Statistical Issues in Measuring Poverty from
Non-Survey Sources
NATIONAL ACCOUNTS
UN STATISTICS DIVISION
Economic Statistics Branch
National Accounts Section
UNSD/NA/MR
1
Is poverty under or over
estimated?

Two sources of aggregate welfare data:

National accounts (NA)



Household surveys (HS)



Per capita Household final consumption expenditure
(HFCE)
Per capita Disposable income of households
Survey mean of consumption
Survey mean of income
Is survey mean typically lower than per capita
HFCE?
Income versus expenditure in the
poverty concept

Consumption surveys are far better than
income surveys for poverty
measurement



Households more willing to reveal their
consumption rather than their income
Conceptualization of non-wage incomes
Recall periods
Some conceptual and measurement
issues in National Accounts

Disposable income

Household final consumption

Two concepts


Household final consumption expenditure
Actual final consumption of households
Comparability between NA and
HS estimate of disposable income




Recording of income of self employed
persons
Imputed rent of owner occupied
dwellings
Exhaustiveness adjustments
Compilation of disposable income in NA
(Table 1.doc)
Components of
disposable
income
Contents covered in national accounts
Estimation method
+
Compensation of
employees
Includes wages and salaries, payments in kind and
employers’ social contributions to pension funds
and other insurance schemes
Should be surveyed
+
Operating surplus
Incomes received by households from using their
own dwellings
Should be surveyed in association with the
estimation of the imputed rental value of owner
occupied dwellings
+
Mixed income
Income left for own use to owners of household
enterprises without business accounts after
deducting from output intermediate cost of goods
and services as well as depreciation and taxes on
production
Should be surveyed in association with survey
on household production
+
Property income
receivable
Interest, land rent, dividends received and property
income attributable to insurance policy holders
imputed as received from pension funds
Should be surveyed. Interest received should be
adjusted to include financial service charges
paid that had been already deducted from
interest received. Property income attributable
to insurance holders must be imputed on the
basis of insurance held.
+
Withdrawals of
income from
quasicorporations
Withdrawals of income for own use by owners of
unincorporated enterprises but with full set of
business accounts such as partnership
Should be surveyed
+
Social benefits
other than social
transfer in kind
Social security benefits in cash, private funded
social benefits (pension benefits), unfunded social
benefits by employers and social assistance benefits
in cash
Should be surveyed and or use of pension funds
data
+
Other current
transfers in cash
receivable
Net non-life insurance claims, current transfers
from government, current transfers from relatives
and others
Should be surveyed and adjusted
to exclude insurance service
charges
+
Social transfers
in kind
Individual final consumption of government and
NPISHs[1]
Imputed by analyzing government
and NPISHs expenditure with
regards to types of households that
benefit
-
Social
contributions
Contributions to social security fund, pension funds
and other insurance schemes
Should be surveyed and compared
with sources of data like
government, pension funds and
insurance companies
-
Property income
payable
Interest and land rent paid
Should be surveyed and compared
with data from financial
corporations and adjusted to
exclude financial service charges,
which are treated in the NA as
final consumption.
-
Taxes on income
Regular income, property and wealth taxes
Should be surveyed and compared
with data from government
-
Current transfers
payable
Net non-life insurance premiums paid and current
transfers from relatives and others
Should be surveyed and compared
with data from insurance
companies and adjusted to exclude
insurance service charges
=
Disposable
income
(adjusted)
Comparability between NA and HS
estimate of household consumption


HS provides information on household
consumption at the lowest detailed level
but
HS can not be directly used for NA
estimates


Conceptual adjustments
Empirical adjustments
Conceptual adjustments

Differences in definitions and concepts








Households production for own final consumption
Services of owner occupied dwellings
Income in kind
Financial intermediation services indirectly measured
Insurance and pension fund services
Direct sales and purchases for business purposes
National concept
Diagram 1 (Diagram 1.doc)
Households expenditures
based on
Household budget survey
Adjustments for purchases of
residents abroad and nonresidents on the domestic territory
- national concept
Household production for own final
consumption
Adjustments for direct sales
and
purchases for business purposes
Services of owner occupied dwellings
– imputed rentals
Goods and services provided by
employers
- income in kind
Other services
-FISIM
- Insurance and pension fund services
Household final consumption expenditure - national National accounts
Empirical adjustments

Differential non-response



Data confrontation




High income households
Exhaustiveness
Retail trade
Enterprise surveys
Others
Diagram 2 (Diagram 2.doc)
Households final consumption expenditure
based on
National accounts
Adjustments for differences in
population
- institutional households
Retail trade data
Adjustments for data
confrontation
Surveys of enterprises
Adjustments for purchases
for business purposes
Household final consumption expenditure
National accounts estimate
Exhaustiveness adjustments for
differential non-response
Others
- Foreign trade data
- Administrative data
Actual final consumption of
households



Shows who benefits from the consumption
Main objective is to enhance comparability of
households final consumption across
countries and over different time periods
Consists of :



Final consumption expenditure of households
Social transfers in kind from the Government
Social transfers in kind from NPISHs
Relationships between the two
concepts
Sector making expenditure
Individual
consumption
Collective
consumption
Total final
consumption
General
Government
NPISHs
Households
X
X
X
Actual final consumption
Households actual individual
final consumption
(= Social
transfers in kind)
(= Social
transfers in
kind)
X
0
0
Government actual
collective final consumption
Government final
consumption
expenditure
NPISHs final
consumption
expenditure
Households final
consumption
expenditure
Actual final consumption =
Total final consumption
expenditure
International debate on non-compliance
between NA and HS survey means

Why use national accounts data
(Karshenas)?



There is significant relationship between the two
measures but it is not always the case that
national accounts consumption would be lower
than survey-based mean consumption
National accounts-based estimates appear to be
more plausible in relation to other non-monetary
indicators of poverty
Calibration of survey means by using external NA
based information as a scale factor for a removal
of discrepancy
International debate on non-compliance
between NA and HS survey means (cont.)

Why prefer household survey data (Deaton)?



Designed to directly measure individual welfare
Yes, under-reporting/non-compliance problems,
but probably produce more accurate measure of
living standard of the poor
No sound basis for using mean from NA and
distribution from household survey for calibration
to remove differences.
Conclusions


Discrepancy between NA and HS means has
important implications for the analysis of
poverty
NA data even not used for direct estimation of
poverty might be of particular help in case of:



Outdated HS data
Missing HS
Practice of HS needs to be improved and
harmonized
Thank You