household food consumption: the influence of household characteristics

18

Click here to load reader

Upload: p-j-lund

Post on 01-Oct-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

41

HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

P. J. Lund and B. J. Derry* Ministry of Agriculture, Fisheries and Food 7

This paper reports an attempt to identify the effects of various household characteristics on household food consumption. The analysis was based on National Food Survey (NFS) data for Grent Britain for the calendar year 1982. Consumption (per person per week) of various food items was related to each of the main household characteristics on which information is collected in the course of the Survey-location (both country/region and type of area), income (generally of the head of the household), household composition, age of housewife, housing tenure and freezer ownership. The results indicated that all these variables are important in the explanation of households ’ food consumption pat terns.

Introduction The National Food Survey is a continuous sampling enquiry into the domestic food consumption and expenditure of private households in Great Britain. Its results are reported in various forms, most comprehensively in the annual reports of the National Food Survey Committee. These include an appendix giving a detailed description of the structure of the survey and provide the most complete tabulation of the results, which are mainly expressed in terms of average expenditure or consumption per person per week. Results are presented for about 200 different food groups, with separate averages being given both for ‘all households’ and for households classified, in turn, according to each of seven different factors. However, in all but one case these factors are considered only in isolation, the exception being in the tables showing cross-classifications by household composition and income group (Tables 17 and 18 of the 1982 report). While this form of descriptive classification is valuable, it clearly does not identify the partial relationships between food consumption and each of the several characteristics. This paper seeks to do this through the application of multiple regression analysis to individual household data (expressed on a per person basis) obtained in calendar year 1982.

The authors acknowledge the assistance and comments of various colleagues in MAFF, particularly Miss S. E. Middleton. Mrs F. C. Snoswell. and M. Ullyett who arranged the extensive computations. The authors also acknowledge the permission of the Ministry to publish this paper though they are solely responsible for its content.

Ministry of Agriculture. Fisheries and Food, Whitehall Place, London SWI 2HH.

Page 2: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

42 P. J . LUND AND B. J . D E R R Y

Methodological Issues For each food+ the dependent variable chosen for analysis was ‘food obtained for consumption’. This variable is defined as purchases from all sources (including purchases in bulk) made by households during their week of participation in the Survey, and intended for human consumption either during that week or later, plus supplies obtained free of payment from an employer, as well as any garden and allotment produce actually consumed during the survey week. This variable clearly gives a better (though obviously imperfect) guide to physical consumption of each food than either the volume of purchases or the value of expenditure and, unlike the latter, does not reflect variations in unit price.

Given that the NFS covers only household food consumption, one determinant of a household’s consumption of food is likely to be the extent to which its members eat out. This has been allowed for by including, as an explanatory variable, a numerical indicator-the household’s average ‘net balance’ 7-of the extent to which the household’s members and visitors relied on its food supplies during the survey week. Second, since each household participating in the Survey does so for only one week, its recorded food consumption is also likely to reflect the time of year of its participation. Some allowance has been made for this factor by including quarterly seasonal dummy variables in the regression analysis.

Of the seven classificatory factors, four are of a qualitative nature and can only be allowed for in multiple regression analysis through the inclusion of sets of dummy variables. These are country/region (9 categories), type of area (6), housing tenure (6) and deep freezer ownership (2). The other three factors-age of housewife (7). househoid composition (1 1) and income @)-are treated in a similar way in the standard NFS tabular analyses and this procedure has been followed in this study. Table 1 lists the separate categories into which each household is classified and shows the percentage distribution of households between them: it also introduces the notation used in presenting the results in Table 2.

The reasons for adopting a dummy variable treatment of these factors include the difficulty of specifying simple functional forms for the relationships between them and food consumption. This problem clearly affects the age variable but is also relevant to income (not least because of the physical constraint on a person’s total food consumption) and to household composition, with its separate numbers of people and age/sex attributes. The dummy variable approach avoids the choice of functional forms at a cost, in terms of degrees of freedom, which can be tolerated in a cross-sectional study with so many observations (7,945). Another factor influencing the choice was the lack of numerical information on the incomes of all households; numerical data (on net household income) is available for only about 60% of the households whereas income-group data based on the head of the household (or the principal earner) is available for the full sample. Finally the adopted procedure permits a more ready comparison with the standard tabular analysis and hence whatever further analysis may be conducted it would appear to be a useful first-stage in its development.

However, some implications of the adopted dummy variable procedure, and of the specification of the income variable in particular, should not be

For NFS purposes food is defined to exclude certain items which individual family members often buy for themselves such as chocolates and sugar confectionery. Soft-drinks, ice-cream. fish and chips and other take-away meals are included if they are brought home to cat. though soft-drinks are currently excluded from the main analyses and from ‘total food expenditure’.

t For details of the construction of net balances, see Appendix A of the annual reports of the NFS Committee.

Page 3: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tab

le 1

C

ompo

siti

on o

f th

e 19

82 N

FS S

ampl

e of

7,9

45 H

ouse

hold

s B

etw

een

[he

Seve

ral C

ateg

orie

s of

eac

h C

lass

ifica

tory

Var

iabl

e

Clo

zrifi

calo

ry

1 V

orio

Me

Fu

ll D

escr

iptio

n

Sco

tland

w

ales

E

ngla

nd-N

orth

--

Yor

kshi

rc

and

Hum

bers

idc

-Nor

th

Wes

t -E

ast

Mid

land

s - We

st M

idla

nds

--Sou

th

Wes

t --

Sou

th

Eas

t/Eas

t A

nglia

Gre

ater

Lon

don

Met

ropo

litan

dis

trict

s an

d th

e C

entra

l

Non

-met

ropo

litan

dis

trict

s w

ith

Cly

desi

de c

onur

batio

n

elec

tora

te p

er a

cre

of

7 or

mor

e 3

but l

ess

than

7

0. 5

but

less

than

3

less

than

0.5

Hou

seho

lds)

co

ntai

ning

I

or m

ore

earn

ers

l24

0 (g

ross

per

wee

k) o

r m

ore

1127

but

less

tha

n 12

40

171

but

less

than

[I27

Less

than

ill

Hou

seho

lds

Ell

or

mor

e w

ithou

t an

ear

ner

less

than

~7

7

Pen

sion

er h

ouse

hold

s (as

defin

ed in

NFS

)

Abb

revi

nlio

n us

ed in

Tn

ble

2

Sco

t. W

ales

N

Y

H

NW

E

M

WM

sw

S

E/E

A

GL

C

Met

ro.

7+

3-

7 0.

5-3

<0

.5

A B

C D El

EZ

OA

P

Per

cen r

age

Dis

rrib

urio

n of

H

ouse

hold

s

1.2

4

.9

1.4

9

.8

10.3

7.

8 9

.8

9.9

3

1 0

12.6

21

.4

18.7

16

.3

14 6

15

.3

6.4

29

.6

26.7

9

.2

2.1

11

.3

14.3

Clo

ss~

ico

rory

V

onob

le

~

Hou

seho

ld

Com

posi

tion

(NB

A

n

adul

t is

a

pers

on a

ged

8 ye

ars

or

mor

e.

thou

gh 'h

ead

of h

ouse

hold

' an

d 'h

ouse

wil

arc

alw

ays

clas

sed

as

adul

ts)

Age

of

Hou

scui

lc

Hou

sing

Te

nure

Free

zer

Cor

enor

ies

Ful

l Lk

scri

prio

n

Adu

lts

Chi

ldre

n I

0 I

I or

mor

e 1

0 2

I 2

2 2

3 2

4 or

mor

e 3

0 3

or m

ore

3 o

r m

ore

4 o

r m

ore

0

I or

2 3

or m

ore

Und

er 2

5 ye

ars

25-3

4 ye

ars

35-44 y

ears

45

-54

year

s 55

-61

year

s 65

-74

year

s 75

and

ove

r

Unl

urni

shcd

: cou

ncil

Unf

urni

shed

: ot

her

rent

ed

Furn

ishe

d. re

nted

R

enl l

ice

Ow

ned:

out

right

O

wne

d: w

ith m

ortg

age

Ow

ning

a d

eep

freez

er

Not

ow

ning

a d

eep

freez

er

A b

brev

rori

on

used

in

Tobl

e 2

IA

IA I

+C

2A

2A

IC

2A

2C

2A

3C

2A

4+

C

3A

3i

A I-

ZC

3+

A 3

+C

4

+ A

< 25

25

-34

35-44

45-5

4 55

-64

65-7

4 75

+

Cou

ncil

Oth

er r

cntc

d R

ent.

[urn

. R

ent

lice

Ow

n: o

utrig

ht

Ow

n: m

ortg

ag,

With

W

ithou

t

Dis

rrib

urio

n P

erce

nro8

r

of H

ouse

hold

s

17.5

3.

1 30

.7

9.8

14

.8

5. I

I .5

7.4

6.5 I .o

2.1

6.3

20

.2

19.5

16

.5

16.2

13

.7

7.5

31.8

7.3

I .2

I .

3 24

.6

33.8

54.5

45

.5

Page 4: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

P. J . LUND AND B. J . DERRY 44

overlooked. First, the income of the head of the household may reflect social class factors rather than total household income*. Second, total household income is likely to be higher relative to that of the head of the household, or the principal earner, the greater is the number of adults (basically persons aged 18 and over). On the other hand the greater is the number of persons in the household, the lower, for given total household income, is the real income of the household after taking account of its greater needs (which vary between adults and children). Recognition of this latter point sometimes leads to the construction, or estimation, of equivalent-adult scales, both for total expenditure and for each particular good or range of goods. Adoption of this approach would naturally lead to the specification of relationships accounting for variations in total household consumption or consumption per equivalent- adult (perhaps measured differently for each item) rather than of consumption per person. It must therefore be accepted that the effects of household composition estimated from analyses of consumption per person represent a complex mixture of separate effects which more sophisticated forms of analysis would seek to disentangle. On the other hand the adopted dummy variable representation of household composition may provide a better indication of the effects of incremental changes in household size than equivalent-adult scales usually permit 7.

Although a more rigorous analysis of the effects of household income and composition might have been possible this would-given constraints on time and computing resources-have detracted from the analysis of the effects of the other variables (region, type of area, age of housewife, housing tenure and deep-freezer ownership). Variables such as these are typically overlooked, or less fully covered, in econometric analyses of household expenditures and it seemed desirable to remedy this by including them within the multiple regression rinalysis. This has been done, though the resulting number of dummy variables did impose constraints on the extent of regression experimentation which could be undertaken. Perhaps the most important of these was that the effects of each variable were presumed to be independent and additive; thus location in a particular region was allowed to increase or reduce consumption (per person) of each food by a set amount rather than by one which might vary according to, say, the household's income or composition. This is contrary to recommended best econometric practice which would allow all factors represented by dummy variables to affect both the intercept and slope coefficients. In this study the large absolute number of dummy variables makes this less restrictive approach computationally impractical, if not impossible T.

The regression analysis was conducted with two main objectives: to assess the relative importance, or significance, of each of the several factors in the explanation of inter-household variations in food consumption, and to identify the patterns of variation within each characteristic. However, the

' It is of note that the Manchester study of 1965 NFS data, reported by Thomas ef al. (1972). which defined income in terms of family net income, identified significant social class effects for total food expenditure and expenditures on most of the individual items considered (milk, particular beverages, various meats).

t For a convenient discussion of equivalent-adult scales the reader is referred to Cramer (1971), pp.161-170. Basically, equivalent adult scales may either be imposed, on the basis of prior judgements, or estimated from the data: the latter approach is however subject to identification problems on which Deaton and Muellbauer (1980) provide more recent references.

f Moreover it raises the more basic question as to why this form of specification is now considered almost obligatory when the effects of a relatively small number of separate classificatory factors are a n a l y d alongside some quantitatively measured ones. but not so when the effects of only quantitatively measured variables are being examined.

Page 5: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS 45

fulfilment of the first of these objectives is not straightforward since it presumes, first, the choice of a particular regression specification and, second, the emphasis to be given to a variable’s apparent contribution to overall explanation and its statistical significance in the conventional sense. For a number of reasons it was decided to include all the considered variables in the finally chosen regression equation; this provided the base for the assessments of both the statistical significance of each classificatory variable and the patterns of variations within them. These reasons were partly general ones which, in principle, apply to all similar regression studies (the arbitrariness of significance cut-offs and the presentational advantage of having comparable equations for all items covered) but also included some specific to this study. These were the cost of rigorously selecting ‘best’ equations even amongst the sets of classificatory variables and the illogicality of excluding a whole set of dummy variables for one or more classificatory variables but not excluding (i.e., condensing) dummy variables within each classificatory set (another possibility excluded by computational limitations).

Given this choice, the statistical significance of each classificatory variable (represented by a set of dummy variables) was examined by conducting an F- test on the difference in overall explanatory power provided by equations respectively including or excluding it (all other variables being included). This procedure was adopted because the more usual I-test of statistical significance relates to individual regression coefficients which, in this case, are those of the dummy variables representing the separate categories within each classificatory variable*. However, this test, like the I-test, is a very strict one since it only allows for the marginal contributions to the overall explanation provided by each variable, considered in turn, on the assumption that the effects of all the others have been allowed for. The consequence is that the sum of the explained variations accounted for in these tests is less than the total explained by the regression equation. An alternative examination-more strictly of ‘contribution to overdl explanation’-was therefore conducted by including the classificatory variables in successive regression equations according to a predetermined order. This was: (i) controlling factors (net balance, calendar quarter), (ii) main variables (household composition, income and region-ordered on the basis of preliminary regression experimentation) and (iii) other variables (type of area, age of housewife, housing tenure and freezer ownership-in that order).

This regression analysis was applied to data on the consumption (per person per week) of each of 40 different though partly overlapping food items and to total household expenditure on food similarly expressed. Between them the 40 food items covered 96% of total recorded food expenditure. The main exclusions from the analysis were the miscellaneous foods for which aggregation of consumption quantities does not make much sense. Expenditure (per person per week) on the 40 items analysed ranged from 0 . 7 9 ~ in the case of oatmeal and oat products to € 2 . 5 5 ~ in the case of total meat. In drawing up the specification of the items, account was taken both of the

meaningfulness of aggregating individual food codes and of the percentage of households making purchases of the items as specified during their week of participation. This latter consideration is relevant because of technical econometric problems which arise when equations are estimated in which the dependent variable regularly assumes zero values, the recording of negative

Moreover, the magnitudes of these individual r-statistics reflect the arbitrarily chosen base and arc not therefore reported. Readers may, however, be interested to note that the mythical base household was one consisting of a single high-earning young adult sharing unfurnished council accommodation in the GLC area with a deepfreezer.

Page 6: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

46 P. J . LUND AND B. J . DERRY

values being constrained by the nature of the variable being investigated*. The literature on this topic is extensive but while the problem in this particular case may be different from that normally examined (being more due to length of recording interval than non-purchase as such) it seems advisable to reduce it by grouping food codes so that the incidence of non-purchase is relatively low. For this reason the number of housholds recording purchases of each of the 40 analysed food items in shown in the table of results. The problem of zero observations clearly does not affect the other analysis undertaken-that in respect of total expenditure (per person per week) on household food.

Results The results of the regression analysis are summarised in Table 2. This shows the following for each of the 40 food items and total household expenditure on food (per person per week):

the statistical significance of each of the classificatory variables as assessed (a) in conventional terms-by excluding it alone and (b) in terms of its contribution to overall fit in the sequential ordering of the variables. the apparent order of statistical significance in (b) of household composition, income of the head of the household and region as indicated by the initial regression experimentation involving these variables. the category within each of the classificatory variables for which consumption (total expenditure) was highest or lowest, after taking account of all other considered factors. the percentage of the variation (R2) in consumption (total expenditure) explained. in a statistical sense, by the equations.

The most notable feature of Table 2 is perhaps the marked contrast between the statistical significance of the various variables (however assessed) and their overall explanatory power. Four of the variables (net balance, composition, region, age) are significant for nearly all foods and a other three (income, area, tenure) for well over half of them yet the highest R values were 0.16 for fresh green vegetables and 0.18 for total food. It must be stressed that these results are not inconsistent and certainly do not invalidate each other. Essentially, they mean that while there is a lot of unexplained variation in the consumption patterns of individual households, and also in their acquisitions between weeks, there is clear evidence that a number of factors have an important role to play in explaining the recorded differences between households. Indeed all the overall fits are clearly significant at the 99% level 7.

?

This type of problem was first explored by Tobin (1958) who proposed a hybrid of probit analysis and multiple regression analysis which was subsequently developed by Amemiya (1973). Dcaton and Irish (1982) have recently considered the question in the specific context of household budget surveys in which the zero observations may be due either to people never buying the commodity in question or simply not doing so in the survey period. Their analysis and evidence (partly based on the UK Family Expenditure Survey) while suggesting that ‘tobit’ is not an appropriate model for analysing Engel curves when zeros are present, does not indicate which alternative would be better (p.23).

t Cramer (1971) remarked that ‘the decision-making process of the individual consumer defies complete description’ (p. 146). He also showed how the apparent ability, as indicated by R’, of particular forms of Engel curves to explain household data on expenditure on various foodstuffs varied considerably according to whether or not the data were grouped (and the extent of grou ing) without much affecting the estimated income elasticities. It should also be

particular, since household sire clearly has a larger part to play in the explanation of total household consumption than consumption per person, the R2 obtained in the present study cannot be compared directly with those obtained in the earlier study of NFS data by Thomas er al. (1972).

noted that R ? will be affected by the precise specification of the dependent variable. In

Page 7: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tab

le 2

Sum

mar

y of

Reg

ress

ion

Ana

lyse

s of

the

Con

sum

ptio

n of

Indi

vidu

al F

oods

and

Tot

al F

ood

Exp

endi

ture

(Hou

seho

ld F

ood,

Per

Per

son

Per

Wee

k)

Stat

istic

al S

igni

fica

nce

asse

ssed

: (a

) in

con

vent

iona

l ter

ms.

by

excl

udin

g on

ly t

he r

elev

ant

vari

able

fro

m t

he r

egre

ssio

n.

(b)

acco

rdin

g to

con

trib

utio

n to

ove

rall

fit

in l

arge

ly p

rede

term

ined

ord

erin

g-w

ith

appa

rent

ord

er o

f im

port

ance

of

hous

ehol

d co

mpo

sitio

n,

regi

on a

nd i

ncom

e gr

oup

indi

cate

d.

Sign

ific

ance

Lev

els

indi

cate

d as

: **

, 99%

; +,

97.5

%;

*, 95

%;

+ , 9

0%;

NS.

Not

sig

nifi

cant

at 90%

leve

l. B

road

Pat

tern

of

Eff

ects

indi

cate

d by

sho

win

g ca

tego

ries

with

hig

hest

and

low

est c

onsu

mpt

ion/

expe

nditu

re.

all o

ther

var

iabl

es b

eing

hel

d co

nsta

nt

Tota

l WUIC

m

eat (

73)

Cod

er 3

1. 3

6. 4

1

Food

ifem

. (as

reris

ks

Rel

atin

g ro

Tab

le 3)

. PL

of

all h

ouse

hold

s pu

rcha

sing

item

dur

ing

(hei

r su

rvey

wee

k.

NFS

Foo

d C

odes

.. .. S

ign.

(a)

NS

S

ign.

(b)

N

S H

igh

QI

Low

Q2

Fear

ure of

Res

ulrs

Bal

ance

.. .*(I) 2A

3+

A

3+

C

*. S

ign

(a)

*+

Liqu

id m

ilk'

(97)

Si

gn (b

) C

odes

4-6

+ .. ..

.. NS

N

S(3

) *

W

NS

sw

G

LC

65

-74

Sco

t. O

AP

7

+

< 25

I S

ign.

Si

sn.

(a)

Ib\

\:

I:

:

Tot

al m

ilk &

cre

am (9

9)

Cod

es 4

-17

Hig

h .-,

I Low

Che

ese.

(6

9)

Cod

er 2

1, 2

3

.. S

ign.

(a)

'+

Sig

n. (

b)

.. S

ign.

(a)

NS

S

ign.

(b)

B

ecf &

vea

l. (5

6)

Cod

e 31

H

igh

I Low

.. ..

Sig

n. (a

) S

ign.

(b)

M

utto

n &

lam

b.

(28)

C

ode

36

Hig

h I Low

Por

k. (

33)

Cod

e 41

.. .. S

ign.

(a)

N

S S

ign.

(b)

N

S

Cln

rriji

caro

ry

Van

able

r

Hou

seho

ld

Inco

me

Are

a A

ge

oj

Com

posi

tion

Gro

up

Op

e

Hou

sew

fi

.. ..

.. ..

IA

Wal

es

3+

A 3

+C

G

LC

<

25

.. I

.. I

.. I

.+

I

..

+i3)

*+

(a

NS

.. "(I

)

.,+::+c

I Ei

I OAAP

I "7

:' I :::

:

Hou

sing

Te

nure

.. .. R

ent

frcc

Ren

t. lu

rn.

.. .. R

ent

free

Ren

t. fu

rn.

.. .. R

ent.

furn

. C

ounc

il

NS

N

S

Oth

er rented

Ren

t fre

e

NS

NS

Oth

er re

nted

R

ent f

rcc

NS

NS

O

ther

ren

ted

Ren

t. fu

rn.

NS

NS

Oth

er r

ente

d R

ent f

rcc

I

Free

zer

Rz

With

W

it ho

u t

+ W

ith

With

out

With

W

ithou

t

6t

h

I '02

With

out

With

W

ithou

t

With

W

ithou

t .. With

W

ithou

t I

8 5 P rn I

0 B U 8 5 5 c z -I I

rn

Z

r

C

rn

Z c: rn

I

0 - 7 % z, n I 3 r

0

c: I

P 21 5 rn E i

v? Q P 4

Page 8: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

T8b

lc 2

(co

ntd.

)

Food

item

. (o

sm

d

trhtiw

to

Tab

k 3)

,

prr

rhp

dru

item

&tin

# t

he

u~

y

mk

.

NFS

Foo

d C

odu

&fM

&h

un

un

cook

ed.

(60)

co

dc 5

5

Ic o

fdl

Lonr

eLdd

c

Poul

try,

uncooked. (

35)

coda

73.

77

Oth

er m

ut

and

mea

t W

oduC

II.

(89)

c

h

6.5

1, 5

8-71

7a~a

PI

Tot

al mu:

(96)

co

da

31-9

4

Fish

' (6

5)

Co

da

1001

27

tw

(72)

co

de 1

29

But

ter'

(47)

C

ode

135

Mar

gari

ne'

(48)

co

de 1

38

TO

I~

1.11

(so)

CD

du

135

-148

Cla

stili

cato

ry V

aria

bles

R2

F

ealu

no

f R

anlu

N

et

Hd

dd

In

com

e A

rm

Age

of

Ho

uin

; F

rm

~

scor

on

Wa

n-

Cor

npar

ition

G

roup

T

yw

Hou

sewi

fe

Tenu

re

+ ..

NS

NS

t

.. Ns

Ns

.07

NS

*'(I)

"(2)

Y

3)

A

N

S

:? %

3tA

3tC

S

cot.

OA

P

3-7

<23

Rm

t. f

urn.

W

ithou

t

.. ..

.. ..

.. S

b.

(a)

Sip

. (b)

1A

N

W

<0

.5

65-7

4 Other

rent

ed

With

+ ..

+ ..

.02

NS

NS

N

S

NS

"(1)

-0)

NS

Ns

NS

W

all3

El

GLC

45

-54

Oth

n rm

led

With

3

tA 3

+C

Y

H

OA

P

0.5-

3 75

+

Ren

t. lu

rn.

With

out

.. ..

.. "(

2)

2A

Sign

. (a)

si

rn. (

b)

Hia

h Q

2 Lo

w

QJ

Sign

. (a)

N

S Si

gn.

(b)

NS

H

iJI

QJ

Low

QI

Sign

. (a)

N

S Si

gn. (

b)

NS

Hii

h 4

2

Lnw

Q4

Sign

. (I)

NS

S

n (

b)

NS

NS

.07

.. ..

.. ..

.. ..

.. ..

.. ..

.*(I)

"(2)

C

7

+

35-4

4 C

ounc

il W

ithou

: **

(I)

IA

.. 3

+A

3

+C

W

ales

O

AP

<

0.5

75

+

Ren

t. fu

rn.

With

..

.. ..

*. ..

+ *.

.05

NS

.. ..

.. Ns$)

"(2)

-+

A

G

LC

55

-64

Oth

er r

mtc

d W

ith

"(1) lA

3+

A 3

+C

S

cot.

OA

P

<0

.5

<Z5

Rat

. fu

rn.

With

out ..

+ ..

.05

NS

.. "(I

) 'W

Y

3)

'+

..

NS

.. ..

.. ..

.. ..

01

IA

N

El

Met

ro.

55-6

4 0t

h-

rent

ed

With

Lo

r*

42

3

+A

3+

C

Wal

es

OAP

<0

.1

<25

Ow

n: M

~t

pt

e W

ithou

t R

iih

*+

NS

.. *+

N

S .0

6 ..

NS

*+

"(2)

"(

3)

'+ ..

.. .. N

A

<

0.5

45

.54

Oth

er re

nted

W

ith

+ ..

NS

NS

3+

A 3

+C

Y

H

D

7+

<2

5 R

mt.

furn

. W

ith

**(I)

..

Sign

. (a)

N

S S

in. (

b)

NS

H

igh

42

Lo

w

QI

Sign

. (a)

N

S S

b.

(b)

NS

Hig

h QI

Lo

w

93

S

ip. (

a)

NS

Sign

. (b

) N

S H

igh

42

Lo

w

QI

sw

. (a)

N

S S

ip. (b)

NS

0

2

1A

3+

A 3

+C

SW

O

AP

0 5.3

75

r R

at.

fun

. W

ithou

t

*V

) <:

.5

45-5

4 R

mt

lret

W

ithou

t IA

W

aln

A

NS

NS

.07

.. ..

.. ..

.. ..

"(1)

"(2)

+ ..

.. ..

.. ..

+ ..

+ .M

.. ..

.. **

(I)

*W

N

W

D <

0.5

45-5

4 Rc

n: f

ret

With

..

N,Y

:)

4+

A

sw

A G

LC

7

5+

O

wn:

Mor

tgag

e W

ithou

t

.+

NS

*+

NS

.06

.. .. ..

NS

NS

.. *.(

I)

"(3)

N

S(2)

NS

..

.. IA

Wal

es

El

<0.

5 45

-54

Ren

t Ir

e

With

2

QI

3+

A 3

+C

S

cot.

A

7+

15

+ O

ther

ren

ted

With

out

?J !- z U f 2 0 ?

!- P

*) -c

Page 9: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tabl

e 2

(cun

ld.)

Fuo

d rt

rm.

lasl

nlr

k n

latm

g t

o T

abk

11.

R of

ail

hour

rhoi

ds

purc

hasi

ng r

um

dur

ing

rhrr

r su

rwy

wtk

N

FS F

ood

Cod

es

Cla

ssifi

caio

r> I

urr

able

s

R2

F

rulu

rr of

Rrs

uirs

H

ou

rrh

dd

In

com

e A

rea

Agr

01

Hou

sing

F

mlr

r &

uson

E

abn

cr

Com

posi

rion

R

rgio

n

Gro

up

Ty

pe

Ho

wrw

~fr

T

rnu

rr

Sug

ar'

(53)

C

ode I50

Su

gu

k p

rcsc

rvn

(62)

C

ode

1501

54

NS

NS

.0

6 ..

.. ..

.. ..

.. ..

.. ..

.. Si

gn. (

b)

'+

*'(I

) "(I

)

<O

J 61

-74

Olh

cr r

cn

id

Wiih

oui

WM

"(

2)

IA

.. S

ign.

(a)

Met

ro.

<25

Ow

n: M

orl

gq

c

With

H

igh

QJ

LO

W

Q2

1+

A

l+C

SW

E

l

NS

NS

N

S

.08

.. ..

.. ..

.. ..

.. ..

Sign

. (b

) "(2)

**(I

) *'(

1 )

..

Sign

. (a)

IA

WM

O

AP

<O

5

65-7

4 R

cnl

lrec

W

ith

It

A l

+C

SW

E

l M

etro

. <2

5 O

wn:

Mor

tnan

c W

itho

ut

Hig

h QJ

L

OW

02 ..

.. Si

pn. (

a)

Sign

. (b)

1 2

I *'

1 H

igh

Fresh

irec

n v

cget

abks

. (73)

Cod

cr 1

62-1

71

I L

ow

Pota

toes

' (6

9)

Cod

er 1

56-1

61

I 1

'1

O

ther

fre

sh

.. ..

I

--

NS

.. .. +

NS

.05

NS

.. .. N

S NS

..

.. ..

.. "(

3)

"(2)

"(I

) S

ign.

(a)

Si

gn. (

b)

IA +

C

N

El

7*

45

-54

Rcn

l Ir

e W

ithw

l IA

EM

O

AP

G

LC

75

+

Ow

n: M

ortg

agc

Wit

h H

igh

w

LO

W

Q2

.. "(1) 2A

2.4

4+

C

Olh

cr p

ro

m4

w

cgcl

ablc

s* (

77)

Cod

er la4

202

To

~d

vc

gct

ibln

(97

) C

oder

156

208

Fres

h fr

uit'

(73)

C

od

a 2

1023

1

"(3)

Ehl

O

AP

75

+

.. ..

.. ..

.. ..

.. ..

t ..

.09

"(2)

**

(I)

"0)

IA

.. ..

.. M

etro

. <2

5 C

ounc

il

Wil

hovt

..

Sign

. (a)

N

S Si

gn.

(b)

Hig

h Q

2 L

OW

Ql

2A

4t

C

scot

. E

l <

0.5

75

t

Own:

outr

ighi

W

ith

NS

N

S 08

'+

'

*+

"(3)

**

(I)

N

Mct

ro.

45.5

4 R

cnt

lrec

With

S

EO

I. O

AP

G

LC

75+

Ow

n: M

orlg

agc

Wit

hout

"(3)

'*(

I)

L.

.. ..

.. ..

.. .. "(2) 2A

Sign

. (b

) ..

.. Si

gn.

(a)

Hig

h QJ

Lo

w

42

2.4 3

c

.. ..

.. ..

.. ..

.. ..

.. ..

.. ..

.* .I

I SE

A

G

LC

43

-54

Rcn

i. (u

rn.

With

"(

2)

IA

Sign

. (b)

..

.. Si

gn. (

a)

Hig

h x:

ir

)r

iA

r

N

nA

p

1.1

4

1s

I

"W

I.

't

Ow

n: o

utri

ght

I Cou

ncil

Rcn

t. lu

rn.

Cou

ncil

I **

I

.I6

I i

ih

I

Wit

hout

With

out

Wit

hout

.. ..

.w

NS

N

S .. ..

.. .+

..

.. Fr

ozen

vcg

ctib

lcr.

(32)

Sign

. (b)

..

NS

C

odes

201

-208

Sign

. (a)

I L

ow

Page 10: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tn

bk

2 (c

onid

.)

.. .. R

cnl.

lurn

. C

ounc

il .. ..

.. ..

.. ..

.. ..

othe

r fr

uir'

(53)

Si

nn. (

b)

NS

'W

"0)

**(I

) co

da 233-246

Hu

h

Ql

LO

W

Q1

A

sign

. (1)

IA

SE

2A

IC

W

ak,

O

AP

.. .. .I

3

Wit

h

Wilh

oul .. ..

.I0

Tol

d lr

uil

(82)

C

okr

2102

48

To

id b

red

(94)

co

des

251-

261

.. ..

.. ..

.. S

in.

(a)

Sig

n. (

b)

Hig

h I

I **

I 27!

I :< 1 *.

:I' I

LO

W

OA

P

.. ..

.. ..

.. *Y

3)

"(I

) "(2

) S

ign.

(a)

N

S

Sig

n. (

b)

NS

H

igh

01

IA

NW

C

Whi

le b

rcd

. (7

6)

Cod

n 25

1-1J

4

&ow

n P

wh

dcm

cd

bra

d'

(39)

I--

Cod

es 2

55.2

56

L 1

.. ..

.. ..

**(I

) *'(2)

**(3

) S

ign.

(a)

N

S

Ca

ko

61 b

irui

ls.

(81)

S

ign.

(b)

N

S

Cod

es 2

67.

2702

77

Hig

h @

Lo

w

Q3

IA

N

ZA I+

C

Wal

es

E2

Sig

n. (.)

Sig

n. (

b)

Hia

h

Sig

n. (

a)

Sign

. (b

)

Oat

s an

d 0.1

pr

oduc

1r.

(5)

Cod

e 28

1

.. ..

.. '+

-0

) **

(a

-(I

) ..

3t

A I

-2C

W

M

D 2A

2C

sw

E

l ..

.. ..

.. ..

'*(I)

**(I

) 'Y

2)

IA

NW

2.4

4

+C

W

M

C

t N

S

NS

N

S

+ N

S

'*(2

1 "(I

)

IA

SC

OI.

E2

3t

A l

+C

EM

C

.. ..

Sig

n. (

a)

Sip

. (b

) H

igh

QI

Low

Q

l

olh

cr b

red.

(45)

C

ode

263

I rm

I

dl

I

I I

3t

A I

tC

I

SW

I E

l

flour

' (2

1)

codc

264

I L

OW

I

02

GL

C

15-4

4 7

+

75 +

.. ..

.. 1

.. 45

-54

j y ..

0.5-

3

.. 0

.5-1

75

+

i;5 1 N;

hklr

Ll.

.. 7+

45

-54

GL

C

75 +

I 6% <iS

hl

clro

. 25

.14

.. .I

1

.. C

ounc

il W

ilhou

l O

wn:

Mor

igag

e I W

iih

I . .-

'+

NS

.O

1 R

cnt.l

rcc

1 Wg;i 1

Cou

ncil

Rcn

l. lu

rn.

Page 11: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tabl

e 2

(coo

ld.)

Food

item

. (a

sIer

tsk

rel~

iin

g 10

Tabl

e 3)

, 41 of

nll h

ouse

hold

s pu

rcha

sing

ilem

dur

ing

thei

r su

rvey

wee

k.

NFS

Foo

d C

odes

Bre

akfa

st c

ncal

r* (

41)

Cod

e 26

2

Oth

er c

erea

ls.

(56)

co

des

28s-

301

Tot

al c

erea

ls (9

8)

Cod

er 2

51-3

01

Tea

. (S

O)

Cod

e 304

Cof

fee.

(3

1)

Cod

es 3

01.3

09

Tot

al b

evcr

ager

(64)

Cod

es 3

04-3

13

Tot

al h

ouse

hold

food

ex

pend

iture

Cla

sslJ

ical

ory

Varia

bles

RZ

F

miu

re of

Res

ulis

N

et

Hou

seho

ld

Inco

me

Are

0 A

ge o

f H

ousi

ng

free

rer

Bal

ance

C

ompo

siri

on

Reg

ion

Gro

up

Type

H

ouse

wife

Te

nure

+ N

S

NS

N

S

NS

N

Y3)

"(

2)

"(I

) N

S

NS

.. .. N

S .0

1 ..

*+

Sig

n. (

b)

.. +

IA +

C

EM

A

0.

5-3

35-4

4 R

ent.

furn

. W

ithou

t S

ign.

(a)

Hig

h 0

3

3A

scot

. O

AP

G

LC

< 2

5 C

ounc

il W

ith

Low

QI

+ +

NS

N

S

.. .. ..

NS

N

S

-02

*+

IA

Sco

t. A

G

LC

.. S

ign.

(a)

N

S

Hig

h Q

s Lo

w

Q2

Sig

n. (

a)

NS

Sig

n. (

b)

NS

H

igh

Q3

LO

W

QI

Sig

n. (

b)

NS

*+

"(

3)

**(I

) '+

(2)

'+

<25

Ow

n: M

ortg

age

With

out

2A

4+

C

WM

O

AP

0.

5-3

75+

Oth

n re

nted

W

ith *. ..

.I0

N

S

NS

.. ..

.. N

S

NS

..

**(I)

"(

2)

"(3)

NS

..

.. IA

N M

etro

. 45

-54

Ren

t fr

ee

With

out

3+

A 3

+C

SW

E

l 0.

5-3

75+

Ow

n: M

ortg

age

With

+ *

0.

.. N

S

l+

64-7

5 R

ent

free

W

ith

+ N

S

10

**(2

) **

(I)

"(3)

..

*. ..

.. ..

.. S

ign.

(a)

N

S

Sig

n. (b)

Hig

h Q

2 Lo

w

Q3

IA

NW

D

2A

4

+C

sc

ot.

El

<0

.5

<25

Ow

n: o

utrig

ht

With

out

.. ..

.. ..

.03

NS

Sig

n. (

a)

NS

Sig

n. (

b)

NS

NS

'.(I)

NY

3)

Y2

)

NS

H

igh

Q2

Low

Q3

.. N

S

.. .. IA

S

cot.

A

7+

35

-44

Ren

t. fu

rn.

With

ZA

4

+C

W

ala

O

AP

G

LC

75

+

Cou

ncil

With

out

NS

7+

64

-75

Ren

t fre

e W

ith

.. ..

.. **

N

S .0

9 ..

.. *+

N

S

NS

*(3)

+

(2)

NS

**

(I)

Sig

n. (

b)

YS

.. S

ign.

(a)

IA

WM

2A

4

+C

sc

ot.

El

<0.

5 <2

5 O

ther

ren

ted

With

out

Hig

h QZ

Lo

w

Q3

.. N

S

*+

**

NS

*+

.I8

.. .* ..

NS

.*

.. ..

.. ';I'

*'(2)

A

G

LC

45

-54

Oth

er r

ente

d W

ith

Sig

n. (

b)

*. **

(I )

Sign

. (a)

IA

3+

A

l+C

SW

O

AP

<

0.5

< 25

Cou

ncil

With

out

Hig

h Q

2 Lo

w

QI

Page 12: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

52 P. J. LUND AND B. J . DERRY

Table 3 summarises the statistical significances of the classificatory variables, under both assessments, in the explanation of the differences in per capita consumption of the 30 non-overlapping food items (totals excluded) and in per capita expenditure on household food. The difference between the two methods of assessing significance is greatest in the case of income, which appears much more significant under the sequential inclusion approach (b) than in the more conventional test (a). This difference probably reflects a particularly strong relationship between this variable and the others in the set.

The exact attribution of effect to particular factors will, to some extent, be influenced by the measurement and specification problems which have already been discussed. These are particularly likely to affect the apparent effects of household composition (because of the differing per capita needs of adults and children), income (of the head of the household/principal earner only) and age of housewife (likely to be related to both the ages and hence needs of children and household income). Thus, it is perhaps not surprising that per capita expenditure and consumption tended to be highest (other things being equal) in single adult households and lowest in the large households with several children. The notable exceptions were for carcase meat, poultry and fresh green vegetables (all highest in adults-only households), and margarine, potatoes and breakfast cereals (lowest in adults-only households). However, the household composition effect may largely explain the marked difference between the position of pensioner households* as shown by this multiple regression analysis and as recorded in the conventional classifications. Thus, although pensioner households typically record above average household food expenditure they are shown by this multiple regression analysis to have (other things being equal) the lowest per capira total expenditure and consumption of many foods including meat, fish, eggs, potatoes, fresh and frozen vegetables and fruit. While this does not mean that the actual consumption of old-age pensioners is below the national average-or below their nutritional requirements-it does confirm the view that the relatively high per capita household food expenditure of pensioner households reflects their typically small size, the paucity of children within them and their greater than average reliance on household food supplies (i.e., fewer meals out).

Not surprisingly, the highest income group (A) had the highest per capila expenditure on food and consumption of many individual foods-with the notable exceptions of margarine, sugar (and preserves), potatoes, white bread and tea. Another marked set of differences is to be observed in the case of income group D (low earning families) which had high consumption of margarine, sugar, ‘other’ processed vegetables, white bread and tea and low consumption of butter and fresh green vegetables. These differences may have implications for forecasting as also may the differences according to age of housewife, especially if these reflect vintage effects rather than age (and associated family circumstances) as such. For most food items the general pattern was for per capita consumption to rise with the age of the housewife. often peaking (as for total expenditure) at the 45-54 years age group, before declining again. To a large extent this pattern may reflect the associated differences between the numbers of adults/children and the equivalent-adult ways of measuring household size and between total household income and that of the head/principal earner. The main interest therefore lies in the notable departures from this general pattern. Younger housewives seem to

* For the purpose of the NFS, pensioner households are defined as ones in which at least three-quarters of total income is derived from National Insurance retirement or similar pensions and/or supplementary pensions or allowances paid in supplementation or instead of such pensions.

Page 13: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS 33

Table 3 Summary of Statistical Significance of lhe Classificatory Factors in the Regressions for 30 Separate Foods (Asterisked in Table 2) and for Total Food Expenditure

Season

Net Balance (a) (b)

Household (a) Composition (b)

Region

Freezer

Statistic01 Significonce

In the 30 separate individual food

consumption equations

99% - 8 9

29 25

21 25

21 2a

13 26

IS 13

21 24

15 15

8 a -

17!4ll - 3 I

- 1

1 -

3 -

1 2

3 5

1 I

2 3

2 2 -

_.

NS - 16 13

I 3

s 3

2 I

7 I

8 7

6 3

I2 12

16 16 -

In total food expenditure equation

99% 99%

99% 99%

99% 99%

NS 99%

99% 99%

99% 99%

99% 99%

NS NS

91.5% 97.5%

have a marked preference for processed and convenience foods such as ‘other’ meat, frozen and other processed vegetables and ‘other’ cereals (which includes rice and pasta) while older housewives consume more milk, carcase meats (particularly lamb), butter and tea.

Three factors-housing tenure, type of area and freezer ownership-were each significant for relatively limited groups of foods while season was generally insignificant. Leaving aside the relatively small tenure groups, households living in unfurnished council accommodation seem to have more clearly identifiable food consumption patterns than do owner-occupiers, both with and without a mortgage. In particular, they consume more ‘other’ meats, processed vegetables, white bread and tea and less cheese, fresh fruit and vegetables, brown bread, breakfast cereals and coffee. The type of area comparisons were also noteworthy in some cases, with the consumption of milk and fresh green vegetables increasing with rurality and that of ‘other’ meats and frozen and other processed vegetables declining according to the same characteristic. In part this pattern reflects the greater abundance of free or cheap supplies, particularly of milk and fresh vegetables, available to rural

Page 14: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

54 P. J. LUND AND B. J . DERRY

households, an availability which is reflected in their lower total food expenditure. Freezer ownership seems associated with high levels of consumption of carcase meat, poultry, fish, fresh and frozen vegetables and fruit, but a low consumption of white bread and cakes. This result may however simply be a reflection of this characteristic’s close relationship with household income.

The final set of differences are those between Wales, Scotland and the (standard statistical) regions of England. These differences, which often occasion some interest, are set out for selected foods in Table 4. This table provides three different estimates of the percentage difference between the consumption of each food in a particular region and the average GB level of consumption. The estimates labelled (a) are based on the results of this multiple regression analysis while those labelled (c) are based on the standard NFS classificatory analysis. The comparison between these is not however straightforward since the form of averaging used in the derivation of (c) is different from that which would be consistent with (a). To overcome this problem the averages (b) have been constructed so that comparisons between (a) and (b) reflect the influence of other factors in the multiple regression analysis while those between (b) and (c) reflect different forms of averaging*. In fact it will be seen that all three sets of figures are very similar, indicating that the allowance for the effects of other factors (and the precise form of averaging) is not very crucial in the case of regional comparisons. In particular it means that the practitioner’s interest in the variation in consumption patterns between regions largely coincides with the theorist’s interest in the specific influence of region.

The regional patterns which are shown in Table 4 could provide the basis for endless comparisons of regional characteristics though they may also reflect the precise locations-within each region-surveyed in 1982. They may also provide the basis for some speculation about future developments. Thus, while the Scots consume less lamb, pork, bacon and ham, poultry, yellow fats, fruit and vegetables, brown bread, flour, breakfast cereals and tea (but more beef, ‘other’ meats, eggs and oats) than their fellow Britons, it may be the relative similarity of their milk consumption which is of greater interest.

Conclusions-The Scope for Further Analysis The study reported in this paper is perhaps best viewed as a multivariate equivalent of the simple tabular classificatory analyses published each year in the annual reports of the NFS Committee. The value of this form of analysis is perhaps best indicated by the case of pensioner households; the study has indicated that their relatively high (per capita) household food consumption and expenditure can be explained by their associated characteristics (small household size, few children, few meals out). The study has also indicated that variables often excluded from conventional demand analyses may be significant, at least for some foods. These factors include housing tenure, type of area and age of housewife, though account should be taken of the relationship between at least the latter and other relevant factors (e.g., the ages of children and numbers of earners within the household). Differences

* Denoting consumption of a commodity by the ith household in quarterj by eij and the number of people in that household by nu where i = I ... h, a n d j = I ... 4, where h, is the number of households in the survey in quarter j .

Page 15: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tab

le 4

D

etai

led

Res

ults

for

Ind

ivid

ual F

ood

Item

s an

d St

anda

rd S

tatis

tical

Reg

ions

IPer

cent

age

diff

eren

ces

in p

er c

apita

con

sum

ptio

n (t

otal

exp

endi

ture

) be

twee

n re

gion

and

Gre

at B

rita

in a

s as

sess

ed:

(a)

on t

he b

asis

of

the

cond

ucfe

d m

ultiv

aria

te a

naly

sts.

(b

) on

the

bas

is o

f th

e av

erag

e pe

r cu

pira

con

sum

ptio

n of

eac

h ho

useh

old

surv

eyed

dur

ing

the

year

-the

aver

age

mos

t co

mpa

rabl

e w

ith (

a).

(c)

on t

he b

asis

of

the

aver

age

of t

he s

epar

ate

four

qua

rter

ave

rage

s of

per

cup

ilu c

onsu

mpt

ion

of p

erso

ns c

over

ed b

y th

e Su

rvey

-the

co

nven

tiona

l N

FS

anal

yses

.

I- vo

d

Mil

k

Che

ese

Bee

f and

vea

l

Mu

ito

n a

nd l

amb

Por

k

Bac

on a

nd h

am.

unco

oked

Po

ult

ry,

unco

okcd

Oth

er m

eat a

nd

mea

l pr

oduc

ts

Fish

""IP

S + 10

+ 14

+

12

~ I2

-

7

- I2

- I5

-

13

- 10

+ 21

+II

t 10

- 17

- I

5 -

1Y

+ 27

+ 32

t 32

1 In

+

17

t II

-8

-

14

- 1

2

- 2

4 - 2

4 -

26

c,

I I

+I

-

I

+z

+I

-

7

-7

+ 25

+

18

+ II

- 54

-

54

- 50

- 53

- 56

-

53

- 13

-

IV

-in

-H

-

I2

-in

+ II

+

I6

+ IY

-5

-

7

-8

- I4

-

I3

- I2

- 22

- 2s

- 23

+s

-

I

+I

- 40

-

40

- 1

7

-7

-

13

- II

+ I3

+

I5

t 15

-I

-

4

-5

+ IY

+

25

+ 27

+ 26

+ 27

+ 1

2

-7

-

6

-7

- I4

-

I4

-1s

+ IY

+

I6

+ 16

- 15

-?

I

- 27

+5

+

2

+ I2

+ IY

I

IY

I IY

7)

-- -

25

- 24

+6

+7

+ 14

+ 21

+ 25

+n

-5

-

5

-7

-7

-

7

-7

-6

- I

?

-0

+ IY

+ 2

2 +

30

- 23

- 27

- 25

+ 29

4

ZY

I 30

II

-

1

+I

+ I3

+

16

+ I5

-0

+

I

-1

,011

tori

! l!d

/fl/

ldS

+J

t

6

+5

+f

i

tY

+

I3

-1

-4

-

3

6 -?

I

23

+m

i

3

tY

3 4 7 n I0

-7

3 --

5 4

- It)

-

I2

.- Y

+.I

t5

+

.l

+5

+

s

+I

- 17

-2

1 ~

25

tn

+

n

+ I3

+ 17

+

I4

+ 12

+ 13

t I

3

17

18

+

7

+ 10

-2

+

I

-3

-2

-

2

-5

.So1

,1h

I4 P

si

+3

+

.I

+5

-1

+

3

+2

+ II

+ I3

+

I1

+ 22

+

I2

tn

t 4f

i + 4

7 +

34

- In

-

IS

17

7

--

9

-0

-n

-

10

- 10

-12

- II

- 10

+?

+

I

+I

I 10

+

10

+II

6 +

I1 *

L

iY

In

t I6

il

t

7

+7

- I6

-

17

- I6

t5

I 10

+

I0

-7

-I

1

- II

-I

-

I

-1

Page 16: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

Tnb

k 4

(con

ld.)

r

Tol

d fr

uii

Whi

te b

rud

Form

of

CW

IlpaliS

OA

W

ak

+7

+ 1

2 +

I3

+ 49

+%

+ 5

1

- 18

-

7

- 17

+ II

+a

t I9

+o

-

0

+3

-1

+I

I +

6

t4

+

8

+8

-I

-

II

- II

-9

-22

- 19

-4

+o

-

3

+I8

+ I5

+ 1

3

+7

+

8

+II

-8

-

13

- I3

-I1

- 12

-9

+4

+

I

+2

-4

+7

+

8

- 54

-61

- 61

-9

- 19

- 25

-6

-

10

- 14

- I9

-

3

-I

-7

- 2

3 - 2

5 -

I

+ I4

+ I4

No

rth

+m

+ 21

+2

0

+ 13

+ 14

+I

-4

-5

-8

- 10

-1

1 -

10

+ II

+ 1

7 + 2

1

- 14

- I5

- 10

i8

+

4

+6

- 34

-40

- 32

i 25

+ 36

+ 36

-I1

- 19

- 19

-2

+

7

t9

Sta

York

s an

d H

um

knid

e

+5

+

6

+l

- I

8 -

I7

- I9

+ 22

+ 1s

+ 2

8

+I

+

2

+3

+2

+

J

*s

-

6

-5

-

3

-0

-2

-

I

- I4

- 20

- 2

1

+ I2

+I5

+ I6

-4

-

9

- 1

0

-7

-

4

-I

ard

Sta

tistic

al R

Nor

th

wes

t

-I

+

o

+o

-

7

-6

-7

+M

+ 26

+ 19

+s

-

I

+I

+ 10

+ 17.

+ I

2

- 23

- 22

-22 - 10

-11 - 10

-9

- 1

6 - I2

+1

+ I

4 + 1

4

- II

- I3

-1

s

+ I3

+ I8

+ 20

'on

Ep

rr

Mid

land

s

-4

-5

-4

-4

-5

-6

-5

+

o

+2

-3

+

I

+7

-8

-

7

-3

+ II

+IS

+ I5

- 12

- 12

-9

-1

-4

tl

+ I5

+ 12

+ II

-6

-9

-

9

+I

+

I

t2

wrs

r M

idla

nds

-4

-2

-

2

+l

+

6

+6

-3

+

2

+6

+ 23

+ 24

+ 23

+8

+ I4

+ 10

+I

+

3

i4

-3

-

3

-2

+3

+

2

+o

+

5

+ I4

+ 10

-7

-1

1 -

12

+ 16

+ J2

+ 29

Sout

h w

est

-9

-

8

-6

-9

-8

-3

- 21

- I5

- 16

- 1

3 -

1

-6

+3

+

5

+7

+ 24

+3I

+ 2

9

+8

+ I

I + 1

2

+3

+

4

t6

-8

-

17

- 19

+4

+

6

t6

- 1

7 - 2

0 -22

Sour

h E

pn

/ E

ast A

nglia

-3

-

6

-8

+I

+

o

+3

-I

-

7

-6

-5

-

1

-9

-7

- I4

- 1

7

+ I4

+ 10

+ II

+4

+

6

+6

+ I4

+ 23

+ 21

-7

-I

S - 1

4

+ II

+

20

+ 13

-7

-I

S -

I7

Page 17: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS 57

m - n = O D = - a t ~ 2 % - - a O P O - c n - o n + + + I l l I I 1 I I I 4 I + I I I + + + + + + - -

- * a - - t N - N 3-n T - N c c n n n n n t - I l l + i I I + + I l l I I + i l l I + + I l l -

eas= a = - - 0 . 0 - - - = N O z c m O N C c l - o

+ + + I + I + + + + I 1 + + + + + + I l l + + + - -

o m 0 C - D D ~ m n - s n m - - o m e - - o - N - o

( 1 1 + + + + + + I I I + + + + + + + + + + + + n n n - -

Page 18: HOUSEHOLD FOOD CONSUMPTION: THE INFLUENCE OF HOUSEHOLD CHARACTERISTICS

58 P. J. LUND AND B. J. DERRY

between regions also appear to be highly significant and in this case the analysis has served to confirm the patterns portrayed by the conventional classificatory analyses of NFS data.

It must, however, be acknowledged that in a number of respects this study could be improved upon and extended. The most obvious directions for improvement and development are perhaps:

the utilisation of the other NFS data: on net household income (which, although reducing the usable sample of Survey households by about 40070, may provide a better guide to pure income effects); on the age (particularly of children) and sex of the household members; on the occupation and industry of earners; on price movements during the calendar year; and on the extent of free supplies available to the household;

(ii) the utilisation of data for other years, whether to increase the sample size, check the consistency of patterns between years or examine trends in them;

(iii) the adoption of particular functional forms with respect to the potential numerical variables (income, household composition), whether or not derived from some underlying theory of consumer demand. Linked to this might be an allowance for specific forms of interaction between variables-e.g., do the same regional differences apply at all income levels?

While it is unlikely that all of these possible lines of development (and there must be more) could be tackled in any one study it seems clear that NFS data provides considerable scope for detailed analysis of household food consumption patterns and trends. It will therefore be of interest that the MAFF is making the data tapes for individual calendar years available, for research purposes, through the ESRC Data Archive as well as selling up-to- date data, at various levels of aggregation and classification*.

(i)

References Amemiya. T. (1973). Regression Analysis when the Dependent Variable is Truncated Normal,

Cramer. J . S. (1971). Empirical Econometrics. Amsterdam: North-Holland Publishing Company. Deaton. A. and Irish, M. (1982). A Statistical Model for Zero Expenditures in Houehold Budgets,

University of Bristol Discussion Paper No. 821128. Deaton. A. and Muellbauer, J . (1980). Economics and Consumer Behaviour. Cambridge:

Cambridge University Press. National Food Survey Committee (annual reports). Household Food Consumpion and

Expenditure. London: HMSO. Thomas, W. J. et al. (Currie, J . M., Hamid Miah, M. A.. Moore, S. A., Rayner, A. J . .

and Stewart, 1.) (1972). The Demand for Food: An Exercise in Household Budget Analysis. Manchester: Manchester University Press.

Tobin, J . (1958). Estimation of Relationships for Limited Dependent Variables, Econometrica,

Econometrica, 41, 997 - 1022.

26, 24 - 36.

* Further details, including a brief note outlining the range and cost of information available, may be obtained from: National Food Survey Branch, Room 419. Whitehall Place (West), London SWI A 2HH. (Tel. 01-233 5088).