1 Somchai Jitsuchon Thailand Development Research Institute Sustainable Growth, Regional Balance, & Social Development for Poverty Reduction in Thailand.

1

Somchai Jitsuchon Thailand Development Research Institute

Sustainable Growth, Regional Balance, & Social Development for

Poverty Reduction in ThailandNESDB-WB

Bangkok, 26 October 2006

Poverty in Thailand & Improving Diagnostic

s

Outline

1. Overview of Thailand’s Poverty

2. Poverty Policy Formulation Fundamental Changes of Policy Architects National vs. Area-based Policy

3. Poverty Diagnostics Tools Poverty Data Small Area Estimation Poverty Map

3

Thailand’s Poverty Overview

Poverty TrendThailand’s Poverty Declined Rapidly over the Past 40-50 Years

If using old definition (before 2004), head-count ratio would be only around 5%

4542

34

28

1915

1820 21

1915

11

51

45

38

33

25

1719

21 2119

16

11

0

10

20

30

40

50

60

1986 1988 1990 1992 1994 1996 1998 1999 2000 2001 2002 2004

Consumption Poverty Income Poverty

0

0.1

0.2

0.3

0.4

0.5

0.6

Gini C

oeffic

ient

But Income Inequality Remains High..One of the World’ Highest

Consequences on Tar get Groups

Destitute poor (absolute poverty) has been dwindling in nu

mber, but some pockets of chronic poverty might exist.

Relative poverty increasingly important stubborn to economic growth, if inequality persists.

began to dominate public debates/policies. More ‘poverty measur

es’ are devised for the relative poor, not the poorest.

Problems of vulnerability also increasingly important, but

still largely neglected.

Rural and urban poverty more linked than in the past, due t

o convergence of economic activities.

7

Poverty Po licy Formulation

Changes in Poverty Policy Architects

In the past national poverty policy either did not ex

ist, or was an unsubstantial part of ‘Nati onal Plan’. Poverty declined mainly thr ough growth process.

Technocrats were thus key (and sole) ar chitects of poverty policy at national lev

el. Politicians mostly influenced sectoral po

- licies, or minor area specific policies.

Changes in Poverty Policy Architects

Present Poverty policy was nationalized by the TR

- T party around the year 2 0 0 0 1 , alo ngwi t h gl obal i nt er est i n pover t y r edu ction. Pol i t i cal success of TRT par t y was p

ar t l y due to thisshi f t . ‘ National Plan’ now plays very little role,

along with its technocrat architects. Pov erty policy was basically transferred to p

oliticians’ hands. Consequently, mostpovertypoliciesar e nowmor e t ar get i ng, mor e sect or

al. Oneexcept i on i s t he uni ver sal heal t h c arescheme.

-National vs. Area based

Mostofthetime(pastor pr esent ), al l maj or pover t y pol i c iesarecent r al l y concept ual i zed and i mpl em entedbycent r al gover nment ’s bur eaucr at i c arms.

However, therehasbeenat t empt t o decent r al i zed i mpl ement ations to ‘local governments’.

Forexample, provincesaregrantedmorepower (fi nanci al and bur eaucr at i c). More room for local initiatives. But most of l ocal eff or t s i s st i l l devot ed t o car rynat i onal pover t y pol i ci es desi gned b ynat i onal pol i t i ci ans.

Consequences The current policy quickly favors the r

elative poor, rather than the absolute poor (except the universal health care

).

There is urgent need for reliable pove rty data at disaggregated areas level (

at least at provincial level).

Also urgent need for high frequency p overty data (at least annually), to sup

port the ‘Poverty Eradication within 3 years’ agenda by TRT party leader.

12

Poverty Diagnostic Tools

Poverty Data Household Surveys on Consumption/Income

Census (pop census, agri cultural census, industrial

census) Administration Records Participatory Reports Hybrids

Thailand’s Pov erty Data

A. Use household surveys (SESs) alone. OK at national/regional level - but inadequate for true area based po licy implementations (e.g. SESs produc

e zero poverty in many provinces).

B. Rural Village Data: Nrd2C and BMN(basic minimum need)Ad hoc ‘poverty line’ composite index (monetary & non-monetary), with ad hoc formula

C. Poverty Registration (TRT part y initiative)completely self-report

- Mis targeting Proble m of

Poverty RegistrationNon-Poor Poor Total

Non-registered 82.0% 71.6% Registered 18.0% 28.4%Total 100.0% 100.0%Within Registered 89.9% 10.1% 100.0%

If not complimented by other datasources,

7 1 .6 % of poor people will be neeeeeeeee

2000Comparing SAE Ma 1 9 9 9 2 ‘

p’Nrd2C Classification

SAE Poverty Map Non-Target ornon-matched

Target Total

Non-Poor 39,781 9,511 49,292

Poor (30% up) 12,296 4,707 17,003

Total 52,077 14,218 66,295

The two maps are significantly differen t. Either (or both) may have the proble m of including the wrong villages as we

ll as excluding the right villages. Whic h one?.

SAE Poverty Map

Simple Idea: Get estimates of household income/consumption on large dataset (usually Census) based on models built on household surveys (SESs).

SESs have both (Y,X) but Census has only X.The models also allow for ‘location effects’

Advantages: •Combine Census’s Large Coverage with SESs’ Reliability.•Esitmated Y’s enable many applications (poverty, inequality, social security).

Limitations: •Only monetary definition of poverty.•Census is every 10 years (may use other dataset---BMN). •Huge data work, complicated econometric procedures.

chchch uy βxln

chcchu

First Map in 2000 (Joint projectNESDB/NSO/WB/TDRI) Use household survey 2000, Census 2000, and

village survey 1999 (provides location variables for rural map)

Second Map in 2002 (Join projectNSO/NESDB/WB/TDRI) household survey 2002, Census 2000, and village

survey 2002

- 2004Third Map in (on going effort)

SAE Poverty Maps in Thailand

Why Validation?•Survey Sampling Errors • Model Error• Omitted Variable problem• Inconsistency between SAE and Nrd2C

Three Field Validations•(1) Nakhon Sri Thammarat province (South) to verify SAE 2000 Map

•(2) Three provinces of Pitsanulok (nor th), Nonbualumpoo (northeast), Ratch aburi (central), to verify SAE 2002 Map.

•(3) Roi Et, Surin, Tak, Maehongson, to verify 2004 Map.

Field Validation

Poverty rates var ied considerably

-Some sub distric ts were clearly b

- etter off (Pai Kho don, Baan Grang)

.

-Central District Pitsanulok (head - count at sub district level)

Baan Grang: Group Inter view (farmers)

Baan Grang - was clearly a better off- sub district, supporting SAE results

Preliminary Evaluation of SAE

method SAE Poverty Map is fairly accurate in

predicting poverty ranking by area.

Poorer (Reality)

Better-Off (Reality)

Poorer (SAE) Predict

Predict few (X)

Better-Off

(SAE)Not

PredictPredict

SAE can benefit from improvement in the accuracy of surveyed income/consumption.

Need to simplify the method (underway), and overcome the theoretical and empirical issues of poverty map updating.

Applications so far•Limited applications at national level.

•More potential at provincial level, mostly through NSO’ provincial offices.

•Some CEO governors showed interest in using SAE maps.

Future Plan•More Promotion to Line Ministries.•Collaborate Ministry of Interior in linking to Nrd2C, BMN.

•Nrd2C/BMN can be used as data sources for ‘location variables’ in producing future SAE map.

SAE Applications