Strictly Confidential © 2014

Poverty measurement:Housing

Why is housing relevant?

Gaerner and Short (2009): “…[including housing in the welfare aggregate]…allows for more reasonable inter-household comparisons, as well as international comparisons, of economic well-being.”

Consumption Aggregate ≈ Indirect Utility Function

It must be as comprehensive as possible(we must consider all goods and services)

Why is housing relevant?

Dominant component of total income(Norris and Pendakur, 2013)

Share of housing consumption increases with welfare levels (Blades, 2009)

13.413.8

14.1

15.916.4

16.817.117.217.2

17.517.617.717.717.918.118.1

18.2

19.119.5

19.819.820.120.2

20.5

21.421.521.7

24.9

10

12

14

16

18

20

22

24

Korea

Norway

Portugal

United States

Sloven

iaBe

lgium

Austria

Estonia

Ireland

Germany

Spain

France

Nethe

rland

sMexico

Australia

Swed

enHu

ngary

Italy

Luxembo

urg

Finland

Poland

Japan

United Kingdo

mSw

itzerland

Czech Re

public

Denm

ark

Slovak Rep

ublic

Greece

Housing as a share of total households’ net adjusted disposable income

Malawi

Burkina FasoKenya

Mauritania

Lao People's Democratic RepublicIndiaHondurasGuatemala

Belize

Azerbaijan

Colombia

Belarus

BulgariaMalaysiaMexico

ChileLithuania

Latvia

HungaryEstonia

Czech Republic

Malta

Israel

Greece

Cyprus

ItalyIceland

JapanGermanyGreenlandFranceCanadaBelgium

Finland

AustraliaAustriaIreland

Denmark

010

2030

Hou

sing

as

shar

e of

hou

seho

ld to

tal e

xpen

ditu

re

0 20000 40000 60000GDP per capita, US$

Source: OECD National Accounts, 2012 Source: authors elaboration on UNStat System of National Accounts

What are the problems?The basics…

The utility is the value of the flow of services from occupying the dwelling rather than the expenditure for purchase it over the period of analysis

Rental markets they work perfectly and all households rent their dwellings. Hence, market rents are a good approximation of dwelling services

However… Many households own their dwellings In other cases, households receive housing free of

charge or at subsidized rates by their employer, friends, relatives, government or other entities

Few households rent their dwelling

Source: authors elaboration on EU-SILC; SEDLAC; http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?fpt=table (USA); http://kostat.go.kr/portal/english/news/1/17/1/index.board?bmode=read&bSeq=&aSeq=273080&pageNo=2&rowNum=10&navCount=10&currPg=&sTarget=title&sTxt= (Korea); authors elaboration on HBS (Albania, Bangladesh, Iraq, Peru)

97.3

94.1

93.9

92.7

92.0

91.0

90.2

88.9

88.8

87.3

86.0

84.9

84.2

83.0

82.3

82.3

81.3

80.8

80.3

80.3

79.8

79.5

79.3

76.5

76.4

75.5

74.0

73.9

72.7

72.7

71.2

70.9

70.8

70.0

69.7

69.4

69.2

69.2

69.0

68.2

67.9

66.8

64.0

63.5

63.1

62.9

62.6

61.8

54.2

52.6

52.5

48.9

46.2

44.0

11 2.67.271

9.0 811.13 2 5.010

2 9 83

136 9 13

1021312

2227 29

11 921

33

20

264449

2 5 3.6 72

8 119.0513

8 1015

71411

814

211214

50 1

192210

0

18 22

7 5

17.7

19.2

20.5

26.1

28.8

29.2

30.0

30.8

31.0

31.8

32.1

36.0

36.5

36.9

37.4

38.2

45.8

47.4

56.0

0

10

20

30

40

50

60

70

80

90

100

RO LTAlbania, 2012

Iraq, 2012

HRPe

ru, 2013 SK

Banglade

sh, 2010

HU BGTh

ailand

, 2009

NO EE ES LV

Nicaragua, 2009 PL

Paraguay, 2011 EL SI IS

Vene

zuela, 2006 CZ FI

MT PT IT

Guatemala, 2011 LU DK

Brazil, 2012 NL

Costa Rica, 2011

Hond

uras, 2011

UK CY

Peru, 2012 BE

Ecuado

r, 2012

Mexico, 2012

El Salvado

r, 2012 SE

USA

Argentina, 2013

Chile, 2011 FR

Dominican

 Rep

., 2011

Bolivia, 2012

Korea, 2010

Uruguay, 2012 AT DE CH

Colombia, 2012

Owner Tenant Subsidized Tenant Non Owners

Theoretical viewpoint When expenditure is used as a yardstick of welfare, it

is important to achieve comparability across household

Including the rent expenditure for renters and not the flow of the service of the dwelling for non-renters would lead to wrong conclusions: Tenants would be better off than Non-Tenants!

Empirical Application

Market Tenants Owners and Non-market Tenants

We assume that tenants report the actual market rent

However, it is important to check if this is an accurate value

Impute what they would be paying if they were renting instead of owning

TEN

AN

TSO

WN

ERS

AN

D N

ON

MA

RK

ET T

ENA

NTS

Welfare value of the flow of services from dwelling

The rent paid by tenants as reported in the survey is supposed to be the true value of the flow of services for tenant, and included as such in the welfare aggregate

Information from external sources may be used

Econometric and statistical methods

Find the best fitting model for rents paid by tenants on the basis of observable dwelling characteristics

Apply the same model on dwellings’ characteristics of home owners and non-market tenants

Infer imputed rent for home owners and non-market tenants

Self-Assessment

Use the value reported by owners (and non-market tenants) as a reply to the question “If you would rent this dwelling, how much would you paid per month?”

Capital Market Approach

Find the capitaliz

ation rate from rents and

market value of

the dwelling

Apply the capitalization rate to the value of the owners occupied dwellings

Collect information on the return rate for housing and operating costs

Rent-to-

Value

User Cost

Payment Approach

Use information about the out-of-pocket expenditures faced by households. For owners, this is the sum of mortgage interest payments, taxes, insurance, maintenance and repairs

External Sources

(i) other households surveys

(ii) administrative register

(iii) newspaper or internet offer notices

(iv) multiple listing datasets

(v) mortgage transactions

(vi) key informant evaluations

Econometric and statistical methods

Find the best fitting model for rents

paid by tenants on the basis of

observable dwelling characteristics

Apply the same model on dwellings’ characteristics of home owners and

non-market tenants

Infer imputed rent for home owners and non-market

tenants

Econometric and statistical methods1. Hedonic models Utility derives from attributes or characteristics of goods

and not from goods per se (Lancaster, 1966) In equilibrium, economic agents observe the prices of

differentiated products and specific amounts of characteristics associated with them. This reveals the implicit prices for the different characteristics (Rosen, 1974)

Housing can be considered as a composite commodity:, ,

No consensus about the specific explicit form the hedonic price function might take

BUT: the relation between rent and characteristics should be non-linear

Location Structural Characteristics

Neighborhood

Econometric and statistical methods1. Hedonic models

The most generic functional specification is:R

In the literature, different specifications and estimation methods are found: Linear, semi-log and higher order Quantile regression, semi-parametric and non-parametric Heckman selection model Spatial Dependency

Econometric and statistical methods1. Hedonic models: Semi-Log

lnR

PROs: The coefficients show approximately the percentage change in the

imputed rent for a given unit-change in the covariate It mitigates the heteroskedasticity problem It is computationally simple It allows the marginal rent-value to be a non-linear function of the

size and quality of the dwellingBUT: The unobserved quality of the dwelling chosen by renters can be

different from that chosen by owners

Econometric and statistical methods1. Hedonic models: Heckman Selection If the choices of tenure type and characteristics of the dwelling are

not independent, the OLS estimation in the market rental sector might be inconsistent (Arevalo and Ruiz-Castillo, 2004)

SOLUTION: Heckman two-stages selection model

R ∑ 1 Where:

1 0

0

Econometric and statistical methods1. Hedonic models - COMMENT

PRO: Simple and fast to implement We can apply the model fitted on tenants on the entire

population and get an estimate of imputed rent for everybody

CONS: We risk to underestimate the prediction for those with

higher flow of services from housing and to overestimate the prediction for those with lower

There is no way to test whether the prediction for owners and non market tenants is accurate

Econometric and statistical methods1. Hedonic models – COMMENT (cont.)

Alert: In some countries, housing and rental markets are not

well enough developed to permit any serious estimate of rental value, and attempts to repair the deficiency using data from a small number of households are unlikely to be effective, however sophisticated the econometric technique (Deaton and Zaidi, 2002, p.38-39)

Econometric and statistical methods2. Stratification

Define a set of relevant characteristics Each characteristic has a set of possible realizations:

(e.g. Region1, Region2, Region3)(e.g. Detached House, Flat)(e.g. 1room, 2rooms, 3rooms, >3rooms)

From which we can define strata of dwellings with homogeneous characteristics:

Take the average rent for each stratum and assign it to each owner-occupied or non-market tenant dwelling in the same stratum

Econometric and statistical methods2. Stratification

PRO: Dwellings in the same strata will be of a more

homogeneous quality, leading to more precise estimates for owners and non-market tenants, since the model is defined on dwellings with similar characteristics

BUT: Increasing the number of strata reduces the average

number of observations per stratum What we learn: We may obtain better predictions for owners if we infer

their rents using information from tenants with dwellings having similar characteristics, possible there is no overlap

Self-assessment This approach is based on data collected about

owners’ estimates of a fictitious market rent homeowners are asked to estimate how much they would pay if they were renting their home

Assumption: owners can estimate rental equivalences This should be less problematic in regions where rental

market is active and well developed (Lanjouw, 2009) BUT…owner-occupiers may over-estimate the true

rental value of their dwelling given the affinity to their property or neighborhood (owner pride factor)

Should be treated with caution and should be tested Sergio’s presentation next Thursday!

III. Distributional Impacts

A theoretical framework Population , made of 1,2,… , individuals, ∈ , , … , distribution of welfare aggregate (w/o

rent) level of welfare aggregate for individual Assume ⋯ , , … , distribution of rents Consider three possible scenarios:

. ∀ ∈ , 0

. ∀ ∈ , ∈ 0,1

. ∀ ∈ , ∈ 0,1

A theoretical framework

wel

fare

agg

rega

te

Individuals

Baseline Scenario A Scenario B Scenario C

Distributional Impact of Rent Imputation

We are interested in:

Distribution Poverty

InequalityLevels

Profiles

Fixed poverty

line

New poverty

line

Ranking

Shape

Shared Prosperity

010

3040

20

0 20 40 60 80 100Percentiles of Population

010

2030

4050

0 20 40 60 80 100Percentiles of Population

DistributionTheoretical Framework

Define a Rent-Incidence Curve:

010

2030

4050

0 20 40 60 80 100Percentiles of population

Scenario A

Scenario B Scenario C

Inequality Decreases

Inequality is constant

Unpredictable Effect

DistributionEvidence from the Literature In general, the literature finds that including rents

reduces inequality Mainly developed countries Mostly income as welfare aggregate Some examples: Guenardand S. Mesple-Somps (2010) for Madagascar and

Cote D’Ivoire: the poor are more likely to own their homes Gasparini and Escudero (2004) for the Greater Buenos

Aires area: large proportion of house-owners at the bottom of income distribution, and income elasticity in housing expenditure <1

Törmälehto and Sauli (2010, 2013) for 29 EU-SILC countries: combined effect of the equalizing gap-effect and dis-equalizing re-ranking effect

DistributionOur Findings – Rent Incidence Curve

10

20

30

40

50

Ren

t inc

iden

ce c

urve

0 20 40 60 80 100Percentiles

OLS Heckman MatchingSelf Assessed Original

IRQ, 2007

10

20

30

40

50

Ren

t inc

iden

ce c

urve

0 20 40 60 80 100Percentiles

OLS Heckman MatchingSelf Assessed Original

IRQ, 2012

DistributionOur Findings - Inequality

Peru, 2013Iraq, 2012

0.2

0.25

0.3

0.35

0.4

Gini Gini Gini

Overall Owners Tenants

Norent OLS Heckman Matching Self_Assessed Original

Inequality increases when using OLS as the imputation method

22.53

3.54

4.55

5.56

p90p10 p90p10 p90p10

Overall Owners Tenants

Norent OLS Heckman Matching Self_Assessed Original

0.2

0.25

0.3

0.35

0.4

Gini Gini Gini

Overall Owners Tenants

Norent OLS Heckman Matching Self_Assessed Original

22.53

3.54

4.55

5.56

p90p10 p90p10 p90p10

Overall Owners Tenants

Norent OLS Heckman Matching Self_Assessed Original

DistributionOur Findings - Re-ranking

PE

RU

, 20

13

Even if inequality does

not change much, there is still re-ranking

OLS1 2 3 4 5 6 7 8 9 10

1 92 7 0 0 0 0 02 8 79 12 1 0 0 0 0 03 14 69 14 3 0 0 0 0 04 19 62 16 2 1 0 0 05 23 59 15 3 0 06 23 56 18 2 0 07 26 56 16 2 08 22 65 13 19 17 73 1010 11 89

Heckman1 2 3 4 5 6 7 8 9 10

1 98 1 0 02 2 96 2 0 0 03 3 92 5 0 0 04 5 89 5 1 0 0 05 6 87 5 1 0 06 7 84 7 1 07 10 82 7 18 9 83 7 09 8 87 510 5 95

Matching1 2 3 4 5 6 7 8 9 10

1 93 6 1 0 0 0 0 0 02 7 81 10 1 0 0 0 0 03 13 71 12 3 1 0 0 04 19 64 14 2 1 0 0 05 22 58 15 3 1 0 06 25 55 16 3 1 07 28 55 13 4 08 24 62 13 19 20 69 1110 13 87

Self_Assessed1 2 3 4 5 6 7 8 9 10

1 93 6 0 0 02 7 82 10 1 0 0 0 03 12 75 11 2 1 0 0 04 15 69 13 2 0 0 0 05 19 63 14 3 1 0 06 22 59 15 3 1 07 24 60 13 2 08 21 65 13 19 18 72 1010 12 88

Original1 2 3 4 5 6 7 8 9 10

1 93 6 0 0 02 7 82 10 1 0 0 0 03 12 74 11 2 0 0 0 04 15 69 13 2 0 0 0 05 19 62 14 3 1 0 06 23 59 15 3 1 07 24 60 13 3 08 21 65 13 19 18 72 1010 12 88

IRA

Q,

2012

DistributionShared prosperity

6

8

10

12

0 20 40 60 80 100pctl

No rent OLS HeckmanMatching Self Original

PER, 2010 - 2013

5

6

7

8

9

0 20 40 60 80 100pctl

No rent OLS HeckmanMatching Self Original

IRQ, 2007 - 2012

Poverty IndicesTheoretical Framework – Fixed Poverty Line

Poverty does not decrease and

rank among the poor is preserved under scenarios

A and B

Poverty does not decrease BUT

rank among the poor might NOT

BE preserved under scenario C

Poverty IndicesTheoretical Framework – Adjusted Poverty Line

Poverty might increase…

…Or remain the same …

…Or decrease

Even under the simplest scenario A …

Poverty Indices – FGT0Our Findings – adjusting poverty line

Iraq, 2012Absolute, Fixed Poverty Line

Iraq, 2012Absolute, Adjusted Poverty Line

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

All Owners Tenants Rural Urban

Tenure Status Type of settlement

Norent OLS Heckman Matching Self_Assessed Original

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

All Owners Tenants Rural Urban

Tenure Status Type of settlement

Norent OLS Heckman Matching Self_Assessed Original

Poverty ProfilesTheoretical Framework – Fixed Poverty Line

R1

R2

R3

, ,

R1

R2

R3 z

1/3 of poor individuals

live in region 1, 1/3 in

region 2 and 1/3 in region

3

2/3 of poor individuals

live in region 1 and 1/3 in

region 2

Profile is the same as the

original distribution. Notice re-ranking!

R1

R2

R3

,

R3

R2

R1

Other examples: urban/rural, minorities, education…

Poverty Profiles – FGT0Our Findings

Iraq, 2013Absolute, Fixed Poverty Line

Iraq, 2013Absolute, Adjusted Poverty Line

00.10.20.30.40.50.60.70.80.9

1

Mal

e

Mar

ried

Empl

oyed

Tena

nt

edu_

1

edu_

2

edu_

3

edu_

4

Coa

stal

Cen

tral

Mou

ntai

n

Demographic Education GeographicalRegion

Share of poor individuals

No Rent OLS Heckman

Matching Self Assessed Original

00.10.20.30.40.50.60.70.80.9

1

Mal

e

Mar

ried

Empl

oyed

Tena

nt

edu_

1

edu_

2

edu_

3

edu_

4

Coa

stal

Cen

tral

Mou

ntai

n

Demographic Education GeographicalRegion

Share of poor individuals

No Rent OLS Heckman

Matching Self Assessed Original

PovertySummary of Results

As expected, poverty decreases by keeping fixed the poverty line and adding imputed rents to the consumption aggregate

Poverty changes very little when the poverty line is adjusted by adding imputed rent in the non-food component

With an adjusted poverty line poverty tends to slightly decrease for tenants and to increase in rural areas (particularly in Iraq and Peru)

The profile of the poor does not seem to change, both keeping fixed and varying the poverty line, according to the imputation method used

Further Readings

Strictly Confidential © 2014