Statistics on household consumption expenditure are used for a

Document Sample
Statistics on household consumption expenditure are used for a Powered By Docstoc
					             Some Uncertainties in Household
            Consumption Expenditure Statistics
                                      A. Sylvester YOUNG
             Director, Bureau of Statistics, International Labour Office, Switzerland

                                           05-09 September 2005 1

           Abstract. The aggregates used for household consumption expenditure statistics for the
           compilation of CPI, national accounts and in welfare analysis are conceptually and in
           measurement terms different. These differences are examined in this paper. Some are inevitable
           and should be taken into consideration when the aggregates are produced. Some others are not yet
           resolved satisfactorily and more work needs to be done on them. National statistical offices should
           take into account the prime purpose of the survey and so decide which of these aggregates should
           be the official series, while offering the possibility to researchers and international organizations to
           have access to the other(s).

1.         Introduction

Statistics on household consumption expenditure are used for a variety of purposes. The
resolution on Household Income and Expenditure Statistics (HIES Resolution) adopted
by the 17th International Conference of Labour Statisticians in December 2003 [1]
recognizes that these statistics can be used, among others, for assessing the economic
well-being of households and individuals; determining weights for compilation of
consumer price indices (CPI) and the corresponding basket of goods and services;
compiling sequence of accounts for the household sector and reconciling national account
estimates with micro-level data. In a survey of countries’ practices carried out in 2003
[2], countries rated these as very important uses of their household expenditure surveys,
in the following order: CPI (89% of countries), welfare analysis (65% of countries) and
national accounts (61% of countries).

The HIES Resolution recognizes that, depending on the purpose, there may be important
conceptual and measurement implications for the production of statistics on household
consumption expenditure. These arise with respect to the treatment of some items of
consumption: durable goods, own-produced goods, own-produced services as well as
services received as transfers from government, from non-profit institutions serving
households (NPISH) and from other households. Also, the choice of an efficient survey
design depends on the purpose of the statistics.

2.         Consumption Expenditure Aggregates

The HIES Resolution proposes two household consumption aggregates. The first,
referred to as household consumption expenditure (HCE), is defined as “the value of
consumer goods and services acquired, used or paid for by a household through direct

    Paper presented to the Conference of Commonwealth Statisticians, Cape town, South Africa

monetary purchases, own account production, barter or as income in kind for the
satisfaction of the needs and wants of its members.” The second, actual final
consumption (AFC), is “the sum of its household consumption expenditure and the value
of consumer goods and services acquired or used by the household through transfers from
government, non-profit institutions or other households.” Thus AFC is an extension of
HCE to include some consumption items received as transfers. Furthermore, the
resolution recommends that AFC is the more appropriate aggregate for welfare analysis,
as it takes into account all consumer goods and services available to households for the
satisfaction of the needs and wants of its members. On the other hand, HCE is the more
appropriate aggregate for the compilation of CPI, especially the monetary components of
HCE ([3], §1.164, pp. 19).

The same 2-aggregates approach is used in the System of National Accounts (SNA) [4],
except that the equivalent of AFC in the SNA does not include transfers of services from
other households. However, probably due to measurement difficulties relating to social
transfers in kind from government and non-profit institutions (STIK), not many countries
currently produce statistics on AFC from household surveys.

The argument in this paper goes beyond these recommendations for HCE and AFC. It
proposes that HCE should be valued differently depending on whether the household
consumption expenditure aggregate is to be used for welfare analysis, CPI compilation or
national account estimates.

Recently there have been two proposals about the relationship between national account
estimates of household consumption and HCE/AFC from household surveys for the
analysis of poverty. The first aims at ‘reconciling’ the value of AFC from household
surveys to that from national accounts [5] and the second to using the ‘more reliable’
national account estimates of AFC for welfare analysis [6]. The approach in this paper is
not inconsistent with the former, depending on how HCE is valued, while it rejects the
latter because of the conceptual basis underpinning the national account estimates. There
are other reasons for rejecting the latter proposal by Bhalla [6], as discussed in the paper
by Havinga et al [5].

3.     Timing and Measurement of Consumption Expenditure

The value associated with HCE depends on both the timing of the registration of
consumption expenditure and the conceptual approach to its measurement. Consumption
can be assumed to take place at the time of acquisition, use or payment for the goods or
services. Often, all three occur at the same time and so the distinction is irrelevant. Since
services are usually consumed on acquisition, in general consumption expenditure on
services should be registered at the time of acquisition or at the time of payment. For
goods, acquisition occurs mostly simultaneously with payment. Even for credit purchases
in which acquisition occurs before payment or the completion of payment, the purchaser
has already incurred a liability to pay at the time of acquisition. So the choice for
registering consumption expenditure of goods is between the time of acquisition and the

time of use. The HIES Resolution recognizes that the appropriate choice should depend
on the purpose for compiling the aggregate.

The value of consumption expenditure can be determined on an expenditure basis or on a
consumption costs basis [7]. In the case of the former, this is the purchase value (amount
paid or obligated to be paid) for the goods and services or the estimated value of goods
that are own-produced. For the latter it is the estimated value of the service flow from the
goods or the value of the actual services.

4.     Non-durable goods and services

The timing of the registration of consumption expenditure for own–produced goods,
especially food items can vary. In terms of its price relevance (CPI), importance should
be given to the quantity produced (i.e. acquired) irrespective of the quantity actually used,
while for welfare analysis it is the latter which is more important. In many developing
countries it is in fact the quantity actually consumed that is recorded in expenditure

There is hardly any difference between the purchase value (expenditure basis) and
consumption costs of non-durable goods, since they are completely consumed at or
shortly after their purchase. Also, the value for actual services is the same under both
approaches, as services are consumed at acquisition. Therefore either measurement
method could be used for these goods and services. On the basis of ease of measurement,
the HIES Resolution recommends using the expenditure basis for them.

5.     Owner-occupied Dwellings and Durable Goods

For durable goods that are short-lived (less than five years) and not very expensive, the
value obtained from an expenditure basis and that from a consumption costs basis would
not be very different [8]. However, for major durable goods and owner-occupied
dwellings (OOD), the values obtained from use of the expenditure basis and the
consumption costs approach could be quite different. Thus a choice between these
approaches has to be made for these items.

The choice recommended in SNA is the expenditure basis for all durable goods and the
consumption costs approach for OOD. The SNA justifies the latter choice on the basis
that dwellings are capital purchases used to provide services to their owners during the
reference period ([4], §9.58, p.211). So it is the estimated value of these services that
should be taken as their consumption expenditure. By definition, durable goods also
repeatedly provide useful services to a consumer over a period longer than the reference
period. They are not directly consumed by the consumer but used as intermediate goods
in the production of these services. Also, durable goods are treated as fixed assets in the
capital expenditure account of enterprises ([4], §10.7, p.218), including household
enterprises, and when leased. Nevertheless, the SNA recommends using the expenditure
basis to value the consumption expenditure of durable goods.

Several reasons are advanced for this different treatment of dwellings and durable goods
([9]; [3], Chapter 14, Box 14.3). Households usually perceive ownership of dwellings as
an investment, unlike ownership of durable goods. Very few households can purchase
their dwellings directly from out of their pockets, while many can do so for durable
goods. Some financing scheme is usually necessary for the former. The pattern of access
to dwellings (ownership, renting, etc.) can vary across countries and time, a variation that
is unique compared to access to most durable goods. Other reasons, less compelling, are
the longer lifetime and much higher costs of dwellings.

Although this different treatment of OOD and consumer durables has been a long-
standing criticism of the SNA ([10], §3.3, p. 272), most countries continue to use this
SNA approach for their national accounts. Admittedly, a change would have far reaching
implications for national accounts, both for the size of nominal GDP and the rate of
growth of real GDP [11]. Also, almost all countries use this combined approach for
deriving weights in the compilation of their CPI. The HIES Resolution recommends
continuing to do so for the sake of consistency with SNA and these country practices
(([1], §§33, 34). However, recently, the point has been made that especially for the
compilation of CPI to be used as a cost of living index, the consumption costs approach
was conceptually preferable for durable goods as well as for dwellings ([3], §23.145,
p.441; [12].

It has been argued that treating all durable goods as fixed assets producing non-financial
services for consumption by households is in fact the only approach consistent with
economic consumption theory. Durable goods do not in themselves produce utility
directly, that is no benefit or welfare is derived from their direct use [13]. So conceptually
this is the approach to be used for welfare analysis. Moreover, a major reason for
favoring HCE over household income when measuring the current material well-being of
households is the relative smoothness of the former over time. Using the purchase values
of expensive durable goods, such as cars, would make HCE lumpy and so detract from
this argument ([14]; [15], §130, p. 31). Also, durable goods could be used both in the
production of household enterprises (as fixed assets) and for household consumption. So
a uniform treatment could be beneficial. Using the consumption costs approach for
durable goods in household consumption would facilitate apportioning the corresponding
expenditure between these two uses. The recommendation in the HIES Resolution is that,
for purposes of welfare analysis, the consumption costs approach is preferable for durable

There is some reluctance in adopting this proposal for durable goods, as it could
introduce a vast number of imputations into the computation of HCE ([3], §1.158, p.18).
However, as noted earlier, by limiting the approach to only major durable goods the
number of imputations required will be greatly reduced. The HIES Resolution
recommends using this method for welfare analysis and for the compilation of weights
for a cost of living index ([1], §34). Major durable goods may be identified on the basis
of a long expected lifetime (over 5 years) and relatively high value, for example a car or
cooker but not socks or a hammer. This approach is also recommended in the World
Bank Handbook ([16], p.103).

6.     Valuation: OOD and Major Durable Goods

When the consumption costs approach is used, the value of expenditure has to be
estimated. This could be done using rental/leasing equivalence or the user costs approach.
The rental/leasing approach uses the rental/lease value of an equivalent dwelling or
durable good in the rental/leasing market. This presupposes that there is a rental/leasing
market for equivalent durable goods or OOD, which may not be so for OOD in certain
geographic areas and for many durable goods. A way round this is to use subjective self
assessment of the rental/lease value by the owner. The rental/leasing approach is widely
used for the valuation of OOD ([3], Footnote 9, 23.21, pp. 422), with about 70% of
countries in the ILO 2003 survey [2] indicating that they used this method. There are also
some other drawbacks in using this approach, one of which is that it may overestimate the
expenditure value of the housing services [15]. While there are increasingly rental/leasing
markets for some durable goods such as vehicles, the difficulty is finding such markets
for durable goods of the various vintages that may be found in households. It is therefore
unlikely that this method could be used for estimating the service flow for durable goods.
Objective assessment for OOD also still poses a challenge for rural areas of some
developing countries, where ownership is the norm and rental markets for housing hardly

The second method for estimating the value of OOD and durable goods is the user costs
approach. This attributes to the value of the dwelling or durable good a portion due to the
use of the services flowing from it (i.e. consumption) and a portion that is ‘investment’,
i.e. the opportunity cost of the capital tied up in the dwelling or durable good. The value
depends on rates of depreciation, appreciation, interest and taxes as well as a current
market value for the equivalent new dwellings or goods ([3], §§23.22 – 23.33, pp422-
424). Current market value could be obtained through self assessment of the owner,
although Frick & Grabka [17] observe that in a study in the US, homeowners tended to
overestimate the value of their dwellings. The user cost approach, especially for OOD,
has not as yet been successfully operationalized. Attempts to do so have so far proved
difficult with results that are implausible, volatile and incomparable across countries
([10], pp.267-269). Iceland has used a very simplified version [18] and the US Bureau of
Economic Analysis uses another version for its national accounts imputation for OOD
[19]. For durable goods, traditional accounting depreciation methods – reducing balance
or straight line depreciation – could be used ([12], 292, p.68; [16], p118). Although these
methods are in fact versions of the user costs approach ([3], 23.55-23.61, pp. 427-428),
they are easier to apply for durable goods as these do not appreciate in value and have
reasonable expected lifetimes.

7.     Non-market services

As mentioned earlier, transfers of individual services from governments and NPISHs to
households (STIK) are important both for welfare analysis and for national accounts
purposes. They are as yet not considered as key to CPI compilation. While the estimated
total value of each of these transfers is obtainable from national accounts, the estimated
value at the individual household level cannot be directly determined from household

surveys, except through self-assessment by households. Some allocation rules are
required to attribute the expenditures of governments and NPISHs to individual
households or to determine the households to be self-assessed. Possible rules are take-up
or entitlement to the goods and services, which can be deduced from survey data. Take
up could disfavor some groups, e.g. the aged, who tend to have greater needs for health
services. Entitlement could lead to double counting of expenditure for those households
who, though entitled to the services, purchase them directly from the market. Also, the
appropriate value to use for the government expenditures is still an open question ([5],
§2, para. 1).

Services from unpaid household work still have problems of ‘acceptance’ as valid
consumption items and of valuation. A vigorous case has been made for their importance
in a complete assessment of national production, and satellite accounts have been
produced for many countries, with assertions that the value of such production could
range from 30% to 70% of GDP [20]. Some countries have even produced an assessment
of ‘extended private consumption’, which measures the consumption of all market and
non-market goods and services. Although methods for estimating their values have been
proposed based on time use data, there is no agreement as yet on which is most
appropriate. Their importance for assessing and understanding the living standards of
households has also been argued. There is no clear indication as to exactly what effect
their inclusion in actual final consumption will have on welfare analysis. The suspicion,
however, is that it would lead to a less unequal expenditure distribution since poor
households consume greater volumes (values??) of such services than rich households,
although this is done at risk to their leisure time. See Young [21] for a review of these

Transfer of services from other households are increasingly becoming important,
especially the transfer of care services for grand-children from grand parents (or older
relatives) to their children living in separate households. However, although this means
that they are relevant for welfare analysis, determining the volume and value of these
services would be even more challenging than for unpaid household services.

This lack of agreement on valuation methods is the main reason advanced in the HIES
Resolution for these services to “be excluded from household consumption expenditure
and from actual final consumption until such time when the valuation of these services is
based on agreed principles.” ([1], para. 50)

8.     Gross/Net Values

Certain items such as expenditures on second-hand goods, non-life insurance premiums
and gambling expenditures can be recorded gross or net of sales, claims and winnings
respectively. The recommended treatment for second-hand goods in SNA and in the
compilation of CPI is to record sales as negative expenditures, which tantamounts to
recording expenditures net of sales ([4], §9.31, p.207; [3], §3.127, p.53).Also, for non-life
insurance premiums and gambling bets, both SNA and CPI regard only the service
component of these premiums and bets as consumption expenditure. These service

components cannot be recorded directly from household surveys but can be approximated
using premiums/bets net of claims/winnings ([3], §§10.156-166, pp.199-200 & §3.51,
p.44). Thus, while for national accounts and CPI purposes these expenditures could be
recorded net, for welfare analysis the preference is to record these expenditures gross,
reflecting the actual payments made by households ([1], paras. 38, 43 and 44).

9.     Survey Design

The household consumption aggregates required for CPI weights are total expenditures
by product item(s) for the reference period, say a year, so as to attribute share of annual
expenditures to item groups. The important requirement is to have a good representation
of expenditures on each product item over the year. Households are used as the medium
for accessing these expenditures. Thus an efficient design depends only on having a
representative sample of households for each time unit, say each month of the year, to get
efficient estimates of these expenditures over the reference period. It is not necessary to
have good estimate of each household’s own annual expenditures for this purpose. The
expenditure surveys used by most countries for their CPI weights are of this type (73% in
the ILO survey [2]), probably because developed countries use this survey almost
exclusively for CPI purposes only, using income statistics for welfare analysis.

For welfare analysis such as poverty and inequality analysis, interest is in total household
expenditure for each household for the reference period, so as to determine distribution of
expenditures. Thus good estimates are needed of the individual annual expenditures of
each household. An efficient design requires not only a representative sample of
households for each time period but also a good estimate of each household’s own annual
expenditures. To minimize recall errors, the latter would require repeat visits to the same
household over the reference period to adequately cover variations in expenditure
patterns of the household over the year. There is less emphasis on the product items
themselves, except possibly for household expenditure on food items and on some basic
non-food items. Detailed product lists are mainly to ensure completeness of the recorded
expenditures. Living standards surveys are usually of this type.

The difference is particularly important with respect to seasonal product items and those
that are purchased frequently, but not necessarily each month. For CPI, the representative
sample of households each month would yield a representative sample of expenditures on
all product items that month. So seasonality is not an issue, nor are products not
purchased each month. For welfare analysis, a household that is selected in the sample for
a particular month would have expenditures from its diary only in that month. Even if
diaries are not used, the recall period for frequently purchased items, especially food, is
usually not more than a month. So the household’s expenditure on off-season items
would not be captured, nor on items that were not purchased that month for one reason or
the other, e.g. earlier bulk purchases. The household’s annual expenditure would
therefore be biased if the monthly value obtained is simply annualized, unless this value
is assumed to be typical of its monthly expenditures for the other months of the year.
Level estimates for groups of households may even out these biases, if the groups are
well-represented across the months. This is not the case at the individual household level,

which would lead to households being misplaced on the distribution scale. A repeat visit
to the same household each major season would be required for an efficient estimate of
the annual expenditure of the household. The alternatives are either to take the risk of
recall errors from asking for purchases for a longer reference period than one month for
such items or to estimate the expenditures for those unrecorded items that are sometimes
used by the household. Both are less efficient.

10.    Conclusions and Recommendations

The consumption expenditure aggregates that should be produced for the purpose of CPI
compilation, national accounts and welfare analysis are measuring different concepts and
doing so in ways that are different (See Annex). These differences and the best treatment
of them need to be acknowledged and discussed. This could lead to solutions and/or
agreements on some issues and so improve the comparability of the statistics.

A pertinent point is that the production of different aggregates, all referred to as
household consumption statistics, could confuse users and affect the credibility of
national statistical offices. The recommendation is that depending on the prime purpose
of the surveys, the corresponding aggregate should be the official series but that analysts
should have access to the other aggregates if necessary. This would also be necessary for
cross-country comparisons of household consumption aggregates.

11.     References

[1]ILO (2003), Resolution concerning household income and expenditure statistics,, Geneva
[2] ILO (2003) Methodological questionnaire on Household Income and Expenditure Statistics, Sources
and Methods, Vol 6,, Geneva
[3] Consumer Price Index Manual: Theory and Practice. (2004) (ILO, IMF, OECD, UNECE, EUROSTAT,
and World Bank);, ILO, Geneva
[4] System of National Accounts (1993). (EUROSTAT, IMF, OECD, United Nations and World Bank.),
Brussels/Luxembourg, New York, Paris, Washington, DC.
[5] Havinga, I, Kamanou, G. and Vu, V. (2005), Bridging the Poverty Measures based on National
Accounts and Household Income and Expenditure Survey, 55th Session, International Statistical Institute,
[6] Bhalla, S. (2002), Imagine There’s No Country: Poverty, Inequality and Growth in the Era of
Globalisation, Institute for International Economics, Washington DC
[7] Garner, T.I. and Short, K.S. (2002), A new ‘official poverty measure’ for the United States:
Experimental measurement and analysis, 27th General Conference of the International Association for
Research in Income and Wealth, Djurhann, Sweden,,, p.16
[8] Diewert, W.E., (2001), Harmonized indexes of consumer prices: Their conceptual foundation, Joint
meeting of the Centre for Economic Policy Research and the European Central Bank on Issues in the
Measurement of Price Indices, Frankfurt, p.64
[9] Franz, A., Ramprakash, D. And Walton, J. (1998), Statististics on the distribution of income,
consumption and accumulation of households (DICAH), EUROSTAT, Vienna, London and Luxembourg,
§2.1.11, p.30
[10] Moulton, B.R. (2004), The system of national accounts for the new economy: What should change?
Review of Income and Wealth, Series 50, No. 2.
[11] Fraumeni, B. and Okubo, S. (2001), Alternative treatments of consumer durables in the national
accounts, BEA Advisory Committee, Washington, DC
[12] ILO (2003), Resolution concerning consumer price indices, , Geneva, para. 15 & 16
[13] Johnson, M., McKay, A.D. and Round, J.I. (1990), Income and expenditure in a system of household
accounts: Concepts and estimation, Social Dimensions of Adjustment in Sub-Saharan Africa, World Bank,
Working Paper No. 10, Washington D.C.
[14] Meyer, B.D. and Sullivan, J.X. (2003), Measuring the well-being of the poor using income and
consumption, Journal of Human Resources, XXXVIII, Supplement, pp.1180-1219.
[15] ILO (2003), Household Income and Expenditure Statistics, Report II, Seventeenth International
Conference of Labour Statisticians., Geneva
[16] World Bank (2000), Designing household survey questionnaires for developing countries. Washington
[17] Frick, J.R. and Grabka, M.M, (2003). Imputed rent and income inequality: A decomposition analysis
for Great Britain, West Germany and the U.S., Review of Income and Wealth, Series 49, No. 4, pp.513-537
[18] Gudnason, R. (2003), How do we measure inflation? Some measurement problems, 7th Meeting of the
International Working group on Price Statistics, Paris,, pp.28-29
[19] Lebow, D.E. and Rudd, J.B. (2003), Measurement error in the consumer price index: where do we
stand? Journal of Economic Literature, Vol. 41, pp.159-201
[20] Goldshmidt-Clermont, L. and Pagnossin-Aligisakis, E. (1999), Households’ non-SNA production:
labour time, value of labour time and of product, and contribution to extended private consumption, Review
of Income and Wealth, Series 45, No. 4
[21] Young, A.S. (2000), Income from households’ non-SNA production: A review, Bulletin of Labour
Statistics, Vol. 2, ILO, Geneva


  Treatment of Components of Consumption Expenditures Depending on Purpose

Expenditure Components           CPI               NA                 Welfare
All Major durable goods          Acquisition (P    Acquisition (P     Service flow,
(newly obtained)                 or E)             or E)              valuation?
All Major durable goods          Not Used          Not Used           Service flow,
(stock)                                                               valuation?
Other goods (second hand)        Acquisition (P    Acquisition (P     Acquisition (P)
                                 net of sales)     net of sales)
Other goods (Own-produced,       Acquisition (E)   Acquisition (E)    Use (E)
transfers or income-in-kind)
All Housing services             Service flow      Service flow       Service flow
                                 (E?)              (E?)               (E?)
Non-life insurance premiums      Acquisition (P    Acquisition (P     Acquisition (P)
and gambling expenditures        net of sales)     net of sales)
Unpaid household services        Not used          Not used, but in   Acquisition
                                                   satellite          (E?)
Social transfers in kind         Not Used          Available from     Acquisition
                                                   NA                 (E?)


NA       : National Accounts
P        : Purchase price
E        : Estimated value (Agreed Method)
E?       : Estimated value (No agreed method)