Near-Poor Vicente B. Paqueo - PowerPoint Presentation

Slide1

Near-Poor

Vicente B. Paqueo, Elvira M. Orbeta, Sol Francesca S. Cortes, and Ana Christina V. Cruz

C H a l l e n g e

And Strategy Development Ideas

Analysis of the

Slide2

 

Persistent poor

TransientsPersistentNon-Poor

% of households8.9

38.6

52.45APIS Panel Data 2004-2010

Persistent poor (non-poor) are households with income less (greater) than official total poverty threshold (TPT) in all APIS survey years during 2004-2010. Transients are households whose income is less than TPT at least once during the same period.

Context 1:

CONCERN ABOUT CYCLICAL POVERTY

households (HH) moving in and out of poverty is quite high

Slide3

Contextt 2:

CONCERN ABOUT “STINGINESS”of the total poverty threshold (TPT)Households above the official TPT at a given year are classified as non-poor due to the “stinginess” of the TPT

They are, therefore, officially excluded from social assistance like the Pantawid PamilyaIn reality, many of them live difficult and precarious lives with high risk of becoming poor any time soonThese non-poor HHs called the “near-poor” in the literature is increasingly becoming a concern in the light of the inclusiveness goal

Slide4

QUESTIONS:

Operationally, who are the Near-Poor? What are the socioeconomic challenges they are facing?

What should be the stance of the Government, given its budget constraints and the more urgent needs of the poor?What (if any) can the Government do to meet the challenges of the near-poor without undermining its ability to help get them out of poverty? What can the near-poor do on their part?

Slide5

Structure of presentation

Part 1: The near-poor: concept, practical estimation methods and profile Part 2: Rationing of Social Assistance When Budget Is Insufficient to Cover All the Near-Poor Households Part 3: Near-poor strategy development

Slide6

 

Definitions of Near-Poor in the Literature

Definition of Near-poor

Poverty Threshold Measure UsedReference

Individuals with family income between 100 and 133 percent of the official poverty threshold

three times the cost of minimum food diet in 1963Orshansky (1966); Heggeness and Hokayem (2013)Individuals with family income between 100 and 125 percent of the poverty thresholds

three times the cost of minimum food diet in 1963Hokayem and

Heggeness

(2014)

Those living between 100 and 150 percent of poverty

National Academy of Sciences (NAS) measure

Ben-Shalom, Moffitt and

Scholz

(2011)

Those with modest incomes - living between 100 to 200 percent of povertySupplemental Poverty Measure (SPM) Short and Smeeding (2012) Workers with per-capita household consumption of between US$2.0 to US$4.0 a day ( i.e., between 160 to 220 percent of poverty) US$1.25 a day (@2005 PPP) absolute poverty lineILO (2013) Vulnerable non-poor – those whose income is within 20% above the poverty line US$2.0 a day – 2009 official poverty line Chua (2013)Those with consumption of approximately 1.2x the poverty line US$1.20 a day 2011 national poverty line in IndonesiaJellema and Noura (2012) Lowest three deciles based on household consumption as covered by the Indonesian health insurance program intended for the poor and near-poorlowest three deciles Harimurti et al. (2013)

Table

6. on study

Slide7

Part 1

The near-poor:

Concept,

Practical Estimation Methods And Profile

Slide8

Finding: Ratio (RAT) of Near-Poverty Threshold (NPT) to Total Poverty Threshold (TPT) Clusters Around 1.1-1.37

 

NPT-TPT ratio (RAT)% above TPT

PhilHealth1.1

10

World Bank1.220

Vietnam1.25

25

US Census Bureau

(

Hokayem

and

Heggeness

, 2014)

1.25

25Orshansky (pre-2014)1.3333Authors’ Estimate1.28 (knife-edge)281.37 (balik-balik)37

Slide9

NPT Amount (in 2014 Pesos) by Different RAT Levels: % over TPT

Annual Per Capita Poverty Threshold

20% = World Bank

25% = Vietnam; US Census

33% =

Orshansky28% = Authors(Knife-Edge)37% = Authors (Balik-Balik)

20,14322,157

10% =

PhilHealth

24, 172

24, 179

25, 783

26, 790

27, 596

Amount equivalent in

Php

Slide10

Comments on Previous RATsBased on income distribution of households

NOT considering the risk of subsequent povertyHow did we estimate our RATs?

Slide11

Methodology for the Balik-Balik RAT estimate of 1.37

We define the Balik-Balik (cyclical) poor as households who move in and out of poverty during a given time period between 2004-2010 (also called transients).We consider the mean household income of the transients (MINC) as the NPT for the Balik-Balik households.The Balik-Balik RAT = MINC/TPT = 1.37

Slide12

Methodology for Knife-Edge RAT Estimate of 1.28

We use the following working definitionPer capita income above the official total poverty threshold (TPT) at a given year, but at high risk of subsequently falling into

poverty soonMetaphorically, non-poor households that precariously live at a knife-edge with little or no buffer against the economic effects of idiosyncratic and covariate shocks

The NPT is chosen at the level of household income that implies a risk of subsequent poverty of at least 50%

Slide13

Methodology for Knife-Edge RAT Estimate of 1.28

RSKj = a + b (1/PCY) for j = e, d whereRSKe

= mean number of poverty episodes in 2007-2010 experienced by the non-poor households of 2004, as a percent of number of survey yearsRSKd = percent of non-poor households in 2004 becoming poor at least once in 2007-2010PCY

= per capita household income

Slide14

Regression

equation estimates relating poverty risk of non-poor households to income

 

RSKdRSKe

Constant

-.2326382 -.189562(-4.43)

(-5.27)1/Per Capita income

9083.147

5663.6

(12.69)

(11.56)

Adjusted R-squared (goodness of it measure)

0.9468

0.9365

Number

of Observations = 10 (mean values of 10 inome classes)

Slide15

Predicted Poverty Risk (RSKj) of Non-Poor Households, Using

NHTS-PR Proxy Means Test Income Estimates

NPT =Mean income per capitaaRAT = NPT/TPTb

Frequency distribution of non-poorcPredictedd

RSKdPredictedd RSKe10,1631.0514%

66%37%

11,129

1.15

12%

58%

32%

12,095

1.25

11%

52%28%13,0721.359%46%24%14,0391.458%41%21%15,0041.556%37%

19%15,9681.655%34%

17%

16,921

1.75

5%

30%

15%

17,905

1.85

4%

27%

13%

18,872

1.95

3%

25%

11%

27,310

2.82

23%

10%

2%

Slide16

Checking out the exclusion and inclusion error rates

RATExclusion Error RateInclusion Error

Rate1.1 (PhilHealth)36%2%1.2 (World Bank)20%

4%1.28 (Knife-edge)12%7%1.37 (

Balik-balik

)8%11%Exclusion Error – those with HIGH risk of subsequently falling into poverty but excluded because of income greater than NPT.Inclusion Error – those with LOW risk of subsequently falling into poverty but included because income less than NPT .

Slide17

Use of RAT is a good practical approach, particularly when panel data are not available to link NPT to RSK

Inclusion error rate is minimal (only 2 percent) at RAT 1.10 and rises moderately to just 7 percent at RAT 1.28 or to 11 percent at RAT 1.37Exclusion error rate is a high of 36 percent at RAT 1.10 but falls to a moderate rate of 12 percent at RAT 1.28 or to a lower rate of 8 percent at 1.37

Memo: The choice of low RAT due for example to tight budget would probably lead to a great number of high risk, near-poor households

Slide18

Exclusion/ inclusion errors are MANAGABLE.

Two ways of managing the errorsIncrease RAT, as additional budget becomes available to reduce the exclusion error rates without unduly raising inclusion

errorsPrioritize high risk near-poor households (methodology in Part 2)

Slide19

Mean Characteristics of Near-Poor (RAT= 1.28)(a) Income and Expenditures Closer to Poor(b) Savings are positive for the near-poor but negative for the poor

 

2010Profiles

All HHsPoor

Near-Poor

Non PoorDistribution (%)10022.9312.15

64.93Total Income1106,385

41,090

57,593

138,567

Wages and Salaries

1

42,900

14,435

22,842

56,703Entrepreneurial Income129,28215,58320,56635,749Total Expenditures194,25043,75156,352119,170Education Expenditures14,7159711,6396,613Medical Expenditures13,060462

9484,373Savings212,135-2,661

1,240

19,397

Notes

1/ The household mean income and expenditure values are in pesos expressed in 2000 prices.

2/ Savings = Total Income – Total Expenditure

Slide20

Mean characteristics of the near-poor: more human and physical assets

 

2010Profiles

All HHsPoorNear-Poor

Non Poor

Education, assets and location

% Did not graduate high school59.483.28

75.5

47.95

% Own Lot

74.4

66.16

70.05

78.12

% with Own Electricity

86.1265.3178.8494.84% with Own Water38.7614.120.4250.89% Houses made of Strong/Predominantly Strong Materials77.0354.2365.4787.24% Urban Household38.8217.7724.548.92

Notes1/ The household mean income and expenditure values are in pesos expressed in 2000 prices.2/ Savings = Total Income – Total Expenditure

Slide21

General implication of near-poor profile

Compared to poor households, the average near-poor households might have some savings and assets that can be leveraged to reduce their vulnerability to shocks and poverty with a little help from the Government

Slide22

Trends in poverty and near-poverty:

year total number of households

poornear poor with RAT = 1.28

near poor with RAT = 1.37number of households

%

number of households%number of households%

2000 15,071,941

3,111,800

20.65

1,539,876

10.22

1,968,982

13.06

2006

17,403,482

3,685,13521.17 1,899,010 10.91 2,429,240 13.962009 18,451,542 3,850,10720.87 2,052,554 11.12 2,656,210 14.42012 21,425,737

4,264,58419.9 2,328,293 10.87

2,992,348

13.97

Source: FIES 2000, 2006, 2009, 2012

Slide23

Trends in poverty and near-poverty:

Source: FIES 2000, 2006, 2009, 2012

Slide24

Part 2

Rationing Of Social Assistance

When Budget Is Insufficient To Cover All The Near Poor Households

Slide25

Defining the challengeHow to ensure that the near-poor HHs with higher RSK are first in line in the selection of near-poor program beneficiaries?

CONTEXT OF THE CHALLENGEAbsence of RSK information in the Listahan data baseDesirability of reducing exclusion error rates, while avoiding inclusion error rates from ballooning Heterogeneity: actual RSK could substantially differ for same income households due to other factors

Slide26

Estimating a multivariate regression equation to generate “predicted RSKe” of households that were not poor in 2004

Relating RSKe (number of poverty episodes per year) to selected risk factors (next slides), using Ordinary Least Squares on APIS Panel Data 2004-2010Selected income and proxy variables in the Listahanan database

With the estimated RSKe equation, the variables can be used to generate “predicted” RSK for each of the households in the ListahanThe % change of RSKe per unit % change in the “predictors” (referred to as elasticity) can be estimated Predicted

RSKe can be arranged in descending order

Slide27

OLS

Regression Equation for Generating “Predicted RSKe” of the

Non-Poor Households of 2004

2004 values of

“predictors”

DefinitionCoefficientsElasticity

Constant

0.4765859***

 

(8.82)

 

Income per capita

Per capita income, average of 2004-2010 in constant 2000 prices

-0.0000006***

-0.30621

(-4.27) Household sizeNumber of member of HH0.0169491***1.16699

(98.35) Home built from Strong MaterialsHH with outer wall and roof made of strong or predominantly strong materials

-0.0151651

 

(-1.51)

 

Own Water

Number of HH with own water

-0.0271347***

-0.22265

(-3.13)

 

Consumer

durables

Number of the following consumer durables a HH has:

TV,Video,Stereo,Refrigerator,Washing

machine,Aircon,Phone

,

Cell

phone,Computer

, Cars, Motorcycles

-0.0193140***

-

1.65286

(-10.5)

Table 16 on study

1

/ * significant at the 10 percent level, ** at the 5 percent level, and *** at the 1 percent level. Dependent Variable is

RSK

e

.

2/ Figures in parentheses are t-values.

3/ See below for definition of

variables

Slide28

Continue: OLS

Regression on the Non-Poor of 2004

Variables

Definition

Coefficients

ElasticityAverage WageAverage Wage per region in 2004

-0.0000010*

-0.06054

(-1.67)

 

Regional Unemployment Rate

Regional Unemployment rate in 2004

0.0018366

 

(1.3)

 Regional Underemployment RateRegional Underemployment rate in 20040.0041283***1.182034(6.44)High School Graduate

Education level of HH head is at least high school level -0.0698544***

-0.8185

(-8.01)

 

Sex (Male)

 

Gender of HH head

-0.0096480

 

(-0.58)

 

Age

 

Age of HH head

-0.0084159***

-5.90034

(-4.25)

 

Slide29

Continue: OLS Regression on the Non-Poor of 2004

Variables

Definition

CoefficientsElasticity

Age-squared

 Square of Age of HH head 0.0000618***

2.043439

(3.09)

 

(-4.15)

 

Self-Employed

Self-Employed HH head

0.0173827**

0.117202

(2.22) Single-0.0183866

 (-0.78)

 

Divorced/Separated

HH Head

Divorced/Separated

-0.0567049*

0.01778

(-1.8)

 

Widowed

HH Head Widowed

-0.0346207*

-0.01659

(-1.91)

R-squared

0.2395

Adjusted R-Squared

0.236

Sample

Size

3904

Continuation of Table 16. on study

Slide30

Arrange households from high to low RSK

Predicted RSK arranged in descending order

Slide31

 

Socioeconomic Conditions

Best(low risk)Good(moderate risk)

Worst

(high risk)Predicted RSKe-0.030.35

0.63Income per Capita

52596.66

13149.96

2123.689

Family

Size

3

7

18

Own Electricity100Own Water100Durables60.060Average Wage14636.342042.918546.9167Regional Underemployment

12.0732124.2323532.3At least High School Graduate10

0

Age

37.36872

59.74662

98

Age Squared

1515.391

3812.529

9604

Urban

1

0

0

Divorced

0

1

1

Illustrative combinations of socioeconomic conditions indicating high-low “predicted

RSKe

”, using 30 randomly chosen HHs

Slide32

Steps in targeting and prioritizing near- poor households for social assistance HOW?

Use equation to calculate “predicted” RSK for each household, using Listahanan data and integrating the new information in the Listahanan databaseRank households based on predicted RSKClassify high risk

near-poor households as those HHs above a specified RSK levelGive priority to the above householdsPlanned application of RSK equation by NHTO

Slide33

Part 3

Near -Poor

STRATEGY DEVELOPMENT

Slide34

Organizing a Near-Poor Strategy

Rationale: why not just focus on the persistently poor?Cyclical and near-poverty is prevalent Dealing with near-poverty is a way of correcting for the “stinginess” of PH poverty thresholdNear-Poor are also victims of policy and program failuresDesirability of preventing

near-poor moving back to poverty In sum, dealing with the near-poor issue would strengthen the development of a sustainable, compassionate and efficiently convergent anti-poverty program

Slide35

ORGANIZING A NEAR-POOR STRATEGY

PrinciplesFocus on win-win interventionsDevelop multi-level multi-dimensional convergent strategyPromote self-reliance – leveraging own resources

Three PillarsReform of failing policies detrimental to poor and near-poor.Addressing pervasive market failuresTime-limited (pantawid) social assistance

Slide36

Pillar I

Reforming current government policies that have been shown to have counter-productive impacts on low income households

Examples Minimum wage and jobs expansion program (JEDI)Rice import liberalization plus a targeted “pantawid” style income support program to help deserving rice farmers adapt to a new policy environment

Slide37

Fixing market failures arising from pervasive under-provision of public goods and services, especially those that hurt the low income population disproportionately.

Examples:

Acceleration of public infrastructure funding and increased absorptive capacityExpanded provision of high schools, complementing CCT extension Strengthening risk-pooling system against natural disastersHealth insurance (reduced out-of-pocket cost for catastrophic expenses)Under-investment in controlling public health threats

Pillar II

Slide38

Provision of time-limited “

pantawid” assistance to low income households. Examples

Development of innovative public-private micro-finance collaboration (i) tweaking the SEA-K program and focusing it on the social preparation of the near-poor to establish their track record and credit worthiness and (ii) providing interested firms this information.

Promotion of enterprise-based employment and human capital development program Voluntary minimum wage waiver program for unemployed and underemployed workers from targeted low-income familiesLabor regulatory reforms and development of innovative financial instruments to expand on-the-job training opportunitiesUse of time-limited CCT-like grant to facilitate transition of farmers to higher productivity and less climate-vulnerable work (to accompany rice importation policy reform)

Pillar III

Slide39

Cost of near-poor program:

OPTIMISTIC SCENARIO

RATPercent of HH in RATj

Total Near-poor HHs in RATjCost in 2014 Billion Pesos% of 2014 Pantawid Annual Budget

1.0-1.1

51,045,7355.46

91.0-1.20

10

2,039,079

8.70

14

1.0-1.30

14

2,908,779

10.55

17Notes: Total HHs in 2015 = 20,956,619.63RAT = ratio of household income to TPTUsed latest official household projection: year 2015Proposed 2014 4P's Budget = P62,614,247,297(Source: Official Gazette: Official Journal of the Philippine Government) CPI 2014 First Quarter(Source:http://www.census.gov.ph/sites/default/files/attachments/itsd/specialrelease/B30_14Q101.pdf)

Slide40

Assumptions re: optimistic scenarioFinancing of consumption deficit when poorDifferent

conditionalities for the near poor? (savings, job, etc_)No leakage due to mis-targeting No leakage due to corruptionImplicationsCost can be much larger

Potentially huge gains from good targetingNeed for experimentation

Slide41

Cost of near-poor program:

PESSIMISTIC SCENARIO (leakage)

RATPercent of HH in

RATjTotal Near-poor HHs in RATj

Cost in 2014 billion pesos

% of Pantawid Pamilya Budget1.0-1.1

5

1,045,735

13.24

21

1.0-1.20

10

2,039,079

24.92

40

1.0-1.30142,908,77934.4555

Slide42

Recommendations

1. Determine NPT, usingRAT = 1.10 for immediate initial use RAT =

1.20 after medium term experienceRAT = 1.28 as long term goal 2. Establish Near-Poor Strategy with the following featuresThree-pillar multi-level and convergent

Promoting self-reliance (near-poor have savings and other assets)Targeted and identified high-risk populationStart small, pilot, evaluate and scale up based evidence3. Develop targeting system that takes into account individual household’s circumstances: large potential gain from good targeting4. Ensure that Pantawid takes into account cyclical poverty and the near-poor phenomenon

5.

Develop linked panel databases that are maintained and updated in statistically sound manner and are highly accessible to researchers

Slide43

THANK YOU