Proxy Means Test (PMT) Configuration#

This guide is for implementers setting up Proxy Means Test scoring for poverty targeting. PMT uses observable household characteristics to estimate income or consumption levels without directly measuring them.

Quick start#

For a quick PMT targeting setup, you can use the sample configuration provided by the PMT targeting pack in SPP Studio:

  1. Go to SPP Studio → Packages and install the PMT Targeting pack

  2. Once installed, go to Studio → Expressions and click Submit for Review to activate it

  3. After activation, the PMT expression becomes available in CEL expression fields under Library

Note

The PMT targeting pack only appears under Library when targeting Group type criteria. This is because the pack is configured to target groups as an example — it will not show for Individual type targeting.

For manual configuration without the pack, follow the step-by-step guide below.

What is PMT?#

Proxy Means Test (PMT) is a poverty targeting method that:

  • Estimates household poverty based on easily observable indicators (assets, housing, demographics)

  • Avoids asking for sensitive income data (which is often unreliable)

  • Produces a poverty probability score that classifies households into poverty levels

  • Is transparent and replicable across assessments

PMT is widely used in conditional cash transfer programs, social safety nets, and poverty-targeted initiatives.

When to Use PMT#

Use PMT When

Don't Use PMT When

Targeting cash transfer programs

Emergency response (too slow)

National poverty registries

Small-scale pilot programs

Long-term program enrollment

Short-term crisis interventions

Verifiable, stable indicators available

Rapidly changing conditions

PMT Components#

A complete PMT model in OpenSPP includes:

  1. Household indicators - Observable characteristics (housing, assets, demographics)

  2. Indicator weights - Coefficients derived from household surveys

  3. Value mappings - How to score each indicator value

  4. Poverty thresholds - Score ranges that define poverty levels

Step-by-Step Configuration#

Step 1: Create the PMT Model#

Go to Scoring → Scoring Models and create a new model:

Field

Value

Name

2024 PMT - Rural

Code

PMT_2024_RURAL

Category

poverty

Calculation Method

weighted_sum

Active

☑ (after configuration complete)

Effective Date

2024-01-01

Description

PMT model for rural households based on 2023 HIES survey

Important: Keep models inactive until fully configured and tested.

Step 2: Add Housing Indicators#

Housing quality is a strong poverty predictor. Add these indicators:

Roof Material#

Field

Value

Code

roof_material

Name

Roof Material

Field Path

household.roof_material

Weight

0.15

Calculation Type

mapped

Required

Yes

Sequence

1

Value Mappings:

Input Value

Output Score

Notes

concrete

0

Strong, permanent material

metal_sheet

2

Durable, semi-permanent

tile

1

Good quality

wood

6

Light material

thatch

8

Temporary material

salvaged

10

Makeshift, poorest quality

Wall Material#

Field

Value

Code

wall_material

Name

Wall Material

Field Path

household.wall_material

Weight

0.15

Calculation Type

mapped

Required

Yes

Sequence

2

Value Mappings:

Input Value

Output Score

concrete

0

brick

1

wood

5

bamboo

7

salvaged

10

Floor Material#

Field

Value

Code

floor_material

Name

Floor Material

Field Path

household.floor_material

Weight

0.10

Calculation Type

mapped

Required

Yes

Sequence

3

Value Mappings:

Input Value

Output Score

tile

0

cement

2

wood

5

bamboo

7

earth

10

Step 3: Add Water & Sanitation Indicators#

Water Source#

Field

Value

Code

water_source

Name

Primary Water Source

Field Path

household.water_source

Weight

0.10

Calculation Type

mapped

Required

Yes

Sequence

4

Value Mappings:

Input Value

Output Score

Notes

piped_dwelling

0

Piped water in dwelling

piped_yard

2

Piped water in yard

public_tap

4

Public standpipe

protected_well

6

Protected well/spring

unprotected_well

8

Unprotected source

surface_water

10

River, lake, pond

Toilet Facility#

Field

Value

Code

toilet_facility

Name

Toilet Facility Type

Field Path

household.toilet_type

Weight

0.10

Calculation Type

mapped

Required

Yes

Sequence

5

Value Mappings:

Input Value

Output Score

flush_sewer

0

flush_septic

2

ventilated_pit

5

pit_latrine

7

none

10

Step 4: Add Asset Indicators#

Asset ownership indicates economic capacity. Use boolean indicators (owns or doesn't own).

Television#

Field

Value

Code

owns_tv

Name

Owns Television

Field Path

household.owns_tv

Weight

0.12

Calculation Type

mapped

Required

No

Default Value

False

Sequence

6

Value Mappings:

Input Value

Output Score

True

0

False

10

Refrigerator#

Field

Value

Code

owns_refrigerator

Name

Owns Refrigerator

Field Path

household.owns_refrigerator

Weight

0.15

Calculation Type

mapped

Required

No

Default Value

False

Sequence

7

Value Mappings:

Input Value

Output Score

True

0

False

10

Motorcycle/Scooter#

Field

Value

Code

owns_motorcycle

Name

Owns Motorcycle/Scooter

Field Path

household.owns_motorcycle

Weight

0.08

Calculation Type

mapped

Required

No

Default Value

False

Sequence

8

Value Mappings:

Input Value

Output Score

True

0

False

10

Step 5: Add Demographic Indicators#

Household Size#

Field

Value

Code

household_size

Name

Number of Household Members

Field Path

household.member_count

Weight

0.05

Calculation Type

range

Required

Yes

Sequence

9

Value Mappings (Range-Based):

Min

Max

Output Score

Notes

1

3

0

Small household

4

6

3

Medium household

7

10

7

Large household

11

99

10

Very large household

Step 6: Define Poverty Thresholds#

Based on your national poverty lines, set thresholds that classify households:

Min Score

Max Score

Classification Code

Classification Label

Color

0.0

25.0

NON_POOR

Non-Poor

green

25.01

50.0

NEAR_POOR

Near Poor

yellow

50.01

75.0

POOR

Poor

orange

75.01

100.0

EXTREME_POOR

Extremely Poor

red

Note: Adjust these thresholds based on:

  • National poverty line

  • Consumption survey data

  • Program targeting requirements

  • Available budget constraints

Step 7: Validate the Model#

Before activating, validate your configuration:

Check Weight Distribution#

Total weight should make sense for your scoring scale. If using 0-100 scale and each indicator scores 0-10, the weighted sum should align with your thresholds.

Example calculation:

  • If ALL indicators show maximum poverty (score 10 each)

  • Weighted sum = (10 × 0.15) + (10 × 0.15) + ... = sum of all weights × 10

  • Make sure this aligns with your "EXTREME_POOR" threshold (75-100)

Test with Sample Households#

Create test cases before going live:

Test Case 1: Extremely Poor Household

Indicator

Value

Score

Weight

Weighted Score

Roof Material

salvaged

10

0.15

1.5

Wall Material

salvaged

10

0.15

1.5

Floor Material

earth

10

0.10

1.0

Water Source

surface_water

10

0.10

1.0

Toilet Facility

none

10

0.10

1.0

Owns TV

False

10

0.12

1.2

Owns Refrigerator

False

10

0.15

1.5

Owns Motorcycle

False

10

0.08

0.8

Household Size

9

7

0.05

0.35

Total

1.0

9.85

Final Score: 9.85 × 10 = 98.5 → EXTREME_POOR ✓

Test Case 2: Non-Poor Household

Indicator

Value

Score

Weight

Weighted Score

Roof Material

concrete

0

0.15

0

Wall Material

concrete

0

0.15

0

Floor Material

tile

0

0.10

0

Water Source

piped_dwelling

0

0.10

0

Toilet Facility

flush_sewer

0

0.10

0

Owns TV

True

0

0.12

0

Owns Refrigerator

True

0

0.15

0

Owns Motorcycle

True

0

0.08

0

Household Size

4

3

0.05

0.15

Total

1.0

0.15

Final Score: 0.15 × 10 = 1.5 → NON_POOR ✓

Step 8: Activate and Score#

Once validated:

  1. Set Active = True on the scoring model

  2. Test on a few real households to verify calculations

  3. Run batch scoring for your full registry:

    • Go to Scoring → Actions → Batch Calculate Scores

    • Select your PMT model

    • Choose target population (e.g., all households)

    • Schedule or run immediately

Urban vs. Rural Models#

Different contexts need different indicators and weights:

Indicator

Rural Weight

Urban Weight

Rationale

Land ownership

0.20

0.05

More relevant in rural areas

Vehicle ownership

0.10

0.18

More indicative in urban areas

Housing materials

0.35

0.25

Stronger predictor in rural

Utilities access

0.20

0.35

More variation in urban

Best practice: Create separate models for urban and rural contexts if coefficients differ significantly.

Updating PMT Models#

When national consumption surveys are updated or poverty lines change:

  1. Create new model version (e.g., "PMT_2025_RURAL")

  2. Update weights based on new survey data

  3. Adjust thresholds to match new poverty lines

  4. Test on sample to compare with old model

  5. Activate new model and deactivate old

  6. Re-score population to get updated classifications

Integration with Programs#

Link PMT scores to program eligibility:

Option 1: Classification-Based#

In your eligibility manager:

Field

Value

Scoring Model

PMT_2024_RURAL

Required Classifications

POOR, EXTREME_POOR

Households classified as "Poor" or "Extremely Poor" are eligible.

Option 2: Score-Based#

In your eligibility manager:

Field

Value

Scoring Model

PMT_2024_RURAL

Minimum Score

50.0

Maximum Score

100.0

Households scoring 50 or above are eligible.

Option 3: Tiered Benefits#

Different benefit amounts based on poverty level:

Classification

Monthly Benefit

EXTREME_POOR

$150

POOR

$100

NEAR_POOR

$50

NON_POOR

Not eligible

Configure in entitlement manager using classification-based benefit tiers.

Common Patterns#

Pattern 1: Philippines Listahanan-Style#

  • 8-12 indicators

  • Focus on housing materials, assets, utilities

  • Thresholds: Poor vs. Non-Poor (simple binary)

  • Calculation: Weighted sum with survey-derived coefficients

Pattern 3: Hybrid PMT + Community Validation#

  • Base PMT score (weighted 70%)

  • Community validation score (weighted 30%)

  • Combines quantitative and qualitative assessment

Are You Stuck?#

Weights don't add up to 1.0? That's okay if you're using a 0-100 scale. Just make sure your thresholds align with the possible score range. Alternatively, normalize weights to sum to 1.0.

Scores are too clustered in one range? Review your value mappings. Make sure high-poverty indicators use higher scores (e.g., salvaged materials = 10, not 1).

Classifications don't match reality? Thresholds may need adjustment. Compare your score distribution to actual poverty rates in the population. Use histogram analysis to find natural cutoff points.

Urban and rural households scoring incorrectly? Create separate models. Urban households may have better housing but face higher costs. Rural households may own land but lack cash income.

Model validation failing? Check that all field paths are valid. Ensure test households have data for required indicators. Review calculation logs for errors.

Need to recalculate scores? You can trigger batch recalculation from Scoring → Actions. Existing scores are archived, not deleted, so you can compare changes.

Best Practices#

  1. Use survey data: Derive weights from household income/expenditure surveys when possible

  2. Test thoroughly: Run calculations on diverse household types before going live

  3. Document assumptions: Note why specific weights and thresholds were chosen

  4. Version models: Track model versions and effective dates for auditing

  5. Monitor distribution: After batch scoring, review the classification distribution

  6. Plan updates: Schedule regular model reviews (e.g., every 2-3 years)

  7. Communicate transparently: Make scoring methodology available to stakeholders

Security Considerations#

PMT scores reveal sensitive poverty information:

Access Level

Can Do

Program Managers

Create/modify models, view all scores

Field Officers

Calculate scores, view classifications

Data Entry

Input household data (no score access)

Beneficiaries

View their own classification (optional)

Configure access rights in Settings → Users & Groups → Scoring Permissions.

Next Steps#

See also: