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.

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 EAnd Strategy Development Ideas

Analysis of the

  Persistent poor

TransientsPersistentNon-Poor

% of households

8.9 38.6 52.45

APIS 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

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 Pamilya

In reality, many of them live difficult and precarious lives with high risk of becoming poor any time soon

These non-poor HHs called the “near-poor” in the literature is increasingly becoming a concern in the light of the inclusiveness goal

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?

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

 Definitions of Near-Poor in the Literature

Definition of Near-poorPoverty 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 1963

Orshansky (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 1963

Hokayem 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 poverty

Supplemental 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 line

ILO (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 Indonesia

Jellema and Noura (2012)

Lowest three deciles based on household consumption as covered by the Indonesian health insurance program intended for the poor and near-poor

lowest three deciles Harimurti et al. (2013)

Table 6. on study

Part 1

THE NEAR-POOR:

Concept, Practical Estimation Methods

And Profile

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

TPTPhilHealth 1.1 10

World Bank 1.2 20Vietnam 1.25 25

US Census Bureau (Hokayem and Heggeness,

2014)

1.25 25

Orshansky (pre-2014) 1.33 33

Authors’ Estimate1.28 (knife-edge) 281.37 (balik-balik) 37

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

Annual Per Capita Poverty Threshold

20% = World Bank

25% = Vietnam; US Census

33% = Orshansky

28% = Authors(Knife-Edge)

37% = Authors (Balik-Balik)

20,143

22,157 10% = PhilHealth

24, 172

24, 17925, 783

26, 790

27, 596

Am

ount

equ

ival

ent i

n P

hp

Comments on Previous RATs

• Based on income distribution of households

• NOT considering the risk of subsequent poverty

How did we estimate our RATs?

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

Methodology for Knife-Edge RAT Estimate of 1.28

• We use the following working definition

- Per capita income above the official total poverty threshold (TPT) at a given year, but at high risk of subsequently falling into poverty soon

- Metaphorically, 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%

Methodology for Knife-Edge RAT Estimate of 1.28

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

• RSKe = mean number of poverty episodes in 2007-2010 experienced by the non-poor households of 2004, as a percent of number of survey years

• RSKd = percent of non-poor households in 2004 becoming poor at least once in 2007-2010

• PCY = per capita household income

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

 RSKd RSKe

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)

Predicted Poverty Risk (RSKj) of Non-Poor Households, Using NHTS-PR Proxy Means Test Income Estimates

NPT =Mean income per

capitaa

RAT = NPT/TPTb

Frequency distribution

of non-poorc

Predictedd RSKd

Predictedd RSKe

10,163 1.05 14% 66% 37%11,129 1.15 12% 58% 32%12,095 1.25 11% 52% 28%13,072 1.35 9% 46% 24%14,039 1.45 8% 41% 21%15,004 1.55 6% 37% 19%15,968 1.65 5% 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%

Checking out the exclusion and inclusion error rates

RAT Exclusion Error Rate

Inclusion Error Rate

1.1 (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 .

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 1.37

• Exclusion 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

Exclusion/ inclusion errors are MANAGABLE.

Two ways of managing the errors

Increase RAT, as additional budget becomes available to reduce the exclusion error rates without unduly raising inclusion errors

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

•

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

  2010

Profiles All HHs Poor Near-Poor Non Poor

Distribution (%) 100 22.93 12.15 64.93

Total Income1 106,385 41,090 57,593 138,567

Wages and Salaries 1 42,900 14,435 22,842 56,703

Entrepreneurial Income1 29,282 15,583 20,566 35,749

Total Expenditures1 94,250 43,751 56,352 119,170

Education Expenditures1 4,715 971 1,639 6,613

Medical Expenditures1 3,060 462 948 4,373

Savings2 12,135 -2,661 1,240 19,397Notes1/ The household mean income and expenditure values are in pesos expressed in 2000 prices.2/ Savings = Total Income – Total Expenditure

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

  2010

Profiles All HHs Poor Near-Poor Non Poor

Education, assets and location% Did not graduate high school

59.4 83.28 75.5 47.95

% Own Lot 74.4 66.16 70.05 78.12

% with Own Electricity 86.12 65.31 78.84 94.84

% with Own Water 38.76 14.1 20.42 50.89

% Houses made of Strong/Predominantly Strong Materials

77.03 54.23 65.47 87.24

% Urban Household 38.82 17.77 24.5 48.92Notes1/ The household mean income and expenditure values are in pesos expressed in 2000 prices.2/ Savings = Total Income – Total Expenditure

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

Trends in poverty and near-poverty:

year total

number of households

poornear poor with RAT =

1.28near poor with

RAT = 1.37

number 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 3,685,135 21.17

1,899,010 10.91

2,429,240 13.96

2009

18,451,542 3,850,107 20.87

2,052,554 11.12

2,656,210 14.4

2012

21,425,737 4,264,584 19.9

2,328,293 10.87

2,992,348 13.97

Source: FIES 2000, 2006, 2009, 2012

Trends in poverty and near-poverty:

Source: FIES 2000, 2006, 2009, 2012

2000 2006 2009 20120

5,000,000

10,000,000

15,000,000

20,000,000

25,000,000

Absolute Number of Households

Poor HH Near Poor HH Non Poor HH

2000 2006 2009 2012

20.65 21.17 20.87 19.90

10.22 10.91 11.12 10.87

69.14 67.91 68.01 69.23

Percentage Distribution of Households

Poor HH Near Poor HH Non Poor

Part 2

Rationing Of Social Assistance

When Budget Is Insufficient To Cover All The Near Poor Households

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 CHALLENGE• Absence of RSK information in the Listahan data base• Desirability 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

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 2004-2010- Selected 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 Listahan

• The % change of RSKe per unit % change in the “predictors” (referred to as elasticity) can be estimated

• Predicted RSKe can be arranged in descending order

OLS Regression Equation for Generating “Predicted RSKe” of the Non-Poor Households of 2004

2004 values of “predictors”

Definition Coefficients Elasticity

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 size Number of member of HH 0.0169491*** 1.16699

(98.35)  Home built from Strong Materials

HH 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 study1/ * significant at the 10 percent level, ** at the 5 percent level, and *** at the 1 percent level. Dependent Variable is RSKe . 2/ Figures in parentheses are t-values.3/ See below for definition of variables

Continue: OLS Regression on the Non-Poor of 2004

Variables Definition Coefficients Elasticity

Average Wage Average 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 Rate

Regional Underemployment rate in 2004

0.0041283*** 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)  

Continue: OLS Regression on the Non-Poor of 2004

Variables Definition Coefficients Elasticity

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 0.2395Adjusted R-Squared 0.236Sample Size 3904

Continuation of Table 16. on study

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300%

10%

20%

30%

40%

50%

60%

52%

47%45% 44%

43% 43%42% 42%

40%38% 38%

36%34% 34% 34% 34%

33%32% 31%

29%

27%26% 26% 25% 25%

23%22%

21% 21%

15%

30 Random Samples for Illustration

Arrange households from high to low RSK

Predicted RSK arranged in descending order

 Socioeconomic Conditions Best(low risk)

Good(moderate risk)

Worst(high risk)

Predicted RSKe -0.03 0.35 0.63

Income per Capita 52596.66 13149.96 2123.689

Family Size 3 7 18Own Electricity 1 0 0Own Water 1 0 0Durables 6 0.06 0Average Wage 14636.34 2042.918 546.9167Regional Underemployment 12.07321 24.23235 32.3

At least High School Graduate

1 0 0

Age 37.36872 59.74662 98Age Squared 1515.391 3812.529 9604Urban 1 0 0Divorced 0 1 1

Illustrative combinations of socioeconomic conditions indicating high-low “predicted RSKe”, using 30 randomly chosen HHs

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

HOW?

1. Use equation to calculate “predicted” RSK for each household, using Listahanan data and integrating the new information in the Listahanan database

2. Rank households based on predicted RSK

3. Classify high risk near-poor households as those HHs above a specified RSK level

4. Give priority to the above households

• Planned application of RSK equation by NHTO

Part 3

Near -PoorSTRATEGY DEVELOPMENT

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 threshold• Near-Poor are also victims of policy and program failures• Desirability 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

ORGANIZING A NEAR-POOR STRATEGY

Principles- Focus on win-win interventions- Develop multi-level multi-dimensional convergent strategy- Promote self-reliance – leveraging own resources

Three Pillars- Reform of failing policies detrimental to poor and near-poor.- Addressing pervasive market failures- Time-limited (pantawid) social assistance

Pillar IReforming 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

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 capacity• Expanded provision of high schools, complementing CCT extension • Strengthening risk-pooling system against natural disasters• Health insurance (reduced out-of-pocket cost for catastrophic

expenses)• Under-investment in controlling public health threats

Pillar II

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 families- Labor regulatory reforms and development of innovative financial instruments to

expand on-the-job training opportunities- Use 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

Cost of near-poor program: OPTIMISTIC SCENARIO

RAT Percent of HH in RATj

Total Near-poor HHs in

RATj

Cost in 2014 Billion Pesos

% of 2014 Pantawid

Annual Budget

1.0-1.1 5 1,045,735 5.46 9

1.0-1.20 10 2,039,079 8.70 14

1.0-1.30 14 2,908,779 10.55 17

Notes: 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)

Assumptions re: optimistic scenario• Financing of consumption deficit when poor• Different conditionalities for the near poor?

(savings, job, etc_)

• No leakage due to mis-targeting • No leakage due to corruption• Implications

• Cost can be much larger• Potentially huge gains from good targeting• Need for experimentation

Cost of near-poor program: PESSIMISTIC SCENARIO (leakage)

RAT Percent of HH in RATj

Total Near-poor

HHs in RATj

Cost in 2014 billion

pesos

% of Pantawid Pamilya Budget

1.0-1.1 5 1,045,735 13.24 21

1.0-1.20 10 2,039,079 24.92 40

1.0-1.30 14 2,908,779 34.45 55

Recommendations1. Determine NPT, using

RAT = 1.10 for immediate initial use RAT = 1.20 after medium term experience RAT = 1.28 as long term goal

2. Establish Near-Poor Strategy with the following features- Three-pillar multi-level and convergent- Promoting self-reliance (near-poor have savings and other assets)- Targeted and identified high-risk population- Start small, pilot, evaluate and scale up based evidence

3. Develop targeting system that takes into account individual household’s circumstances: large potential gain from good targeting

4. 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

THANK YOU