles closely allied to it.
This device would very considerably extend the degree of coverage
of certain items. But it is important to show how far the inference
is, in fact, justified, since otherwise errors may be introduced into
the final results. Before such an extension is made of a particular
case, the evidence for its applicability should be examined in detail.

i That is, in effect, to weight the price ratios for each food group by the full
weight of all the items in the groups instead of by the weight of the items for
which price ratios were actually obtained.

ERRORS IN BASIC DATA

4- Detail of Individual

Commodities.

The extent and importance of the gaps and errors in the basic
data on quantities can best be understood in connection with a
review of some of the problems affecting the items as actually met
with in the different countries. The specific problems include the
following points: (1) Absence of detail in the family budget study
where the specific item is not listed or shown separately, usually
because it is deemed not to be of sufficient importance in the particular country to warrant itemisation, but occasionally because
though important it is included with one or more related items in a
total such as bread, coffee and coffee substitutes, beef and veal, etc.
The latter raises the general problem of what is a commodity for
purposes of price ratio analysis. (2) Absence of detail as to qualities
is common to many items in family budget studies, but on the
other hand too much detail may render the results almost unusable.
For example, if all the many kinds and brands of tea are specified,
each with its own price, the lack of agreement in brands and kinds
consumed in the several countries might render comparison of
prices difficult, if not impossible. The usual recourse in such a case
is to obtain the quantities consumed on the one hand and the prices
of the most typical brand or kind in the several markets on the
other. This point, however, will be discussed further under "Prices".
The quantities of the 18 commodities as used in the calculations are given in table X I . These are drawn from the summary
table presented in "An International Survey of Recent Family
Living Studies, Part I I : Food Expenditure and Consumption
Habits". 1 The figures are compiled from the data of the original
sources by converting the quantities consumed as given in the
source to quantities consumed per consumption unit of standard
composition, as described in the same article. 2
The notes as given in the table throw light upon the adjustments made to obtain the quantities consumed for the commodities
chosen for the calculations. Those commodities were chosen for
inclusion where (1) prices were available in all countries, and (2)
quantities were available or could be estimated to a reasonable
degree of accuracy. In some cases usually affecting items of lesser
importance estimates were required, as explained in the footnotes.
In a few cases, perhaps, estimates have been made even where the
basis of estimate left something to be desired, in order to give as
large a series of commodities in as large a number of countries as
possible for a test of the group basket method.
i International Labour Review, Vol. X X X I X , No. 6, June 1939, pp. 822-27.
2 ibid., p. 821.
2 4

TABLE XI.

Foodstuffs

Austria

I SI i

Wheaten bread.. 18.6
81.1
Flour
27.4
6.3
1.5
7.2
16.6«
0.1
3.0
2.3
2.3"

Milk (litres), .,, 158.6
3.2
Butter
3.9
Eggs (number)..
191

Belgium

QUANTITIES CONSUMED PER ADULT MALE CONSUMPTION U N I T
TO FAMILY BUDGET STUDIES OF WAGE EARNERS' FAMILIES

Bulgaria

195.4
1.2
7.2
1.8
1.6

211.7"

21.5
8.0
0.0
2.8

5.8'
13.6'
9.0>
1.0»
5.5"

2.9
4.8

145.1
5.3
17.5
192
2.3
3.1
56.2 220.2
24.0
Coffee b e a n s . . . . 2 . 4 « 14.7
6.7

50.5
4.9
1.3

30.7
6.2
0.5
70
11.5
19.1
9.3
0.3

CzechoDenslovakia
mark

8.7
15.6»
0.0
10.2

28.3
159.4
10.3
4.0
0.9
9.2
6.3
7.7
1.7

0.6
3.1

2.2«
17.9

1.5
2.9

154.5
3.7
3.8
169

105.6
5.7
9.6
227

151.2
0.2
5.2
7

2.0
100.5
28.8
0.8

0.4
92.6
35.0
6.8

2.2
205.4
28.6
2.1

84.8
52.7
3.8
0.2
10.5
9.7
0.6
1.9

16.7
76.9
13.2
1.0

Estonia

Finland

France

Germany

Great
Britain

Hungary

11.9
28.0
78.3
6.2
0.6
11.0
9.2
0.0
1.5

115.7«

70.5«

14.4«
3.5
8.0
1.6«

81.0«
27.0«
54.9
4.2
0.5
5.3
9.8
0.5
1.6

130

15.9
1.5
4.6
8.4

21.4
90.9
15.6
2.0
2.4
6.8
8.3
0.6
1.6

1.2
1.2

3.7
10.3

3.5
3.0

7.2
0.0

96.4
5.8
5.8
106

158.6
5.1
6.2
161

61.5
3.0
8.0
121

2
4

10.8
62.7
12.1
2.4

2.9
162.7
17.6
1.4

3.0
70.0
20.0
0.0

113.1
0.2
0.6
80
4.3
72.1
21.3
0.8

1.1'
4.6
331.6
1.1
14.7
61
1.6
114.7
29.8
6.1

2.4
3.4
3.9

1
Quantities converted by the I.L.O. into quantities consumed per standard consumption unit: for statement of method see "An International Survey of Recent
Family Living Studies; Part II. Food Expenditure and Consumption Habits",
International Labour Review, Vol. XXXIX, No. 6, June 1939, p. 821.
« Bread in source, assumed to represent wheaten bread.
• Estimated from 108.0 kg. of bread in source, divided 75 per cent, wheat, 25 per
cent, rye, in same proportion as wheat and rye form of total production plus imports
less exports.
«Estimated from 121.6 kg. flour covering wheaten bread, flour, and macaroni:
100 kg. flours 130 kg. bread.
« Distributed according to proportions derived from quantities consumed, p. 53,
of League of Nations Report on "Statistics of Food Production, Consumption and
Prices".

18.0
1.9

16
5
5
5
3
0

3
28
5
0

• Figures for vea
cluding beef and vea
' Figure for fres
Britain and the Neth
> Estimated amo
including bacon and
' Figure for anim
'• Estimated.
««Figure for vege
«i Coffee.

ERRORS IN BASIC DATA

Certain special problems, however, relating to specific commodities are discussed below.
Two problems are raised in regard to bread: one, a question
of method, whether bread should be treated as a single commodity,
and if so, how it should be priced, or as two or more commodities,
that is, principally wheaten bread and rye bread; and the second
that of a series of questions of detail to estimate the quantities of
wheaten and rye bread consumed in those countries where family
budget study data or other sources do not give the requisite information. The first question is resolved tentatively in favour of
separate figures for wheaten and rye bread, partly in order to ascertain how much difference it makes in the final result whether
bread is treated as a single item or as two items; this question is
then deferred to the next chapter where the problem is discussed in
connection with the question of adaptation to local customs or
allowable substitutions.
The importance of the second question is indicated by the
series of notes attaching to bread quantities in the table. For a
number of countries the sources used do not show whether the
bread consumed is wheaten bread or other kinds. In most of these,
however, the absence of characterisation merely means that wheaten
bread is used so exclusively that it has not been thought necessary
to provide for the alternative use of rye bread. This is true for
Great Britain and the Union of South Africa, and probably for
Italy and Bulgaria, where judging from production and import
and export figures, wheaten bread seems to be used almost exclusively. In Switzerland, production, import, and export figures for
the available supplies of wheat and rye give a ratio between the
two of 93 to 7, suggesting that bread is principally wheaten bread.
In Hungary, on the other hand, figures of available supply are in
the ratio of about 75 to 25. An analysis of the figures of available
supply in relation to the relative consumption of wheaten and rye
bread in wage-earners' families according to family budget studies,
indicates that in countries where the consumption of rye bread
predominates, the percentage of rye bread consumed is usually
greater than the ratio of rye to wheat in the available supply, while
in countries where the consumption of wheaten bread predominates, the percentage of rye bread consumed is usually less than the
ratio of rye to wheat in the available supply. 1 In the case of Hungary

i Figures on production, imports, and exports are given in the INTERNATIONAL
INSTITUTE OF AGRICULTURE, International Year Book of Agricultural Statistics.
See also N. JASNY, "Die Zukunft des Roggens", Vierteljahrshefte zur Konjunkturforschung, Special No. 20, Berlin, 1930, p. 95 and elsewhere.
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

the estimation of consumption assumes that 75 per cent, of the
bread is wheaten and 25 per cent, rye. 1
The point might be made in passing that during the war period
special regulations in certain countries specify that bread shall be
composed of a mixture of grains. Thus in Switzerland the so-called
"federal" bread must contain other grains than wheat in specified
proportions. For purposes of calculation in such a case only one
method appears to be feasible, namely, to use the commodity
"bread" in conjunction with its price. When such commodity and
price are compared with a country using wheaten bread exclusively,
part of the difference in the price must obviously be attributed to
the difference in the quality.
The questions in regard to other commodities in the main
merely involve minor problems of estimate. In the case of commodities of small importance, errors in estimates have little effect
on final results. The estimates for the amounts of different meats in
Bulgaria and the Union of South Africa are those the importance
of which is greatest, but in these cases the total amount is given in
the family budget reports and only the distribution among the four
meats haB to be estimated. In the case of lard, in a number of
countries the total for animal fats was. used, the quantity of the
specific item "lard" not being available. The amounts involved,
however, are small.
The second general question is raised in those cases where
details of family budget studies in one country cover a number of
specific kinds and qualities of an article, for example, cheese (seven
kinds in Norway), while in others all these articles are grouped
together under one heading. For purposes of food analysis one may
make shift perhaps with the grouping if the details are not available :
for purposes of price comparisons too much detail may even prove
embarrassing. Other items for which detail is often shown are milk :
skimmed, whole, condensed, dried, other; meats, with specification
of the kind of cut or other description of qualities. The problem of
qualities raises the difficult question of definition in uniform terms
applicable to all countries. In practice this question may best be
treated with price problems, since it is of much greater importance
that the prices should be correct than that the quantities should
be exact.
Peas and beans are grouped together since the family budget
studies in some countries did not show the quantities of each
separately.
i A shift in the estimated consumption of wheaten and rye bread in Switzerland from 100 per cent, wheat to 50 per c«nt. of each leads to changes in the final
price ratios averaging about 1 per cent.

ERRORS IN BASIC DATA

A number of commodities were excluded because of problems
of quantities or prices. As to fish, must the quantities of each kind
of fish be obtained separately, or is it permissible to consider the
total quantity of fish consumed as a single item, and, if so, how is
"fish" to be priced? Is it legitimate to consider one kind of fish
important in the consumption of one country as equivalent to the
same quantity (or calorie value) of a different kind of fish important
in the consumption of another country? A somewhat similar
problem is whether tea and coffee are allowable substitutions one
for the other, and on what terms ? Pending a satisfactory and general answer to these questions, these problems have been avoided
in the calculations as presented above by excluding the commodities from their scope.
5.

Errors in Quantities in Relation to Error in the Final Price Ratio.

Errors in quantities affect the final price ratio only by affecting
the weights attaching to the individual price ratios. The price
ratio for a single item — is not affected at all by errors in the
quantities. Since the price ratios of the individual items vary, the
effect of a given error, say of 2 per cent, in the quantity of a particular item, upon the final ratio will vary depending upon two
further elements: the deviation of the individual price ratio from
the final average price ratio and the importance of the particular
item in the whole budget. If the deviation of the individual price
ratio from the final average price ratio is zero, the error in the
quantity has no effect upon the final result. The positive error of
2 per cent, attached to the weight of a particular item tends to be
offset in part by a corresponding underemphasis attached to the
weight of all other items, since the deviation in price of the item itself is in an opposite direction to the deviation in price of "all
other items". Thus, if the price deviation is positive and the error
of 2 per cent, is also positive, the weight of the particular ratio is
overemphasised but the effect is partially offset by the corresponding underemphasis upon all other items, which have a negative
deviation. If, however, the price deviation is negative and the
error of 2 per cent, is positive (or if the latter is in the opposite
direction to the former), these two effects partially offset each other,
and the result is further diminished by the corresponding effects
upon the other group, and the final effect is relatively smaller.
To take a specific illustration, if the item in question is a very
important one, bread, and the price deviation substantial, the
effect on the final ratio may be perceptible; with an item forming
2 t,

INTERNATIONAL COMPARISONS OF FOOD COSTS

50 per cent, of the budget, (in Italy bread is only 38 per cent.), a
price deviation of 10 per cent., an error of 2 per cent, in the quantity
of the item will, considered by itself, produce an error of .7 or .4 in
the final ratio, depending upon whether the deviation is in the same
direction as the error or in the opposite direction. If, as in most
cases, the item is less than 5 or even 10 per cent, of the food budget,
the final error is correspondingly reduced and is almost negligible.
Errors in the different items, which are not correlated with
one another, tend to offset one another in the final result.
PRICES

Sources of Price Data.

The sources of the price data are the current compilations of
prices collected usually for purposes of cost of living analyses.
In particular, the sources used in the calculations given above are
the prices collected by the International Labour Office in its annual
survey as of October 1938.1 The data are furnished by the competent statistical services of the different countries and collected
by them either specifically on request of the International Labour
Office or, more usually, for their own cost-of-living analyses. In
general, the Office list of commodities has in the past been relatively
limited, since it was found that in longer lists many items were
missing for a certain number of countries. In short, the list has
been restricted to commodities in common consumption in practically all countries.
The prices are those given for the principal cities in each country
(usually omitting metropolitan cities since prices for these might
not be representative) and are presented in table X I I . The list of
cities in each country for which prices are given is shown above,
p. 27.
2. Errors in Data.
Methods of collection. It may be helpful to state briefly the
principles and techniques of the collection of price data. 2 First, the
items to be priced, which must be, in general, commodities in
demand by the wage-earning class, must be defined in simple and
clear terms. This definition involves, of course, as much of a description of quality as is necessary to identify the item properly. Secondly, attention must be given to the sources of the quotations. These
i "Retail Prices in Certain Countries in October 1938", International Labour
Review, Vol. X X X I X , No. 3, March 1939, pp. 411-416.
2 See INTERNATIONAL LABOUR O F F I C E : Methods of Compiling Cost of Living
Index Numbers, Studies & Reports, Series N, Statistics, No. 6, esp. pp. 33-40,
48-64.

TABLE XII.

Items

Wheaten bread..
Flour (wheaten).
Macaroni
Beef
Pork»
Veal5

Austria

Belgium

RM.

FT.

1.11
.42
.45
.43
.75«
1.60
1.73
1.80«
1.87
1.73
.80

Milk (litres),
Butter
Eggs (number)..
Peas, beans7 . . . .
Coffee (beans)...

.30
2.40
3.14
.10
.47
.09
.84
5.60

AVERAGE RETAIL PRICES OF 18 FOODSTUFFS IN 19 COUN

Bulgaria

Czecho- Denslova- mark
kia
Levas
KÍ.
Kr.

2.10
2.31'
2.49
2.88
7.49

2.49'
2.35
2.78
3.00
7.93

Estonia

Finland

France

GerGreat
many Britain

E.Kr.

F.Mk.

FT.

RM.

.93
.22
.35
.80
1.40«

.58
.20
.38
.65
.90

10.83
4.06
5.00
5.97
8.00«

.71
.33
.44
.49
.78
1.64
1.62
1.79
2.08

Hungary

Pence

Pengö

4.6
4.6'
4.2
5.5
11.5

.42
.36
.39
.90
1.20

2
2
2
2
3

15.7
24.3
15.2
15.4

2.35
1.58
1.34
1.60

9
12
14
13

5.00
5.00'
4.10
15.00
20.00«
12.55 19.00
25.18 27.50
25.00« 25.00
25.00« 25.00
13.21 41.00
10.00 40.00«

12.63
12.25
10.00
11.50

1.45
1.85
1.70«
1.48

.60
.88
.68
.53

13.10
15.75
13.25
10.29

3.08
3.08«
4.31
4.19
8.86
10.11
20.02
11.73
14.60

14.75
12.60

2.15
1.34

1.42
.95

14.50
15.00

12.20
12.12

2.14
1.68

16.1
13.4

1.95
2.60

8
8

1.7«
6.00
15.59 26.00
63.00
26.51
2.00
.87
3.20 11.00
5.25
.62
3.25 22.90
19.60 145.00

1.55
14.00
20.00
.75
2.95
.83
6.23
40.00

.35
2.00
3.10
.13
.75«
.15
.52
4.79

.11
1.50
1.70
.05
.34
.07
.46
4.00

1.87
23.00
30.70
1.18

1.92
20.70
27.78
.91
5.20
.76
5.47
21.26

.23
2.00
3.12
.12

6.2
24.3
37.7
2.1

.30
2.30
3.00
.10

1
14
15

.60
4.08
.76
4.74

5.4
1.7
5.5
59.1

.46
.09
1.28
8.40

5.50
1.19
5.74
20.00

1
Price per kg. except milk (litres) and eggs (number). From "Retail Prices in
Certain Countries in October 1938", International Labour Review, Vol. XXXIX,
March 1939, pp. 412-416.
s
Estimated from price ratio wheaten bread to rye bread in another year or years.
> Estimated as equal to price of wheaten bread (see text).

6
30

* Estimated.
* 2nd quality, exc
*7 Average of 1st a
Average price fo

INTERNATIONAL COMPARISONS OF FOOD COSTS

are generally limited to retail establishments (grocers, etc.). serving
working class communities. For purposes of showing changes in
prices over a period of time, price comparisons are commonly
limited to price changes occurring in identical establishments. No
similar restriction is, of course, possible in comparing prices in
different localities, the only recourse being to restrict the comparison so far as possible to shops serving the same types of clientele.
Thirdly, quotations are usually secured for a number of establishments, thus tending to eliminate the effects of quotations markedly
out of line with others.
The difficulty of defining an item and its quality in sufficiently
precise terms varies with the particular item. Thus, dry groceries
are subject to less variation than fresh vegetables, for example, and
the kind and qualities can be more easily defined. In some markets
international or trade standards have been set up, for example,
in the marketing of cotton, wheat, and other staples. These standards are based on trade practice and serve trade needs. In retail
markets of food, however, no such international standards are as
yet available, though in many areas or countries local or national
laws or customs provide effective standards through the regulation
of labelling or the specification of minimum requirements for certain
commodities. Hence, for present purposes the task of ensuring
uniformity in the items to be priced is usually entirely a matter of
definition of the item and its quality.
In some cases differences in quality may exist which affect
consumption and price but which are not easily put into terms of
specifications for obtaining prices of identical qualities in other
countries. Fresh fish in a seaport town may be of much higher
quality in regard to freshness and palatability than fish sold in an
inland market. "Fresh" eggs are notoriously difficult to define in
terms of their exact degree of freshness; the average freshness
doubtless varies in relation to size of city and nearness to rural
sources of supply. In other cases identical qualities may not exist,
as in the case of bread, where in an emergency a Government prescribes an admixture of rye or maize to wheat, in making bread for
the market.
These variations in quality may be responsible for differences
in prices due solely to the difference in quality, and thus the value
of the data for purposes of determining the true price ratio between
the two commodities is impaired. 1
1
The requirement often made, that in each market the commodity be of a
quality locally consumed by the wage-earning group, may in some instances
conflict with the requirement that qualities priced in the different markets must
be identical. Where qualities are different, the problem is raised as to whether
{continued on next page)

EBROBS IN BASIC DATA

Averaging a number of quotations for a single market and
averaging the quotations for a number of markets tends to eliminate
unusual errors or lessen the influence of individual quotations
subject to such errors. Such a procedure tends to improve the
accuracy of the results, provided that the general plan of collection
of the data is properly aimed to obtain substantially identical
qualities.
Prices are always specific. Thus, a price is for lard, not for
animal fats; for coffee beans of a particular brand, not for "coffee"
implying a combination of coffee and coffee substitutes; for a particular kind of cheese, not for cheese in general. In this respect the
price data do not usually fit exactly with the quantity data. The
quantity data, being derived from family budget studies, often
do not have the detail of brand, or specific kind or variety of the
item consumed, or if the detail is available in the original schedules
or accounts it may usually be omitted in the published reports.
Hence, the consumption data commonly give the amounts consumed of cheese, of coffee, etc., often without further detail. In
such cases the usual practice is to obtain prices for a specific brand
commonly consumed by the wage-earning classes, and to consider
the quantity of the item consumed as if it were all of the specific
brand that is largely consumed by the population groups under
consideration. 1
In some cases price quotations may be wanting, particularly
for articles not in demand in a particular country. Where this
happens, either the item must be omitted from the comparison
entirely or the price must be estimated : it is not correct to set the
price at zero. It is true that where the price is wanting the quantity
consumed in the particular country is zero, and hence does not
figure in P 0 : but in Pi where the weights are the quantities consumed
in the second country, the quantity is not zero and the price is
required. In the present calculations, where price data are missing,
the item is usually omitted; except in a few instances, usually of
relatively minor items, where price data were omitted from one or
two countries only, the item was retained if a satisfactory estimate
for the missing items could be supplied. Notes are given in the
table wherever an estimated price is used.
the poorer quality in one market is an allowable substitution for the better quality;
that is, whether as a result of a price "bargain" and the adjustment of consumption to it, the lowering of the standard by reason of the lowered quality is offset m
part or wholly by an increase in the standard by reason of the larger quantity
consumed at the iower price—this is a specific case of the problem of substitutions
which1 is discussed in chapter IV.
With more complete details, a more exact procedure of comparing prices of
specific brands with their quantities might be followed: it would be of interest to
compare the results obtained with the two types of procedure where full data were
available.
„ ,

INTERNATIONAL COMPARISONS OF FOOD COSTS

In one case, that of the price of rye bread in certain wheat
consuming countries, the problem of estimation is of some importance: rye bread is an important item in the diet in a number of
Countries. Omitting this item entirely is not an acceptable alternative since it would affect the price ratio in many countries and render
it less representative. An examination of the relative position of
rye and wheaten bread prices shows that in the rye consuming
countries the rye price ranges from 24 to 83 per cent, of the wheaten
bread price1, while in the wheat consuming countries the rye price
is either just a little below or a little above the wheat price. Accordingly, in the four countries Bulgaria, France, Great Britain, and
the Union of South Africa, where no rye price was reported and
where it could not be estimated from the returns of other years2
the price of rye bread was set equal to the price of wheaten bread.
This gives the same result in these cases as the alternative procedure
of taking the price of "bread". It has the additional justification
that in a country where rye bread is not consumed in any considerable quantity, a newcomer from a rye consuming country would
naturally tend to substitute wheaten bread for rye.
Form of quotation. One type of error is in the degree of approximation involved in the form of quotation. Thus, the price of potatoes in Great Britain, in its published prices, is quoted to the
nearest farthing: for a unit of 7 pounds priced at around 6d. to 8d.,
this means an error varying up to 2 per cent. ; the price per egg, at
around 2/^d., expressed to the nearest farthing, has an error, due
solely to the form of quotation, varying up to 5 per cent. Each item
has its own degree of approximate price, depending on the currency
unit used in the quotations. Usually, if prices are quoted for a
number of markets, an average can be given to a much closer degree
of accuracy than is customarily obtained for purposes of publication.
If the figures are to be used for comparing prices of identical qualities in different markets, it seems unwise to suggest a high degree of
accuracy by presenting decimals for price quotations, which in fact
may be subject to considerable error. In any case, this type of error
is perhaps not of great importance, since such errors tend to cancel
out in forming an average price ratio for a number of food items.
8. Relation of Errors in Prices to Errors in Final Price Ratios.
In contradistinction to errors in quantities which affect the
final ratios only through their effect upon the weighting given to
1
Except Czecho-Slovakia, where the rye price is about 94 per cent, of the
estimated
wheat price based on average ratios in October for some four years.
2
Estimates of rye prices from data obtained in other years were made in the
cases of Belgium and Italy.

ERRORS IN BASIC DATA

the different price ratios, errors in prices affect each price ratio
directly. If in a comparison between two countries one price is
correct and the other is affected by'an error of 2 per cent., the price
ratio for that particular item will be in error by 2 per cent.—a
positive error if the error is positive and in the numerator of the
ratio fraction, or negative and in the denominator; or a negative
error if the error is negative and in the numerator, or positive and
in the denominator. If both prices in a comparison are subject to a
2 per cent, error: (1) the price ratio may be subject to a positive
error of 4 per cent., if a positive error of 2 per cent, in the numerator
is combined with a negative error of 2 per cent, in the denominator ;
(2) to a negative error of 4 per cent, if a negative error of 2 per cent.
in the numerator is combined with a positive error of 2 per cent.
in the denominator; or (3) may be substantially correct if the
errors in numerator or denominator are both positive or both
negative. This illustration indicates the building up or cancellation
of errors in prices : they may be termed "subtractive" (algebraically),
that is, the percentage error of the denominator is subtracted from
that of the numerator; in practice errors are not all of uniform size
so that where errors of numerator and denominator have the same
sign, their influence is merely offset rather than cancelled.
Errors in the price ratios of the various commodities have an
influence upon the final price ratios in proportion to the importance
of the particular commodity. In general, these errors in price ratios
for the different commodities tend to offset one another in the general
average, since there is no connection in general between errors
affecting the prices of different commodities. A possible exception
to this last-mentioned relationship, however, may be found if errors
in qualities priced in a particular market tend to too high or too
low a quality, that is, if the qualities locally consumed by wageearning groups are biased, as compared with other markets, in the
direction of too high a quality : such an error might affect more or
less all the items. This is, however, one aspect of the general problem of pricing instructions and definition of items to be priced, and
they should be so drafted as to avoid the possibility of systematic
differences in quality.
EXCHANGE RATES

The data used for calculating exchange rates between countries
are taken from the figures published in the Monthly Bulletin of
Statistics of the League of Nations. They are applied in converting
the prices of the several countries into units of a common currency,
namely the Swiss franc. For this purpose, exchange rates for
2 ¡t

INTEBNATIONAL COMPARISONS OF FOOD COSTS

October 1938 are required between the currency of each country
and the Swiss franc.
Two methods of conversion are possible. If all countries were
on the gold standard, the exchange rates in normal times would
fluctuate between the upper and lower "gold points", and the gold
equivalents of the standards in the two countries would be an
appropriate exchange rate. Since, however, relatively few countries
were on the gold standard in 1938, but most of them were in 1929,
the gold equivalents of 1929 could be used in conjunction with the
percentages of currency depreciation in each country between 1929
and October 1938. The second method takes the average exchange
rates as quoted on the exchange markets in October 1938. In the
figures here given, the first method was used. 1
Two points require comment. Though exchange rates between
countries are in most cases fairly stable, in a few cases, notably
France, thé rates were in transition to new values in anticipation
of an expected devaluation of the currency. In such circumstances,
the ratios as calculated, though correct for the particular period,
would not long remain significant. Secondly, the fact that in many
countries exchanges were not free, leads to difficulties in interpreting
the results. If the exchange quotations are artificial, the ratios,
when costs of food only are compared, are equally so. Where comparisons are made between real wages in two countries, however,
neither of these difficulties affects the final results: since the exchange rate appears in both prices and wages, they simply cancel
out, leaving the ratio between the real wages in the two countries
unaffected. But for comparisons involving the cost of food only,
the difficulty arising from exchange restrictions must be borne in
mind. Among the countries for which price ratios are shown, in
1938 such restrictions were found in Austria, Bulgaria, Czechoslovakia, Denmark, Estonia, Germany, Italy, and Poland. 2
1
Owing to the method of calculating the depreciation of currencies by means
of official
rates on New York, the two methods are essentially the same.
2
Since the manuscript was completed, new figures for (ireat Britain based
on a broad family budget study have become available. They have not been
introduced into the tables, however, since to incorporate them would have
required extensive recalculations, and because the primary objective is the
exposition of method.

CHAPTER IV

SUBSTITUTIONS
In applying the price ratio technique it sometimes happens
that in a particular country a commodity included in the basket of
another country or group is not itself largely consumed and that
the price quoted is out of line with what would normally be found
if the item were more generally in demand. This may be caused by
the fact that in the particular country a cheap and good substitute
for the other article of food is obtainable and is favoured by the
local population. These considerations suggest, on the one hand,
that the price ratio technique in such a case presents a certain
artificiality of method and, on the other, that a substitution procedure in which the local product is substituted for the item consumed in the other country may give a better cost-of-food ratio.
The point of departure for the consideration of the problem of
substitutions is the discussion of the formula which expresses the
ratio of the costs of food in two communities.
In its most general form, this formula may be given by J =—^—where the standards of food consumption in the two countries are
the same. The formula gives merely the ratio of the costs, of food
in the two countries and cannot be used as a food cost ratio at the
same standard of food consumption as distinguished from a ratio of
the costs of two different standards, without a satisfactory method
for determining that the standards of food consumption as represented by the respective quantities of numerator and denominator
are in fact equivalent.
Another approach is afforded by the formula for the food price
ratio as developed above. According to the usual explanation, the
ratio P 0 = —
is a maximum for the price ratio starting with the
S po Ço
quantities of region 0 : with the new set of prices in region 1 a consumer with the habits and preferences of region 0 can make adjustments in quantities purchased in the numerator such as will mean
equal satisfactions at less expense—in other words, he can secure
an equal standard of food consumption and make savings in the
amount spent, which will then be less than 2 pi qB. (A similar
reasoning shows f\ to be a minimum, since the denominator
2 Po 9i is a maximum starting from the quantities of region 1.)
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

Both these approaches suggest the possibility of substituting
in the numerator in place of the quantities of the denominator other
quantities which will yield the same standard of food consumption
and will serve as a valid food cost ratio at the same food standard
in place of the food price ratio hitherto discussed. The discussion
of the formula suggests, in fact, that the food cost ratio is a better
measure of the desired ratio than the food price ratio itself, provided that satisfactory means are available for determining it.
The questions must therefore be examined, by what feasible methods
can such a food cost ratio be determined subject to the condition
that numerator and denominator give equal standards of food consumption, and does the result differ materially from the result shown
by the food price ratio ?
T Y P E S OF SUBSTITUTIONS

Substitutions may be considered as of two types, the first,
where there is substantial identity between the original items and
the items substituted, and the second where there is no identity
but merely approximate equivalence. Of the first type may be
considered the natural and the synthetic products of extract of
vanilla, beet and cane sugar, etc. Substitutions of one for the other
are made without question in ordinary price ratio calculations, and
they may be dismissed without further consideration here. The
only problem perhaps is on the borderline, whether two items that
are nearly alike can or cannot be considered identical for the practical purposes of price ratio calculations. Of the second type are
such substitutions as rye bread for wheaten bread, beef for mutton,
rice for macaroni, etc. A whole series of gradations can be set up
in the extent of substitutions: (1) substitution of a given quantity
of one item for a given quantity of another; (2) substitution of a
set of quantities of various items in a substitution group for a
different set of quantities of the same items ; or (3) the substitution
of an entirely different set of quantities of all items for the original
quantities.
CRITERION FOR SUBSTITUTIONS

The criterion for substitutions is the savings in cost of the
substituted over the original food items. In other words the compelling motive is price. Preference for the substituted item on
account of its giving greater satisfaction or having a higher food
value than the original item, at the same price, is the same motive
in other terms. The procedure of increasing consumption of items

SUBSTITUTIONS

when price falls, and diminishing consumption of other items when
price rises, is the application of the housewife's principle of economy
in food purchases. Of the many possible sets of quantities that may
be substituted for the original set in the numerator of the ratio P 0 ,
that set should be accepted which gives the smallest total sum and
yields the same standard of food consumption as the original set,
that is, will yield the maximum savings over the original cost.
BASIS FOB EQUIVALENCE

A crucial question is the basis for equivalence of food standards,
or for determining when the standards of numerator and denominator are equal or equivalent. Two bases may be examined: the
nutritive and the economic.
The nutritive basis accepts as a test of equivalence the findings
of nutrition experts concerning the food values of different foodstuffs in terms of their food elements—calories, proteins, minerals,
and vitamins. This can be applied to substitutions of a single item
for another, of a group of items for another group, or, with more
difficulty, to the substitution of an entirely new set of items for the
original set.
The economic basis accepts as a test of equivalence the verdict
of the individual consumers, as shown by their acts, that one set of
items yields more utilities or "satisfactions" than another set. In
effect, this question is not put to the test of any individual preferences : the preference is inferred by means of theoretical reasoning.
The exact position may be made clear as follows. With a given
sum of money in region 0, a consumer will adjust his purchases in
such a way t h a t his total "satisfactions" will be a maximum; a t
the margin, where specific alternatives are tested, his purchase of
any item will depend on whether that item will yield him more
satisfaction than the purchase of any alternative item; this depends,
therefore, upon the relative marginal utilities, to the individual
consumer, of the several items available in the local market, and
these marginal utilities of the different consumers are adjusted to
the prices prevailing in the local market. When such a consumer
with his scale of preferences adjusted to local prices in region 0
moves to region 1 with a different set of prices, he will in general,
by changing the quantities, be able to increase his "satisfactions"
in expending the sum 2 p x qa which is sufficient for him to purchase
exactly the quantities he purchased in region 0. In other words,
he can obtain equal satisfactions at a less cost and can therefore maintain the same "standard" of satisfactions as before at less
expense than S pi q0.
^

INTERNATIONAL COMPARISONS OF FOOD COSTS

Exactly similar reasoning applies to the consumer who has
adjusted his food consumption to the prices prevailing in region 1
when he moves to region 0. He also can increase his "satisfactions"
by changing his consumption to conform to the new set of prices in
region 0, and can do so at less cost than the sum 2 p 0 QiThe points to be noted in this reasoning are (1) the "satisfactions" are based upon "utilities" which are adapted to local
prices, in such a way that the marginal utilities to each individual
consumer of each "last" item purchased all exceed the marginal
utility of their money cost. They are therefore essentially economically determined. (2) When the satisfaction or preference scales
of individual consumers from the two markets are compared, there
is no basis for deciding whose scale of preferences shall govern.
(3) True comparisons of food standards should be based in some way
upon quantities consumed rather than upon many values.
These points may be made clearer by a specific illustration.
If the satisfactions derived from the consumption of 1 kilogram of
bread and 1 kilogram of milk and 1 kilogram of oranges are compared, analysis shows that the satisfactions are indicated by the
marginal utilities of the different items. If, in one market, prices
of 1 kilo, of bread = 1 kilo, of milk = l kilo, of oranges, relative
satisfactions or marginal utilities tend to become adjusted to these
prices. If then an individual with a preference scale so adjusted
buys in another market where z/± kilo, of bread equals \Y¿ kilo.
of milk equals Y% kilo, of oranges, he will shift his consumption and
his preference scale until his marginal utilities tend to agree with
this new set of prices. He will tend to consume more milk and fewer
oranges and consider that his "satisfactions" have been increased.
On the other hand, the individual whose preferences have been
adjusted to the second set of prices will, if transferred to the first
region, tend to consume less milk, more bread, and more oranges,
and consider likewise that his satisfactions have been increased.
Equal "satisfactions" thus appear to depend upon values which
vary with prices in the market.
As a result of changes in prices in a given market, quantities
consumed change. Changes in prices in the same market are
relatively easy to follow in their effects on changes in consumption.
In the case of comparisons between different markets, the exposition
is complicated by the facts that prices are different rather than
changed and that quantities consumed are different and may or
may not accord with what one would expect if both prices and
quantities referred to the same market. Assuming for ease of exposition that prices and quantities in the two markets react as if
they were in a single market, and that currencies and commodities

SUBSTITUTIONS

are identical, then if a price is reduced, the quantity consumed tends
to be increased. If more than one commodity is involved, those
with lower prices tend to have larger quantities and those with
higher prices tend to have smaller quantities.
The crucial question concerns the standard of food consumption which depends not on prices but on quantities. If some items
have increased and others decreased consumption, how are the
different items to be combined ?
If the test "satisfactions" is adopted, the different items are
combined on the basis of their marginal utilities in the market concerned, which is, at one remove, according to the relative prices of
the different foodstuffs. When two markets are involved, as in case
of price ratios, the question as to which of the two sets of marginal
utilities should govern is left unanswered.
If the test "food values" is adopted, the different items are
combined on the basis of their respective food elements and a
definite answer can be offered to the question as to whether two
standards are or are not equivalent. To take a specific case, if two
communities are identical in all respects except for their respective
prices of milk and bread, and the prices are such that in one community relatively more bread and less milk is consumed, whereas in
the other relatively less bread and more milk, the nutritive appraisal
would prefer the second, while the economic appraisal, strictly,
would appear to adjudge the preference in the light of the respective
prices of bread and milk. In practice, the economic test can be
utilised only by means of complex mathematical methods for the
solution of the problem of food cost ratios.
MATHEMATICAL METHODS

I t may be worth while, in the search for feasible methods for
calculating food cost ratios, to consider the attempts to solve these
difficulties by mathematical methods. The underlying assumptions
are that all conditions in the two communities compared are identical, except only such changes in consumption as are induced by price
differences: tastes, habits, customs, etc., are the same.
One of the most interesting of these attempts is Wald's method. 1
He restricts the analysis to the comparison in time of prices in the
same community and utilises the demand curves for each item
which he considers applicable to the two times compared. This
avoids the problem of trying to equate both "satisfactions" derived
i A. WALD, "New Formula for the Index of Cost of Living", Econometrica,
Vol. 7, No. 4, Oct. 1939, pp. 319-331.
24

INTERNATIONAL COMPARISONS OF FOOD COSTS

from consumption preference scales in two different markets.
Utilising the "Engel curves" of each item as derived from family
budget data—showing the total cost of each food item purchased as
total expenditure increases—and assuming for ease of computation
that all such curves are straight lines, he reaches a formula in which
the food cost ratio is
/ =

Pi (g° + X gi)
S Pa (go + X 5i)

where X is a fraction giving the ratio of the marginal utilities of
money in the two situations.1
This result is substantially identical with that of Bowlty's
analysis. In this result, however, the values óf the quantities of
region 1 are given more weight if they are on the whole larger.2
In commenting on this result, four points may be made. First,
though the exact significance of the simplifying assumptions is
difficult to appreciate, it appears to give a result which takes
account only of price differences rather than one which takes
account of all substitutions under the conditions envisaged above.
Secondly, the difference in practice between the results of the
substitution formula and the average of P0 and f\ depends essentially on the value of X: if X is large the result tends to move towards Pi, if small, towards P 0 , while if X is equal to 1, the result is
approximately the average of the two. Thirdly, if "Engel curves"
have to be calculated for each food item in each food area, the work
of calculating a single ratio on this plan might be so great as to
render unfeasible the method as applied to the task of calculating
an entire set of ratios between each pair of countries. Fourthly,
the method is restricted to cases where data are available for calculating the "Engel curves".
Another mathematical approach is Frisch's so-called "double
expenditure" method.3 On this basis, the quantities of numerator
and denominator of the price ratio fraction are considered equivalent where Q — 1, Q being a quantity ratio of the same type as P
is a price ratio, i.e., where
i If all the Engel curves are straight lines through the origin, the formula
is simplified into the Fisher formula. Ibid., p. 331.
2 See Ragnar FRISCH, "Annual Survey of Economic Theory: The Problem of
Index Numbers", Econometrica, Vol. 4, No. 1, Jan. 1936, p. 28; also, A. L. BOWLEY,
"Note on Professor Frisch's, 'The Problem of Index Numbers' ", Econometrica,
Vol. 6, No. 1, Jan. 1938, pp. 83-84.
3 The criterion is stated by Frisch as Sp 0 qt Spi q¡ = Spi go Spo q<¡. See
Ragnar FRISCH, he. cit., p. 29. Also "The Double Expenditure Method", Econometrica, Vol. 6, No. 1, Jan. 1938, p. 85. See also Robert M. WOODBURY, "Quantity
Adjustment Factors in Cost of Living Ratios", Econometrica, Vol. 8, No. 4, Oct.
1940, pp. 327-38.

SUBSTITUTIONS
X Po gì S p i gì

\ S po Ço S p i Co

This method, however, for the points to which the criterion
applies, i.e., where Q = 1, gives the same ratio as that given by the
Fisher formula for this point.
This, then, substitutes for an answer to the general question,
what effect do substitutions have upon the food price ratio, the
specific answer that where Q = 1 substitutions as envisaged by the
formula have no effect, or that they have the same effect upon both
ratios, leaving the final answer the same as that given by the Fisher
formula.
PRACTICAL TECHNIQUES

The use of the nutritive basis of equivalence seems to offer a
practical technique for calculating the effect of substitutions. Under
this plan, the different food items can be assigned to "substitution
groups", within which substitutions may be made on the basis of
approximate equivalence of nutritive elements. For purposes of
simplicity of statement, we may consider these substitution groups:
I.
II.
III.
IV.
V.
VI.
VII.
VIII.
IX.
X.

Wheaten bread, rye bread, (flour)1, rice, macaroni.
Beef, pork, mutton, veal, etc.
Margarine (butter) 2 .
Milk, cheese.
Fresh (leafy, green, and yellow) vegetables.
Other fresh vegetables.
Citrus fruits.
Other fruits.
Coffee, tea (cocoa, chocolate).
Sugar, honey.

Substitutions may be allowed of one item for another item
within the same group, subject to the reservation that in a particular
case some extra or supplementary items may have to purchased in
order to maintain the standard. For example, if macaroni is substituted for rye bread, a supplementary purchase of tomato sauce or
1
T h e substitution of flour for bread is in part a difference in the degree of
preparation, and represents an increase in the home baking of food : to avoid the
complication entailed by the difference in the degree of preparation, wheaten
flour may be converted into its equivalent wheaten bread (7-10ths of a kilo, of
flour—1 kilo, of bread) before testing the amount of saving possible by substitutions. Other differences in t h e degree of preparation may be involved in other
cases, e.g., macaroni for bread, b u t these are of relatively small importance.
2
Butter and margarine are obviously not perfect substitutes, since the former
has vitamins which t h e latter lacks.
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

cheese may be required to keep food values unimpaired. All such
questions must be left to the nutrition expert. For present purposes, to test the importance of substitutions in influencing the food
cost ratio, the present scheme of substitution groups may be tentatively accepted, subject however to the reservation that where
supplementary items are needed, the saving indicated by the
substitution procedure may be too large and overstate the true
saving.
Having established the groups within which substitutions may
be made, the next step is to establish a basis of equivalence within
each substitution group. This may be kilograms, as 1 kilogram of
wheaten bread = 1 kilogram of rye bread, or calories, as 100 calories
of beef = 100 calories of mutton, or in terms of other nutritive
elements. The basis of equivalence should take care of the variations in the percentage of waste, as in the proportion of bone in the
different cuts of meat, etc. The quantities and prices in each substitution group must then be converted into this new basic unit.
The limits to which substitutions are allowed are established by
taking the percentages prevailing in the numerator country as
limits. For example, in Denmark, as a numerator country, rye is
cheaper than wheaten bread and 82 per cent, of the bread consumed
in Denmark is rye. Consequently the limit of permissible substitutions of rye for wheaten bread is set at 82 per cent, in all ratios
where Denmark is the numerator country. Finally the items are
arranged in order of the prices prevailing in the numerator country,
and the amount of each item is calculated, beginning with the most
expensive, by taking the proportion (of the remaining or unallocated quantity in the substitution group as expressed in the new
units) found in the numerator country if less is consumed there than
in the denominator country or the proportion found in the denominator country if less is consumed there than in the numerator country.
These amounts are then multiplied by the prices as expressed in the
new units. The sum of these amounts times prices is then substituted
for the somewhat larger original sum in the numerator of the price
ratio fraction. 1
In each food ratio the denominator is unchanged.
SUBSTITUTIONS AND THE GROUP BASKET M E T H O D

The preceding discussion has been in terms of substitutions as
affecting the ratio between two countries, when the individual
1
T h e use of t h e average price per new unit in the substitution group, though
simpler t h a n the method described above, will not give the same results, since
it may require substitutions in some cases where costs are increased by the
substitution.

SUBSTITUTIONS

baskets are used in the calculations. With the group basket method
no substitutions are made between countries in the same group,
since for these the same basket is used. Substitutions are thus limited
to cases arising between countries in different groups, and the
amounts of savings are calculated on the basis of the differences
between the group baskets in conjunction with the price differences.
This simplification effects a marked reduction in the amount of
calculation necessary, though it gives a slightly different result
than if individual baskets were used. 1
RESULTS

Results of substitutions were examined for two cases: namely,
that of rye and wheaten bread and that of beef, pork, veal, and
mutton. In the case of bread, substitutions are made kilogram for
kilogram, since the calorie content per kilogram is substantially
the same; whereas in the case of the four meats, "equivalent beef"
units were used, based largely upon calorie content: pork was
converted into equivalent beef units at the rate of .8 kg. pork equals
1 kg. beef; veal at the rate of 1.2 kg. veal equals 1 kg. beef, and
mutton at the rate of 1 kg. mutton equals 1 kg. beef.
The net effect of substitutions involving the four meats upon
the final ratios was not great. The largest changes (affecting 4
ratios, i.e., two ratios and their reciprocals) were between two and
three per cent. ; three-fourths of the changes were of less than half
of one per cent.; twenty were between one and two per cent. The
possibility of substituting the cheaper for the dearer meats up to the
limits set by the proportionate consumption in the country where
the substitution is made according to the group basket of that
country does not produce substantial changes in the results.
The effect of substituting rye for wheaten bread (and vice versa)
is more considerable* especially in ratios between rye consuming
and wheat consuming countries. Table X I I I gives the ratios for
the different countries as calculated following the substitution procedure outlined above. Of the total of 342 ratios, 262 are potentially affected: the rest, besides the.27 intra-group comparisons, are
cases where the proportions of the two kinds of bread in conjunction
with the price difference do not permit of savings from substitutions.
1
An alternative procedure is also possible, to use the same total quantities
in the substitution group in numerator and denominator (as before) but to utilise
the proportions of the numerator country basket (instead of those of the group
basket corresponding to the numerator country) in calculating the amounts for
the numerator. With this procedure, corrections appear also in ratios between
countries of the same group.

SÎ5SSS

N O O O O
r-lCOCOOOIN

IMCOlO^OO
Oi-INr-"^

8S3S

CO I N CO CM. CN

SîgSigS

"#t~ioiocN
o i O H O œ

N O O 0 0
H O O M

^ t - IO O i-l
00 *)< • # » "-i

oo es os i—i o
o o œ o o i o

S-s
i o i—i co co »o

OOOOOSTÌ«

HMCDOS'*
THOOOOt^CO

¿ti
O Ol

i-H a s e s a r *

OS <N I N C O O
IO ( D b . C D O

M W O O H
O O O H O O

COCNTtlcNIN

o coas as
ooooot-

IO OS CO CO
OOSOSOO

Z i
U Ì C D H T ) ! CM
CD (M CC C O O

OOOiHOO
oooooosco

i-ICNTtHCDCD
t * ri* T}I CO Ti*

• * C S « w O
OS IO CD 00 CO

OOPlrtN

So-Sgg

O—ICO I M O
CO CO <N C O O

55-

t^-TjHCOOO
H H O O !

H O N O
oosoooo

C M l O i O M * >H
i-HCNi-li-l

OS CO 00
OS 00 00

CM - H CM. l-l »H

rt^^H^CM

M O O 0 0 H
N O O f O O

« » N O N
COCOt*.t~0

¡SgcSSSg

ssssg

M O O - # N
CDOOOOt^O

«'a

ssssg

§££§§

^Ht~^HCO00
>o i o co i o oo

^IrtrH-H

t~00OSO
t > CO CO CO

ooooot~©
Qg

va
H

IOOST-I

OSO

« • - H IM CO CO
t - 0 0 OS 00 I N

H N O H 1 3
- * O S Ó O S 00

C O M r t O N
cNosomoo

o ^ c o o c o
CO CO 00 t*» i-l

Oi-lt-OSi-t
r - o o oo t - I M

S Te ts< pT )o< -s<o0 äSSSS 5Sf£S^
CM (M CM i-l i-l

^H^i-(r-*CN

i-l^i-Hi-l

• C3

Si
Ï Ï CS

g E o

• a

•ß.3 S ja a
3 Cü 3 N J£
«JfQCQüQ

IS ^ ¿IN -S S

OJTJ
-• rt Cü ri
a g u a *»
3 Jâ a S ça
n .S t< o t*

3 tì 0)

0"o

¡nszZfri

Sî

& £¿?

o
c

CCOQl-Jp

SUBSTITUTIONS

Of the 262 cases, 124 or nearly half, involved changes so slight
that they did not alter the ratio when expressed to two decimal
places, and 55 more altered it only one point in the second place:
thus two-thirds of the ratios potentially affected were not altered or
were altered in only a negligible degree. Of the rest (83), 25 involved a change of two points, 14 of three points, 8 of four points t
and 36 of five or more points. A small number of cases, therefore,
involved a considerable change in the ratio. The largest percentage
change was 19 per cent. (15 points) in the case of the ratio Estonia/Bulgaria, with 16 per cent. (20 points) in the reciprocal
ratio Bulgaria/Estonia. Almost equally large was the change in
the ratios between Italy and Estonia: Estonia/Italy 19 per cent.
and Italy/Estonia 15.5 per cent. The third largest change was 18
per cent, in the case of the ratio Great Britain/Estonia with a
corresponding change in the reciprocal ratio 15.5 per cent. Estonia/
Great Britain.
An examination of these large changes shows that the question
is of greatest importance when ratios between a wheat-consuming
and a rye-consuming country are concerned, and that they are
greater when the disparity between the proportions consumed is
greater, when the quantity of bread consumed is greater, and when
the difference between the rye and wheaten bread prices is greater.
An important point is the direction of change in any given
ratio. 1
Any given ratio, P , is the average of the two ratios, P 0 and P i ,
for the two countries concerned : the substitution procedure affects
the quantities of one country in P 0 and those of the other country
in Pi. In terms of the wheat-rye substitution, assuming that the
rye price is less than that of wheat in both countries and that the
wheat consuming country is in the denominator, the effect of the
substitution is to lower P 0 and leave P i unchanged; the average of
the two thus tends to move towards P i since in general P 0 exceeds
Pi. If the rye consuming country is in the denominator, P 0 is unchanged while P i is increased, and the average tends to move
toward P 0 . 2 8
1
So far as net changes in the aggregate are concerned, since half the ratios
are the reciprocals of the other half, a change in one ratio produces an approximately equal (percentage) change in its reciprocal in the opposite direction.
2 This result may be compared with those reached by Bowley and Wald. Wald's
conclusion is t h a t the true ratio, when allowances are made for price changes and
their consequences upon the demand for the different items consumed, may be
expressed :
j = 2 Pi (go + X qrQ
S po (ç. + X g,)
where X is a factor t h a t modifies the q¡'s. This means t h a t where X is greater than
» ,
(continued on next page)

INTERNATIONAL COMPARISONS OF FOOD COSTS

•

SCOPE OF SUBSTITUTIONS

As to the scope of substitutions, theoretically, an indefinite
number of possible substitutions, each equivalent to the original
food basket (or part of basket), may be set up by nutrition experts :
of these only those which represent savings need be considered, and
in practice, only t h a t one—covering the whole food basket—which
represents the maximum savings is important. The analysis made
in the preceding section suggests, furthermore, t h a t for practical
purposes a particular proposed substitution need be considered only
if the savings are substantial, since otherwise the final ratios will
not be materially affected.
On the one hand substitutions may be limited to the locally
available items which offer inexpensive and preferred alternatives
in the original food budget, or on the other hand may be extended
over the whole range of the food budget, so as to secure 100 per
cent, coverage of the food basket by using substitutes for all items
1 the value of 1 moves in the direction of Pi, whereas if X is less than 1 it moves in
the direction of Po. In Bowley's formula as modified and interpreted by Frisch,
the condition that X is greater than 1 corresponds to the ?i's being greater than
the go's and the prices in region 1 lower: in this case the value of the ratio moves
towards Pi. In the illustration above, the price of rye bread was lower than that
of wheaten bread and more would, therefore, in theory, be consumed in the
numerator country: in this case the value of the ratio with substitutions tended
to move towards Pi. Though the mode of description of the two cases differs,
they appear to correspond in their effects.
3
Reference may be made here to the method of calculating a cost-of-food
ratio suggested by the Swedish Social Board for comparing relative costs in Detroit and Stockholm. See INTERNATIONAL LABOUR OFFICE: A Contribution to the

Study of International Comparisons of Costs of Living. An Enquiry into the Cost of
Living of Certain Groups of Workers in Detroit (U.S.A.) and Fourteen European
Towns, Studies and Reports, Series N, Statistics, No. 17 (Second revised
edition), pp. 69-70.
This method proposed that the food budget be divided into the "staple"
foods and the non-staple foods, the former including flour, bread, breakfast foods,
potatoes, peas and beans, sugar, syrup, and oleomargarine, and the latter all the
rest, including the protein-containing foods, the protective foods, etc.; that an
average price per calorie be calculated for each group, and that the ratios of these
average prices per calorie in each group be weighted by the proportion that each
group formed of the total value of the Detroit budget. In essence, this is a procedure of substitutions with two substitution groups, in which, however, substitutions are made not merely when savings can be made, but irrespective of savings
the proportions of foods consumed in the numerator community (Stockholm) are
substituted for those of the denominator community (Detroit). The calculation
was made only for the Po ratio Stockholm/Detroit; this happens to be the one
in which large savings appear; and the hew ratio for Po is found to be reduced by
24.1 per cent. This reduction is somewhat less than would have been shown had
substitutions been limited to those cases where savings would have appeared
(30.0 per cent.). Had the Pi ratio been calculated following the method of the
Swedish Social Board, it would have some been 8.1 per cent, below the original
Pi; but in theory, Pi is a minimum; following the method of restricting substitutions to cases where savings can be made, the ratio would have been raised 1.4 per
cent, above the original P\. No special importance should be attached to these
values since the substitution groups were taken with far too wide a range and the
basis of equivalence in calories only cannot well be justified.

SUBSTITUTIONS

not locally available. A broad scope, however, raises difficulties
of determining the equivalence of the substituted for the original
foods and increases the volume of calculations, especially where
special computations have to be made to provide for equivalence
by adding in the cost of supplementary items deemed necessary
by nutrition experts.
A practical compromise to reduce calculations to the minimum
is to restrict the scope of substitutions to those which lead to substantial savings.
CONCLUSIONS

The fundamental assumption of the substitution method is
that the substituted foods are equivalent in food standard to the
original foods. If this is true, there appears to be no valid objection
to accepting the results. If the substituted foods are of equal or
superior standard to the original foods and cost less, the substitution ratio would appear to be superior to the ratio obtained by
comparing prices. But if the substituted foods are of lower standard
than the original ones, the lower cost may be due in part or wholly
to their lower standard, and the result of such a substitution can
not safely be accepted, even though in a given case it may be
possible that the result is better than that afforded by the price
ratio technique.
The acceptability of the substitution method in its specific
applications thus involves the acceptability of the specific substitutions. For example, nutrition experts would probably consider
that the substitution of rye bread for white bread improved rather
than impaired the diet, and doubt would arise, perhaps, as to
the advisability of admitting the substitution of white bread
for rye. In regard to the basis of equivalence adopted in the calculations described above for the four meats, much depends upon the
actual cuts of the different meats commonly purchased in the different countries and on the proportion of lean and fat ; the calculation
was made as a crude test of the importance of these substitutions in
affecting the final ratios without implying that the equivalents used
were satisfactory from the point of view of nutrition.
There remains the question whether any elements not concerned with nutrition must be included in the appraisal of equivalence of food standards. Can the equivalence of food standards be
determined on the judgment of nutrition experts, or have utilities,
satisfactions, or preferences of consumers also to be considered?
There is no problem where nutrition experts and preferences of
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

consumers agree; in such a case, the substitution ratio can be
definitely preferred to the price ratio. In the long run they doubtless
tend to coincide. Preferences of consumers educated to nutrition
values tend towards the purchase of the nutritionally most valuable
foods for the money.
Tastes and habits of consumers in a particular market are
often formed by price conditions; the cheaper foods, preferred at
first on account of price, are presently preferred on account of taste.
To some extent the reverse effect may be found : foods may be preferred on account of their being expensive, and their consumption
favoured because it appears to belong with higher economic status.
Difficulties arise, however, in trying to harmonise preferences derived from two different markets in order to derive a satisfactory
food price ratio between them.
The whole question may be summed up as follows : the classical price ratio technique offers a method of comparing prices in
different markets which is for practical purposes substantially valid;
in certain cases, where equivalent foods can be substituted for the
original foods, the substitution technique offers a ratio that is preferable to that obtained by the price ratio technique, but its acceptability in the given case is dependent upon adequate proof that the
condition of equivalent standards is satisfied.
As a practical proposal, therefore, its use should be limited to
those cases where, on account of special local conditions, large
savings can be expected by substituting an inexpensive and good
article of food for a general item which in the local market appears
to occupy an exceptional position as a luxury item or is not available
at all.

CHAPTER V

CONCLUSIONS
In summary, the following conclusions have been reached.
(1 ) The group basket method appears to afford a satisfactory
technique for international comparisons of food prices subject to a
small margin of error.
(2 ) A satisfactory technique for grouping countries according to
similarity of proportionate consumption is developed. The number
of groups should in practice be sufficient to provide representation
of the major types of food consumption in the various countries
compared.
(3 ) A technique is presented for correction factors which substantially reduce errors.
(4) The method utilises the cost of the group basket (with correction factors) as the basis of comparison between countries in the
same group, and the geometric average of the ratios obtained by
using the two group baskets, corresponding to the two countries,
when two countries in different groups are compared.
(5 ) A technique for extending the results to other cities in the
countries included is described.
(6) A technique for extending the results to cities in countries
not included in the original analysis is presented, provided that the
new country or countries can be assimilated to one of the groups
already established. If the new countries do not belong to one of
the groups already set up, a new group or groups are required.
(7) Differences in consumption in different economic groups in
many cases do not appear to cause large differences in results. In
cases where greater accuracy is required a special technique is
suggested.
(8) Sources of error in quantities and prices are examined. The
need for careful description (qualities, etc.), of articles to be priced,
for comparable sources for price data, and for special methods for
checking up data for use in international or interregional comparisons is emphasised. The extension of the list of articles to be
priced is desirable.
24

INTERNATIONAL COMPARISONS OF FOOD COSTS

(9) In cases where the prices of particular articles are out of line
or where the item is not available, or where a cheap substitute is
locally preferred, a substitution ratio may give a more satisfactory
result than the price ratio technique. Details for a substitution
technique are worked out. It is suggested that the substitutions
be limited to cases where large savings are to be expected and where
local preferences favour the substituted item; and that they be
restricted to such as, in the judgment of nutrition experts, are
nutritionally equivalent. All ratios using substitutions should carry
notes indicating the exact type and nature of the substitution.
(10) Climatic differences in food requirements are neglected on
account of lack of data. In general, if differences in food habits and
conditions are too great, the food price ratio, the substitution ratio,
and the food cost ratio all lose their significance.
(11) In using the results, care should be taken not to place too
great reliance upon the exact decimals of the ratios, since errors
from various sources, up to two or three per cent, or even more,
may affect the figures.

APPENDIX I

RESOLUTIONS ADOPTED BY INTERNATIONAL CONFERENCES, THE COMMITTEE OF STATISTICAL EXPERTS,
AND THE GOVERNING BODY OF THE INTERNATIONAL
LABOUR OFFICE

(1) Second International Conference of Statisticians convoked by the Social Science Research Council (U.S.A.)
(May 1930).

B.—INTERNATIONAL COMPARISONS OF COST OF LIVING.

III. In order to see how far different methods of comparing
real wages agree with one another, the Committee recommends as
desirable: . . .
(d) that in view of the fact that cost-of-living index
numbers depend more on the prices than on the composition
of the budgets, and that retail prices of the same article vary
considerably according to quality, precautions be taken with a
view to ensure that, as far as possible, the comparability of
prices is maintained. It would be desirable before making comparisons that the statistical offices of different countries be
requested to furnish detailed information as to the quality of
the articles indicated in the budgets, and for any observations
to which these budgets may give rise from the point of view of
international comparisons.
(2) Fourth International Conference of Labour Statisticians
(May 1931).

(6) The International Labour Office should continue to
collect and publish statistics of the retail prices of the articles of
food consumed by working-class families in the towns for which
2 4

INTERNATIONAL COMPARISONS OF FOOD COSTS

particulars as to wages are collected; particulars should also be
collected of the prices of fuel and light in these towns. The statistics
hitherto published should be extended as far as possible and in
publishing the information the Office should give such indications
as can be obtained as to the scope of the data.
(8) In those countries in which no family budget enquiries
have been made during the last ten years, such enquiries should be
undertaken at the earliest possible date, on the lines laid down in
the resolutions adopted at the third Conference of Labour Statisticians in 1926.
(9) The wages and prices information referred to in the preceding paragraphs should be collected and published by the International Labour Office at annual intervals. The Office should
address each year to the statistical offices of each contributing
Government a questionnaire soliciting the fullest possible particulars
of wages and retail prices and rents relating as nearly as possible
to the month of October. In view of the fact that there are important seasonal variations in the prices of certain foodstuffs, such
prices should be obtained several times a year.

(11) Following upon the publication of this volume (Yearbook
including statistics of wages, rates of wages, earnings, and retail
prices, etc.), there should be prepared and published in the International Labour Review an article in which the data contained in the
published volume should be discussed from the standpoint of determining the relative levels of purchasing power of wages in the various
countries. In this article the difficulties of the procedure should be
fully and frankly disclosed. Indices on various bases should be
given, with the purpose of indicating the various points of view from
which comparisons can be made and of preventing any one figure
being regarded as authoritative for all purposes.
In view of the fact that the index numbers of purchasing power
of wages serve different national purposes, it is impracticable for
the International Labour Office to compute all the comparisons
which are possible between a series of countries and between
different occupations and industries. The series of index numbers
compiled by the Office should be illustrative of the methods by
which the data may be used for computing further series.
(13) The calculation of these index numbers by the use of
information as to working-class consumption furnished by family

APPENDIX I

budget enquiries should be made not only as hitherto on a fixed
international budget but also on the basis of various national or
regional budgets.
(14) The Conference realises that comparisons cannot usefully be made between countries of widely differing habits and
customs ; and that the closer are the consumption habits in different
countries, the more trustworthy are the comparisons likely to be.
(15) The Conference is unable to recommend the proposal
to compare countries of dissimilar conditions by progression through
countries with intermediate conditions owing to the lack of information on such conditions at the present time and to the absence
of any criterion for measuring the degree of dissimilarity among the
intermediate countries.
(16) In all comparisons between two or more countries,
however, allowance should be made for the difference in articles
consumed in each of the countries by basing the calculations successively on the list of important articles consumed in one country
and the list of important articles (but not necessarily the same
articles) consumed in the others.
(17) It is also desirable that as regards certain articles of
food for which direct comparison of prices as between one country
and another is impracticable, the International Labour Office should
explore the question of supplementing the present method of
calculation by a method in which account would be taken of the
possibilities of comparisons of prices on a basis of nutritive value.

R E COMMENDATION

The Conference recommends that the Governing Body place
the question of the supply of the data required for the purpose of
these international comparisons of wages and cost of living on the
agenda of a future International Labour Conference with a view to
the framing of a Convention binding the Governments which ratify
it to collect and supply the information at regular intervals.
It recommends that a small committee of experts representing
the competent national statistical authorities might be set up. The
function of this committee would be to assist the Office in its work
of developing and publishing wage and cost-of-living statistics and
in preparing for any future international conference which might
appear desirable in the near future.
2 4

INTERNATIONAL COMPARISONS OF POOD COSTS

(3) Committee of Statistical Experts (1st Meeting, December
1933).
RESOLUTION I

(1) As regards the cost of food, the Committee recommends
that the International Labour Office continue :
(a) its theoretical research, not only on the lines indicated
in its memorandum 1 , but also in other directions and particularly from the point of view of physiological standards;
(ò) its practical studies, and the periodical publication of
international indices of the cost of food.
(2) Before indicating some of the principles on which it is
possible to express an opinion at present, the Committee recognises
that, on the one hand, the absence, in very many cases, of recent
enquiries into family budgets classified according to incomes and
localities, and showing both the expenditure and the quantities
consumed, and, on the other hand, the defects in the statistics of
retail prices, present serious obstacles to the application of any
method of comparison.
In order, however, to enable the International Labour Office to
utilise all the data of sufficient value available in the different countries, the Committee recommends that in each case the procedure
should be that which is best adapted to the character and the value
of these data. It is nevertheless obvious that relatively summary
methods can alone be used when the data necessary for more refined
calculations are at present lacking. The following principles therefore apply only to comparisons relating to the countries in which the
statistical information available enables more elaborate methods to
be applied.
(3) In accordance with the recommendations of the fourth
International Conference of Labour Statisticians, the comparisons
should be made on the basis of national or regional budgets of family
expenditure; further, indices obtained by different methods should
be calculated and published simultaneously.
For the purpose of these index numbers, there should be included within the same "region" only those countries, parts of
countries, or towns in which the habits of consumption and the
price systems are sufficiently comparable; in the first place, the percentage of the total food expenditure represented by the expenditure
on articles of common consumption should be sufficiently high;
1
Now published as Part I of International Comparisons of Cost of Living,
Studies and Reports, Series N, Statistics, No. 20.

APPENDIX I

the further criteria suggested in the above-mentioned memorandum
may then be applied, and account taken of variations in comparability according to the level of incomes.
In principle, "regions" should be formed by towns and not by
countries; consequently, they should not necessarily be constituted
in such a manner as to cover the whole of a country; they may, for
example, include certain towns and exclude others belonging to the
same country. When, however, the available data do not admit of
the separate consideration of several towns within a country, a
single town or an average covering several towns may be taken, and
in this case, in default of appropriate data, it will be admissible to
utilise, for the purpose of weighting, average quantities relating to
the country as a whole. Geographical proximity and comparability
from points of view other than that of food consumption may be
taken into account in allocating a town to one region or another in
those cases in which it is almost or wholly a matter of indifference
to which region a town is allotted. It will be expedient to restrict the
number of regions as far as possible.
In order to reduce the calculations involved, the number of
towns should be reduced so far as possible to the capital and, where
necessary, some other representative towns in each country.
(4) In each "region" a town should be chosen as "regional
centre", well comparable with the other towns in the "region". The
Office is invited to prepare a primary system of "regions" which
will serve as a basis for its calculations. The Office might, in addition, study other systems of "regions" with a view to the ultimate
adoption of the system which would yield the best results.
(5) For the comparisons within each region, it will be sufficient
to calculate one index between each town and its regional centre.
(6) For comparisons between the regional centres a formula
should be adopted which takes account of the consumption in each
of them. Indices of this kind should be accompanied by an indication as to the degree of confidence which can be placed in them so as
to avoid any misinterpretations.
(7) For the calculation of an index relating to two towns belonging to different regions the Committee is of opinion that it is
impossible to formulate a general rule applicable to all cases. If
questions of this nature arise, the Office should select in each particular case the method of calculation which appears the most appropriate.
(8) The publication should include any necessary reservations,
and the Committee expresses the opinion that it should contain
not only the computed indices, but also the original data used in
the calculations.
- ,

INTERNATIONAL COMPARISONS OF FOOD COSTS

(4) Committee of the Governing Body on Cost of Living and
Wage Statistics (December 1930).
Extract from resolution adopted :
(1) The Committee recognises the difficulties encountered
in making international real wage comparisons and fully appreciates
the work of the International Labour Office in this domain and the
successive improvements introduced into the methods employed.
It considers however that certain modifications are necessary in
order to make those comparisons of greater value . . .
(2) . . .
(e) To avoid any errors due to differences in the selected qualities of the various commodities, the International Labour Office
should ask for a careful description of quality in its questionnaire.
(/) In order that the degree of reliability of the indices relative
to countries with widely differing customs, needs, and habits may be
judged, the index numbers should be computed and published, not
only on the basis of an average international budget, but also on the
basis of different budgets . . .
(3) . . .
(j) The method of making comparisons between countries
with widely differing conditions of living by progression through
countries with intermediate conditions should be examined and
introduced into the calculations if it should appear practicable
and advantageous.

APPENDIX II

TABLE 1. CORRECTION FACTORS TO P'„ AND P\ BY COUNTRIES
Correction factors

Austria
Belgium
Bulgaria
Czecho-Slovakia
Denmark
Estonia
Finland
France
Germany
Great Britain
Hungary

Italy
Netherlands
Norway
Poland
Sweden
Switzerland
Union of South Africa
United States
1
Country of denominator.
' Country of numerator.

2 4

Ci«

1.0694
1.0504
1.0237
.9647
1.0984

.9351
.9520
.9769
1.0366
.9104

1.0465
1.0553
1.0967
1.0432
1.0819

.9556
.9476
.9118
.9586
.9243

1.0331
1.0415
1.0402
1.0710
1.0593

.9680
.9602
.9614
.9337
.9440

1.0045
1.0426
1.0464
1.0000

.9956
.9591
.9557
1.0000

—it-ososT*

ta m

• * ! D M H O
t-- I O O S - * IO
i-H I - I T - I I - H

COT^CCt--*
I Û T C OSOCO
COTHIOCOCO

1-3

O sì

I-

II
ëî
°a

iti
lì

li
05 (>>

.ss
ss.
si
<u

_i-H —I ^ j - l

—i to to to

00 C 0 - * T J <

i-l

OS CO © i—( OS
COOOCNCOrH
» O r t O l O

N CO CO I O CO
CO CO I N T h e o
-*OSCN00OS

IN IN C t o n i o
© O S CO —l IO
INOOIOOO

CO OS 00 I N - *
CO Tfl • * (N T(<
CDCOOINCO

OS IO IO • * CD
C N 0 5 C N - * IO
0 0 - * t o t » IN

CN —I —I

—I ^H

W O I O T Ì O

CDl-OOt-O

T-H t-H i-t

i-H

H 0 0 O ' *

• * H O H

liîO—IIN
H Ö H O

o »o »-i os
O O C 5 N
© O S OS 00

05 00 IO <N
i-H CO CO IO

oor-t-co

W O I O O H
C S M H O O

i—I

—I —I —I r

OCNOOCOCD

0>t-0-*CO

t-noon

oooo—icoos

T-H

O O I O O
CO OS t— O
W H H O

O O N H O

Ü.ÜS §!§§!

inoooMo

CM — l i O C M
H O O O S

CSCOt-iO
T*-*

TJIIO

o o en oo
ocococo

CO CO C O I N - *

00 IO c o t - I N

C O O - * lOCO
O O H O O

H H O C O T H

I1ISÌ2

•*CNCO—ito
•*-*—ioOiO
00 00COO5 TJH

N H 0 5 0
t - C O C M 00
H O O »

t--(N © - *

t - - * ose»co
t - I O CO CO I O

io osco—< •*
o> oo oo t» os
— H N ^ H

co io co e o o

OS t - — 1 0 0
H O H C «

oocNcot00 IO I N CM
COCMCNf-

I N <N I N 1-H 1-1

Tt<l-<O0Tf<
TU CO IO CO

•*01THOOO)

§§§§§

eooioeo-*
OSQCOCDH
» O H I O O Ì

CO i O IO »O OS
HOOOr-l<N
t-<NOO0OtN

H co io oo
OSi-iOO-*
t-t-t-CO

IO t o t o H I N
C O t - t - C N CN
IO 00 CO CO OS

OCOCOCO
tOOOO—i
CO IO COkO

co to co os co

i-irtoo^i-i

TjHCOr-lrH
• * CO CD 00

I-—IOS00CN

©oooot-

r- oo os co oo

•*—itOtN
• * IO IO OS
OS 00 OD t -

OSINOOSOS

lOiOCOCOOO
H I O H O Ï H
•-leOlOIXN

CO CD 00 IO CO
I N 00 Til CO CO
COCM O S r - i i N

t-OioCNtt - C O O S - * ÍO

Qq r-4 i-H

I-H

©l-OO-HCM

t-ooiooeo
cot-cot-t-

8 C00D 0H0C- 0* Ot r

—icocot-o

CM t - C O CS CO
COt--*rtO
T U I N <N C O O

-*IN

O S T - Q

K 3 H N O O
00to<NOi-l
CO CO CO O OS

IOOCN o o

<NCSO0t-t>
CO IO CO 00 OS
T f C N C N IO OS

• * OS 1-1 TX 00

r-t>csoocN

« M O N O
CO ( N O — i 00
I O »-H O O O

OS CO IO OS CO
1-1 tv IN—100
IO CO IN CO OS

CO CO OS CM 00
t- CO CO OS CO
t~ CO OS 00 CO

OSO-* i-l-*
N O n i O N
lOOCOOS-t

O S - * C O - * 00
lOtOOCOO

t> oo a> oo co

CST)<-*CNOS

H W t O r t P I

ININtNH—I

M
öS

•
-

•

= S 'S O 03

'S.3 3.a

•s .au? 8 a

CÎFQ

3 . 3 d 9 »a
+= d <s Si ?

OONt-rtO

sss; SoI O H H O
I O CO CM CO
OS 00 OS t OCMCNCM
iHCQOtr

O O - i i H O

»
S tH- tH- t OS- t 1-H

r-IHH—I-H

—I CM —I —I I N

oocot»-*

—I HH —I —(

ai
-0
13

••2-13

S3 c3 <D

O ^

N S

OIOTPCOO
OOO-ftDOO
tOCMCNCSCM

CSCN i o t » t »
CO -^1 O CO CO
osioior-rn

001-Cat--*
00 r i tOCOCO
O O O r t C O

00 tO 1-1 O r-l
oo oo oo t o t o
U5 O i-i 1-1 i—I

os «* os t o t o
iot--H>ot".
to ( N M t O O

lOOOOCOH
i-i o o - * t o
O O O O O O S CO

o> t o ó o s
OCSOCM
O00O00

CO CM »O »O OS
rH C D t ~ t ~ - l O
( O H O H H

t - CM IO rH CO
OOrHOiTfl
t - - * CO CO OS

OSCNOOOIO
OSCO »O CD »O
t > O S Ó O S CO

f O t i *
OOOSCO
iHOoooo

CM t ^ O S O S OS
IO i * CO IO H I
lOOOSOOS

OS»!* I O C O O
CO CO 00 CM IO
I O CN r-( CD 00

OOCOCMlN
( N - * lOOSCN
t-ooosoocM

©lOCOCM

TtiTji r o c o c ò
oo c o - ^ i r ^ os
cot»t-t»t>

00 rH CO IO 00
ocooo^co

rH 1-H OS i—I O
( D O N N O
I O CO C O C C O

pi»
Q
O

W
H
W

ÏJ

H
M

M
œ

<!
m

¡s
co

(Ofl'JO

ot-œg
(NCSOO

oooooot»
tooio

THE
CTO BER

o
«S
oœ

fiO

H Pu
H O

<!
•4 ou
P «>,

ISsSS

O b " * c 9 0
r-i (N i o t o IM
t o CM CO t O O

t^cO-tfOrH
H^lrJIOlO
ooosooco

OOCOrH

C O - * I N OS CO
COOCOOOCO
>Or-IOOO

NOtDMOO
00 C O - * r-l CO
- * » 1 C M t o OS

N H O O l i J
•Ot-OiO-tf

C O f - C O C O rH
rH l O O IO t o
•>* r i IO CM 00

COOCMlflN.
TACO OS 1 * 0 0
rHOOCNlOr-l

u
r-J H

< fl
(J

CM 00 OS C R N
tO r-l 00 CO t o
CO CO CM COCO

C (H

SAm
O W

IOCOTJICOO

OCOrH OOS
I O OS O O OS

SA
<i H

mA
p

«3

COR
ÍES

H OS
U rH
W ¡z¡

N
O Q

FTER
KETS

< «¡
«
a,^m

i*f
•3'í

LUE

ta o
O Pu
(71 p

«•a

0
W
«1 ou
t»

CM C O l O O
OS CM t O O
f > OS OS 00

OtOOOOO
OCOt-t-CM
O O H O O

ooio>oto

O H N I Ì O Ì
O N O H O

!£§<!

Nt-CJiOOrH

r-l CO CM CM
OOOSOO

xh.cot-.to

r-l CO »O CO OS

3!cB

tOncOCOCO
0 > O r - l O O
M i t - 0 0 00 00

t^r-lO-*-*
Tticgoio-*
rHOOO-tft-

-*rHOtOCM
CO CM CO I O - *
I O CO f > t o OS

OOOr-l to
OS CM tOCC

Ttnat»Tîeo
rH 0 0 CM CJSOO
i f l - N N N

Siili

O J O K Î C Û to
COCO r-l t o r-l
i o t o t * t o os

t-oo-*-*

Ol-OSt-iO

S tr »H tl -OcOoSt oO

TtlOS00t»CM
OSCMOSTflCO
-*io>o>ooo

1-1 CM t o t O O
coiOCM n o
• * 00 OS O S O

gggl

M O J O *

omooooo
loooioca

r-C tO t - • * Tfl
n t - . 00 00 OS
- * O r i IO00

r—e» i o oo rn
r-l IN r i —I 00
t^ooosoo n

OSrl IOTH

OSOSOCM00
tO-*Or1lO
CMt-OCOlO

r H 0 0 t » C O OS
H I CO CO IO tO
NlOQOCON

( N CM CM CO 00
osCMOoOn
l O t O N I i 00

CO 00 tO 0 0
r i OSCS r H
»O »O tO i o

OSOOOSrH
lOOOSCOOS
• * O 0 0 03 00

00 " 1 IO CO OS
t»t~t~(NO
CMOSOlOCO

C O O C O 00
ot^osco
t^t^toto

O Os OSOS t »

ot»ooioco
TtUNgSTH-H
t O O O O t CM I N CM rH i-l

ooosoosto
OS I N 0 0 O tO
IN HICOiOrH

C O t O r H OS
OCMIN-*
colloco

r i r i r i r i CM

ri ri ri ri

8 It ~N 00t - i00O -00*

nooocoto
OSOOt~.CN

f - C M - i • * »O

00 Tf 00 T(l 00
COt-lOOOOS
CO tO t o t o tO'

GO
rJ
ri
(U ( I .
O

C O O T ) ( OS CM

OS OS 00 OO

cot-oso-*

CATI
COM

¡3
O

S§

mm
um

t-ooooo<N

oot-t-o

00t-OO00
r-l t o iO O tO

" * t - l O C O tO

i¿ O

OS r-l C O N
iOtOlO-tf

to to co io

•*-*oto
N t D t o IO

imu
IO*O t o o

oooot^t-

.' cä

«
m
<
H

li
Ö

•'i
:>
o
.S
s
a
os
s . 3 C3J3 a
o"3 s

<¡m«oQ

• -o

si«

03 -O
•Ga ia Sest ìa +»
'a

•s .a 2 s a
H P H Í ^ Ü Ü

2 4

^-1

• 03 .

• fe 03"O

J S «'S

03 CJ O * o

¡S M r «

TABLE 4. VALUES OF SUBSTITUTION RATIOS P'o, WITH SUBSTITUTIONS B E T W E E N R
CATION OF CORRECTION FACTORS, et, CALCULATED BY THE GROUP BASKET
18 COMMODITIES I N 19 COUNTRIES, PRICES AS OF OC

Numerator country
Denominator
country

Austria

Belgium

Austria
Belgium
3.176
Bulgaria
Czeeho-Slovakia.
Denmark
Estonia
1.4741
Finland
France
Germany
Great Britain .
Hungary
.753
2.057
Italy
Netherlands..
Norway
1.299!
Poland
Sweden
Switzerland.. .
Union of S.Africa
United States. 1.4421
1

Bulgaria

CzechosloDenvakia
mark

538
1.221 1.500
1.091

Estonia

Finland

France

Germany

Great
Britain

Hungary

.341
492
.714
.956 1.577
.647
.708

2.490
1.928

765 .6
1.529 1.3
1.618
385 1.2
1.4
553

.830
.803 1.132
.394
.646 .912

2.133
2.187
1.761

1.804 1.5
1.761 1.6
7
.843
1.418 1.2

.619 11.021
.616
.567

1.650

1.048
1.319 1.
1.214 1.0

.622
.561
.555
.485

1.6O0
1.442
1.512
1.336

1.352 1.
1.219 1.0
1.218
1.121 1.

950

1.8

1.259
1.237
.996

.9
.791

.944
.854
.856
.791

.972

.765

.783
.632

1.6

In calculating final substitution ratios, if no value for P't is given here, the value in Appendix II, Table 2 is used.

TABLE 5. VALUES OF S U B S T I T U T I O N RATIO P'i, W I T H S U B S T I T U T I O N S B E T W E E N RY
CATION OF C O R R E C T I O N FACTORS, ei, C A L C U L A T E D BY T H E G R O U P B A S K E T M
C O M M O D I T I E S I N 19 C O U N T R I E S , P R I C E S AS O F OCT

Numerator country
Denominator
country

Austria

Belgium

Bulgaria

CzechoDonslomark
vakia

Estonia

Finland

France

Germany

Great
Britain

Hungary

.315
.679
Czecho-Slovakia. 1.860
Estonia
France

.917

.819
.667

.794

2.866 1.401 1.046 1.547 1.413
2.033
.634
1.002
1.052

.402

1.328

.519

1.306 .654 .618 .722 .644
1.432 .753
.832 .706
1.682
.715
.829
1.521
.750
.555
2.254
1.111
1.014
1.238
1.121
.532
1.607
.792
1.426
.611
.703

Switzerland
Union of S. Africa
United States. .. 1.359

.670
.645

.713

.533

.437

1.004

1.2
1.0

1.205 1.245 2.541 1.547
.883
1.097

1.6
.9

.808

.469

.457

.554
.631
.715
.671
.935

.705
.568 1.186
.9
.622 1.366 .773 1.044
.908
.731
1.1
.821
.661
.8
1.216
.980
1.5
2.034
.867
.698
.8
.620
.770
.9

.573

.590

Great Britain....
Hungary.. :
Italy
Netherlands
Norway

Ita

.568

.733

.856

1.171

i In calculating final substitution ratios, if no value for P'i is given here, the value in Appendix II, Table 3 is used.

Publications of the International Labour Office
Studies and Reports, Series B, No. 33

Studies in War Economics
CONTENTS
I.

E c o n o m i c O r g a n i s a t i o n for T o t a l W a r , w i t h S p e c i a l
R e f e r e n c e t o t h e W o r k e r s . By E. F . P E N R O S E .
A general introduction to the study of war economics. It discusses
the underlying principles involved in changing from a pre-war to a
war economy, dealing with changes in productive activities, problems
of war finance and foreign trade, and methods of distributing the war
burden in the most equitable manner.

II.

W h o s h a l l P a y for t h e War? A n A n a l y s i s of t h e K e y n e s
Plan.

By E. J. R I C H E S .

A detailed summary and discussion of the proposals put forward
in Mr. J. M. Keynes's How lo Pay for the War. Various amendments
are suggested and the applicability of the Keynes Plan to other countries
is briefly discussed.
III. R e l a t i v e W a g e s i n W a r t i m e .

By E. J. R I C H E S .

An analysis of the nature and causes of wartime movements in
the relative wages of different groups of workers, and of the problems
involved and safeguards needed in attempts to limit such movements.
IV.

C o n t r o l of F o o d P r i c e s .

By A. S. J. B A S T E R .

A discussion of the objectives and limitations of food price control. The problems of administration and enforcement of price regulation and the prevention of profiteering are investigated.
V.

H o u s i n g P o l i c y a n d War E c o n o m y . By Carl Major W R I G H T .
A discussion of the place of housing in an effective defence effort.
It examines the ways in which housing policy has been adjusted to
new conditions both in belligerent and in neutral countries.

VI.

T h e Effect of W a r o n t h e R e l a t i v e I m p o r t a n c e of P r o d u c i n g Centres, with Special Reference t o t h e Textile
I n d u s t r y . By E d i t h Tilton D E N H A R D T .
An analysis of the effect of war on international trade and, through
trade, on industry in different countries. The position of the textile
industries throughout the world is discussed, as well as the way in
which the war is likely to affect the development of these industries.

199 p a g e s

P r i c e : $1.00; 4 s .

The Labour Situation in Great Britain
A SURVKY: M A T — OCTOBER

1940

Studies and Reports, Series B, No. 34
The purpose of this survey is to present a general picture of the manner in
which Great Britain adapted its administrative machinery and its labour and
social policies and practices to the needs of total war.
The survey was prepared in the London Branch of the International Labour
Office by Mr. A. D. K. OWEN, Stevenson Lecturer in Citizenship at the University
of Glasgow and Secretary of the Civic Division of P.E.P. (Political and Economic
Planning), and Mr. Neil LITTLE, a member of the Geneva staff of the International
Labour Office.
CONTENTS
The Political Background
The Machinery of Control
Economic Policy
The Regulation of Labour Supply
Employment and Unemployment
Wages and Earnings
Hours of Work and Holidays
Industrial Welfare
Industrial Health and the Prevention of Accidents
Industrial Relations
The Standard of Living
56 p p .

P r i c e : $0.25; I s .

Industrial Safety Survey
(Published quarterly in English)

This periodical is intended to act as a link between those in all
countries who are interested in problems of accident prevention, and
to make international co-operation possible.
In addition to articles by experts on specific aspects and problems
of accident prevention, the Survey contains notes on the activities of
safety associations, new laws and regulations relating to safety precautions, extracts from official reports dealing with accidents, reviews
of books and periodicals, and reproductions of new safety posters
issued in different countries.
40 p p .

P r i c e : $0.25; I s .
2 4

Labour Supply and National Defence
Under the title "Labour Supply and National Defence" the International Labour Office has just published a report originally prepared
for a meeting of government officials, employers and workers of the
United States and Canada held under the auspices of the Office
on 12 April 1941.
PART I. PROBLEMS OF LABOUR SUPPLY
Chapter I : Allocation of Man-power between the Armed
Forces and Industry.
Reservation of Workers for Production and Release of Key Workers from the
Armed Forces in Great Britain, the British Dominions, France and the
United States.

Chapter II : The Control of Employment.
Restrictions on Engagement and Dismissal of Labour in Great Britain, Francei
Germany, Australia, New Zealand, South Africa, Canada, the United States
and Japan.

Chapter III : Vocational Adaptation of Labour Supply.
Vocational Training in Schools and Special Centres, Intra-plant Training and
Upgrading, Apprenticeship, etc., in Great Britain, France, Germany, Australia, New Zealand, South Africa, Canada, the United States and Japan,
with a Section on Arrangements for Training of Indian Workers in Great
Britain.

Chapter IV: The Mobilisation of Labour Resources.
Dilution of Skilled Labour, the Distribution of Experienced Workers and the
Mobilisation of Labour Reserves in Great Britain, France, Germany, the
British Dominions and Japan.

PART II. PROBLEMS OF INFORMATION AND
ORGANISATION
Chapter V: Information.
Inventories of Labour Supply and of Labour Requirements in Great Britain,
France, Germany, Canada, the United States and Japan.

Chapter VI : Organisation.
Administrative Organisation and Special Machinery for Co-operation with
Employers and Workers in Great Britain, France, Germany, Canada, the
United States and Japan.

In conclusion the report contains a series of general principles
which the experience of the countries studied has shown should be
kept in mind in the interests both of effective policy in the present
emergency and of future readjustment to peace economy.
", . . . a useful contribution to the discussion of Canada's war effort. . . ."
The Gazette, Montreal.
". . . . the International Labour Office in what is probably one of its most
important publications of recent years, has just issued . . . a desirable
and informative volume." The Financial Times.

Studies and Reports, Series C, No. 23
245 pp.

Price: $1.; 4s.