## Section I. Detailed Technical Assessment and Recommendations

## Source details

**Canonical URL:** [Section I. Detailed Technical Assessment and Recommendations](https://www.imf.org/-/media/files/publications/tar/2024/english/tarea2024060.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tar/2024/english/tarea2024060.pdf.md)
- [Structured JSON version](/-/media/files/publications/tar/2024/english/tarea2024060.pdf.json)

---

### Introduction and mission scope
- TA mission: January 22–26, 2024, under the Data for Decisions (D4D) Trust Fund to assist Bangko Sentral ng Pilipinas (BSP) with developing hedonic methods for the Residential Property Price Index (RPPI).
- Objectives completed:
  - Assisted BSP to apply hedonic methods for the compilation of the RPPI using existing loan information.
  - Provided practical training to BSP Department of Economic Statistics (DES) staff.
  - Helped devise a strategy for dissemination of revised results.
  - Made recommendations on accessing and using more comprehensive administrative data sources.
- Participation and commitment:
  - A team of six DES staff attended daily meetings; senior BSP management, the IMF Resident Representative for the Philippines, and BSP staff attended the concluding presentation.
  - Experimental quality-adjusted indices were developed for condominiums in the NCR and houses in AONCR; these two segments represent approximately 86percent of the residential property market based on bank loans.

### Data sources: current status and recommended short/medium/long-term approach
- Findings:
  - BSP’s Bank Quarterly Report on Residential Real Estate Loans (RRELs) reporting became mandatory under BSP circular 892 (issued November 2015).
  - First data collection: 2014Q1; recommended index compilation start: 2016 because the template became mandatory in late 2015.
  - Average number of quarterly observations:
    - 2014: 4,800
    - 2015: 6,700
    - 2016: 9,800
    - Average over 2016–2023: 9,700
    - Lowest quarterly observations: 3,900 in Q2 2020 (onset of the Covid19 pandemic)
  - Loan data contains required property characteristics for hedonic methods: housing type, property type, location, floor area, lot size, number of bedrooms, number of floors, and effective age.
  - Template collects both appraised value and acquisition cost; acquisition cost is the most suitable price measure for the RPPI. Banks report acquisition cost denominated in Philippine peso (PHP).
  - Coverage limitations: loan data excludes (i) cash purchases, (ii) transactions financed by loans from developers, and (iii) transactions financed by the Home Development Mutual Fund (HDMF).
  - Real Property Information System (RPIS) is a new centralized database under DoF-BLGF’s Land Governance Reform Project; BSP involvement is planned in phase two and could broaden coverage.
- Recommended actions:
  - Use the loan data to compile the RPPI in the short to medium term.
  - Devise a longer-term strategy to broaden coverage by identifying new data sources (e.g., listings data from websites or administrative data).

### Compilation methods: hedonic approach, software, and processing
- Findings and recommendations:
  - Current published index: quarterly Residential Real Estate Price Index (RREPI) using a basic stratification approach; methods should be updated for compatibility with international statistical standards and to apply hedonic methods for accurate quality adjustment.
  - Adopted but unpublished 2022 mission recommendations (examples):
    - More detailed stratification for the NCR and AONCR.
    - Apply weights at a lower level in the aggregation structure.
    - Use acquisition costs instead of appraisal values.
    - Impute data for strata with no transactions in a quarter.
    - Use annually updated fixed weights based on share of total transaction value.
  - Current mission recommendation: use the time dummy hedonic method with a four-quarter window length to compile quality-adjusted indices.
  - Hedonic approach implications:
    - Replace existing outlier detection (interquartile range) with Cook’s Distance (regression-based approach).
    - Transfer data processing from MS Excel and Stata to R; initial R scripts were developed for BSP to refine.
- Data inspection and preprocessing:
  - Use visualizations (histograms, scatter plots) and descriptive statistics (mean and median price per region) to inform variable inclusion.
  - Exclude non-market transactions via a lower threshold on acquisition cost; test different thresholds per stratum.
  - Merge some property characteristic categories into homogeneous groups where per-quarter transaction counts are low.
- Specification and model application:
  - Authorities should decide on stratification for the RPPI; strata are used for compilation and dissemination and must contain sufficient observations each quarter.
  - BSP should specify the regression model for each stratum; demonstrated for two test strata:
    - Condominiums in the NCR.
    - Houses in AONCR.
  - Experimental hedonic indices: for condominiums in the NCR the hedonic index displayed less volatility than the published RREPI.
- Operational changes:
  - Transfer processing from Excel to R and set up monthly procedures and internal documentation.
  - Apply Cook’s Distance for outlier detection.
  - Calculate sub-indices and aggregate to higher level indices, including national RPPI, using annual weights.
  - Apply robustness checks to regression results for each stratum.

### Stratification, outlier detection, regression specification, robustness, and aggregation
- Stratification:
  - Extensive discussion on appropriate stratification; authorities should invite feedback from users and stakeholders.
  - Strata should contain sufficient observations each quarter to ensure stable indices.
- Outlier detection:
  - Recommendation: use Cook’s Distance instead of interquartile range by quarter and strata.
  - Application steps:
    - Run regression including all relevant characteristics to calculate Cook’s Distance for each observation.
    - Rerun regression excluding observations with high Cook’s Distance.
    - Suggested threshold: four divided by the number of observations used in the regression.
    - Apply outlier detection separately for each stratum and each rolling window (four quarters of data).
- Regression model specification:
  - BSP should specify the regression model for each stratum.
  - Demonstrated specifications:
    - Condominiums in the NCR:
      - Dependent variable: log price.
      - Regressors: (i) log of floor area, (ii) location, (iii) number of bedrooms, and (iv) new v pre-owned v foreclosed.
    - Houses in AONCR:
      - Same as above with lot area added.
  - Data quality note: investigate quality of the number-of-bedrooms variable; zero bedrooms may indicate missing values for bank submissions before 2022.
  - Method for calculating the price index from time-dummy coefficients was demonstrated.
- Robustness and diagnostics:
  - Ensure estimated coefficients or shadow prices are plausible in sign and magnitude.
  - Conduct stability checks on coefficients for each rolling window and each stratum.
  - Analyze regression residuals for trends.
- Aggregation and weights:
  - Aggregate sub-indices (strata indices) to higher-level indices including the national RPPI.
  - Weighting approach:
    - Use annually updated fixed weights.
    - Weight reference period: the previous calendar year (example: indices for the four quarters of 2023 should use 2022 as the weight reference period).
    - Weights calculated by summing acquisition cost for transactions in a given stratum and dividing by total acquisition cost for all strata during the weight reference period.

### Experimental hedonic indices and commercial property considerations
- Experimental hedonic indices and findings:
  - Experimental indices compiled for two strata.
  - Key details:
    - Figure 1: index for condominiums in the NCR using hedonic methods (calculated in a single regression rather than a four-quarter rolling window due to time constraints).
    - Indices were also compiled using data from Q1 2014, whereas BSP should use loan data from 2016 onwards.
    - Figure 1 shows published RREPI and updated stratification index; both hedonic and updated stratification indices display less volatility than the published RREPI.
    - Figure 2 shows hedonic indices for condominiums in the NCR and houses in AONCR.
    - These two strata account for about 86 percent of the overall weight of the national RPPI.
  - Recommendation: complete calculation of sub-indices using hedonic methods and aggregate to the national RPPI; calculate intermediate higher-level indices (e.g., national index for condominiums) and compare with current published indices.
- Commercial property considerations:
  - Further analyze coverage of loan data for commercial property before computing price indices.
  - Issues:
    - Commercial transactions often not financed through bank loans; larger transactions may be financed via SPVs and REITs.
    - Representativity concerns for loan information.
    - Fewer observations and heterogeneous transactions increase complexity and potential volatility/bias.
    - Data requirements for hedonic methods are greater for commercial property; additional characteristics likely needed (e.g., prime v non-prime, existing tenancy, age of building, building material).
  - Current template captures appraisal cost of land only; full appraisal value or acquisition cost of the property is not collected—insufficient for a Commercial Property Price Index (CPPI).
  - Recommendation: amend data collection form to collect full appraisal value/acquisition cost and additional characteristics; meet with real estate experts and industry stakeholders.
  - Prioritization: disseminating revised RREPI should be prioritized ahead of developing commercial property indices.

### Dissemination, stakeholder engagement, and publication timetable
- Findings and recommendations:
  - Published RREPI should be revised as soon as possible; agreed workplan envisages publishing updated indices by December 2024.
  - Communication with users and stakeholders required during revision:
    - Provide advance notice of revision and solicit feedback.
    - Draft technical and non-technical communication documents explaining reasons and extent of revisions.
    - Consider harmonizing the indicator name from RREPI to RPPI.
    - Draft and publish a detailed methodological document on the authorities’ website.
  - Obtain approval from the BSP Monetary Board prior to publication.
- Key dissemination milestones and target dates (from RPPI Workplan Table 2):
  - Feb-2024: H — The BSP should use the loan data to compile the RPPI in the short to medium term.
  - Jun-2025: M — Devise a longer-term strategy to broaden coverage by identifying new data sources.
  - Feb-2024: H — Decide on the stratification to be used for the RPPI.
  - Mar-2024: H — Use the time dummy hedonic method with a four-quarter window length.
  - May-2024: H — Specify the regression model for each stratum and apply robustness checks.
  - Jun-2024: H — Apply Cook’s Distance for outlier detection.
  - Jul-2024: H — Calculate sub-indices and aggregate to higher level indices, including national RPPI, using annual weights.
  - Aug-2024: H — Transfer processing from Excel to R and set up monthly procedures and internal documents.
  - Sep-2024: M — Consult with users/stakeholders on revisions to published RREPI.
  - Sep-2024: H — Re-design the press release with a new structure of graphs and tables.
  - Sep-2024: H — Prepare communication documents, both technical and non-technical.
  - Oct-2024: H — Obtain approval from the BSP Monetary Board.
  - Nov-2024: H — Run Q3 2024 data using hedonic methods and prepare final publication on BSP website.
  - Nov-2024: M — Provide advance notice to users of upcoming revisions.
  - Dec-2024: H — Publish the revised RPPI.
- Priority recommendations and target dates (Table 1):
  - Feb-2024: Decide on the stratification to be used for the RPPI. — Responsible: BSP
  - Jul-2024: Calculate the sub-indices and aggregate to higher level indices, including national RPPI, using annual weights. — Responsible: BSP
  - Dec-2024: Publish the revised RPPI. — Responsible: BSP

### Technical and operational recommendations (concise)
- Use acquisition cost (contract price/actual selling price at time of contract) denominated in Philippine peso as price measure.
- Implement time dummy hedonic method with a four-quarter rolling window for quality adjustment.
- Replace interquartile-range outlier detection with Cook’s Distance (regression-based).
- Transfer data cleaning, outlier detection, index calculation, and aggregation to R; refine mission-provided R scripts.
- Exclude non-market transactions using stratum-specific lower acquisition-cost thresholds.
- Group sparse categorical variables into meaningful homogeneous groups to ensure model stability.
- Specify and document regression models for each stratum and run robustness checks; publish detailed methodological documentation.
- Engage users and stakeholders, provide advance notice and explanatory material, and re-design dissemination outputs (press releases, graphs, tables).
- Consider renaming published indicator from RREPI to RPPI for harmonization.

### Officials met during the mission
- Mr. Francisco G. Dakila, Jr., Deputy Governor, Monetary and Economics Sector, BSP
- Ms. Iluminada T. Sicat, Senior Assistant Governor, Monetary Policy Sub-sector, BSP
- Mr. Redentor Paolo M. Alegre, Jr., Senior Director, Department of Economic Statistics, BSP
- Ms. Haydee R. Paulino, Deputy Director, Department of Economic Statistics, BSP
- Ms. Willa Boots J. Tolo, Bank Officer V, Department of Economic Statistics, BSP
- Mr. Jeremy L. De Jesus, Bank Officer V, Department of Economic Statistics, BSP
- Ms. Gloria A. Cubinar, Bank Officer V, Department of Economic Statistics, BSP
- Mr. Richer King R. Supnet, Bank Officer IV, Department of Economic Statistics, BSP
- Mr. Gian Paolo T. De La Torre, Bank Officer II, Department of Economic Statistics, BSP
- Mr. Christian Kenneth R. Collado, Research Analyst II, Department of Economic Statistics, BSP
- Ms. Hazel Anne V. Paguirigan, Junior Central Bank Associate, Department of Economic Statistics, BSP

*Source: Section I. Detailed Technical Assessment and Recommendations (tarea2024060).*

### Section I. Detailed Technical Assessment and Recommendations ...................................................... 4

### Section I. Detailed Technical Assessment and Recommendations

### Introduction and mission scope
- A technical assistance (TA) mission was conducted from January 22–26, 2024, under the Data for Decisions (D4D) Trust Fund to assist the Bangko Sentral ng Pilipinas (BSP) with developing hedonic methods for the Residential Property Price Index (RPPI).
- Objectives completed:
  - Assisted BSP to apply hedonic methods for the compilation of the RPPI using existing loan information.
  - Provided practical training to BSP Department of Economic Statistics (DES) staff.
  - Helped devise a strategy for dissemination of revised results.
  - Made recommendations on accessing and using more comprehensive administrative data sources.
- Participation and commitment:
  - A team of six DES staff attended daily meetings; senior BSP management, the IMF Resident Representative for the Philippines, and BSP staff attended the concluding presentation.
  - Experimental quality-adjusted indices were developed for condominiums in the NCR and houses in AONCR; these two segments represent approximately 86percent of the residential property market based on bank loans.

### Data sources: current status and recommended short/medium/long-term approach
Findings:
- The BSP’s Bank Quarterly Report on Residential Real Estate Loans (RRELs) reporting became mandatory under BSP circular 892 (issued November 2015).
- First data collection: 2014Q1; recommended index compilation start: 2016 because the template became mandatory in late 2015.
- Average number of quarterly observations:
  - 2014: 4,800
  - 2015: 6,700
  - 2016: 9,800
  - Average over 2016–2023: 9,700
  - Lowest quarterly observations: 3,900 in Q2 2020 (onset of the Covid19 pandemic)
- Loan data contains required property characteristics for hedonic methods: housing type (condominium, duplex, townhouse, etc.), property type (new, pre-owned, foreclosed), location (island, region, municipality, etc.), floor area, lot size, number of bedrooms, number of floors, and effective age.
- Template collects both appraised value and acquisition cost; acquisition cost is the most suitable price measure for the RPPI. Banks report acquisition cost denominated in Philippine peso (PHP).
- Coverage limitations: loan data excludes (i) cash purchases, (ii) transactions financed by loans from developers, and (iii) transactions financed by the Home Development Mutual Fund (HDMF).
- Real Property Information System (RPIS) is a new centralized database under DoF-BLGF’s Land Governance Reform Project; BSP involvement is planned in phase two and could broaden coverage.

Recommended actions:
- The BSP should use the loan data to compile the RPPI in the short to medium term.
- The authorities should devise a longer-term strategy to broaden coverage by identifying new data sources, e.g., listings data from websites or administrative data.

### Compilation methods: hedonic approach, software, and processing
Findings and recommendations:
- Current published index: quarterly Residential Real Estate Price Index (RREPI) using a basic stratification approach; methods should be updated to ensure compatibility with international statistical standards and to apply hedonic methods for accurate quality adjustment.
- Previous 2022 mission recommendations that were adopted (but not reflected in published RREPI) included:
  - More detailed stratification for the NCR and AONCR (aggregation of elementary indices at district or city/municipality level).
  - Apply weights at a lower level in the aggregation structure.
  - Use acquisition costs instead of appraisal values.
  - Impute data for strata with no transactions in a quarter.
  - Use annually updated fixed weights based on share of total transaction value.
- Current mission recommendation: use the time dummy hedonic method with a four-quarter window length to compile quality-adjusted indices; hedonic approach allows for quality adjustment within and across strata.
- Hedonic approach implications:
  - Replace existing outlier detection (interquartile range) with Cook’s Distance (regression-based approach).
  - Transfer data processing from MS Excel and Stata to a suitable statistical software package (R); initial R scripts were developed for BSP to refine.
- Data inspection and preprocessing:
  - Visualizations (histograms, scatter plots) and descriptive statistics (mean and median price per region) were produced to inform variable inclusion in regression models.
  - Exclude non-market transactions via a lower threshold on acquisition cost; test different thresholds per stratum (higher thresholds for higher-priced strata, lower for lower-priced strata).
  - Merge some property characteristic categories into homogeneous groups where per-quarter transaction counts are low (location, property type, number of bedrooms, number of floors).

Specification and model application:
- Authorities should decide on stratification for the RPPI; strata are used for both compilation and dissemination and must contain sufficient observations each quarter.
- BSP should specify the regression model for each stratum; the mission demonstrated model specification for two test strata:
  - Condominiums in the NCR.
  - Houses in AONCR.
- Experimental hedonic indices: for condominiums in the NCR the hedonic index displayed less volatility than the published RREPI.

Operational changes:
- Transfer processing from Excel to R and set up monthly procedures and internal documentation.
- Apply Cook’s Distance for outlier detection.
- Calculate sub-indices and aggregate to higher level indices, including national RPPI, using annual weights.
- Apply robustness checks to regression results for each stratum.

### Dissemination, stakeholder engagement, and publication timetable
Findings and recommendations:
- The published RREPI should be revised as soon as possible; agreed workplan envisages publishing updated indices by December 2024.
- Communication with users and stakeholders is required during the revision process:
  - Provide advance notice of revision and solicit feedback.
  - Draft communication documents (technical and non-technical) explaining reasons and extent of revisions.
  - Consider harmonizing the indicator name from RREPI to RPPI.
  - Draft and publish a detailed methodological document on the authorities’ website.
- Obtain approval from the BSP Monetary Board prior to publication.

Key dissemination milestones and target dates (from RPPI Workplan Table 2):
- Feb-2024: H — The BSP should use the loan data to compile the RPPI in the short to medium term.
- Jun-2025: M — Devise a longer-term strategy to broaden coverage by identifying new data sources.
- Feb-2024: H — Decide on the stratification to be used for the RPPI.
- Mar-2024: H — Use the time dummy hedonic method with a four-quarter window length.
- May-2024: H — Specify the regression model for each stratum and apply robustness checks.
- Jun-2024: H — Apply Cook’s Distance for outlier detection.
- Jul-2024: H — Calculate sub-indices and aggregate to higher level indices, including national RPPI, using annual weights.
- Aug-2024: H — Transfer processing from Excel to R and set up monthly procedures and internal documents.
- Sep-2024: M — Consult with users/stakeholders on revisions to published RREPI.
- Sep-2024: H — Re-design the press release with a new structure of graphs and tables.
- Sep-2024: H — Prepare communication documents, both technical and non-technical.
- Oct-2024: H — Obtain approval from the BSP Monetary Board.
- Nov-2024: H — Run Q3 2024 data using hedonic methods and prepare final publication on BSP website.
- Nov-2024: M — Provide advance notice to users of upcoming revisions.
- Dec-2024: H — Publish the revised RPPI.

Priority recommendations and target dates (Table 1):
- Feb-2024: Decide on the stratification to be used for the RPPI. — Responsible: BSP
- Jul-2024: Calculate the sub-indices and aggregate to higher level indices, including national RPPI, using annual weights. — Responsible: BSP
- Dec-2024: Publish the revised RPPI. — Responsible: BSP

### Technical and operational recommendations (concise)
- Use acquisition cost (contract price/actual selling price at time of contract) denominated in Philippine peso as price measure.
- Implement time dummy hedonic method with a four-quarter rolling window for quality adjustment.
- Replace interquartile-range outlier detection with Cook’s Distance (regression-based).
- Transfer data cleaning, outlier detection, index calculation, and aggregation to R; refine mission-provided R scripts.
- Exclude non-market transactions using stratum-specific lower acquisition-cost thresholds.
- Group sparse categorical variables into meaningful homogeneous groups to ensure model stability.
- Specify and document regression models for each stratum and run robustness checks; publish detailed methodological documentation.
- Engage users and stakeholders, provide advance notice and explanatory material, and re-design dissemination outputs (press releases, graphs, tables).
- Consider renaming published indicator from RREPI to RPPI for harmonization.

*Source: Section I. Detailed Technical Assessment and Recommendations (tarea2024060).*

### 26. The authorities should decide on the stratification to be used for the RPPI. There was

### 26. The authorities should decide on the stratification to be used for the RPPI. There was

### Stratification for RPPI
- Extensive discussion on appropriate stratification for the property loan data; various options assessed.
- Importance: strata used for both compilation and dissemination of indices.
- Authorities should invite feedback from users and stakeholders to the extent possible.
- Strata should contain sufficient observations each quarter to ensure stable indices.

### Outlier detection
- Recommendation: use Cook’s Distance for outlier detection rather than the existing interquartile range by quarter and strata.
- Rationale: regression-based method identifies outliers with greater precision and preserves the sample to the extent possible.
- Application steps:
  - Run a regression model including all relevant characteristics to calculate Cook’s Distance for each observation.
  - Rerun the regression excluding observations displaying a high Cook’s Distance value.
  - Suggested threshold for outlier determination: four divided by the number of observations used in the regression.
  - Apply outlier detection separately for each stratum and each rolling window, i.e., four quarters of data.

### Regression model specification
- BSP should specify the regression model for each stratum.
- Demonstrated model specifications for two test strata:
  - Condominiums in the NCR:
    - Dependent variable: log price.
    - Regressors: (i) log of floor area, (ii) location, (iii) number of bedrooms, and (iv) new v pre-owned v foreclosed.
  - Houses in AONCR:
    - Same as above with lot area added to the regression model.
- Remaining strata can be specified using a similar approach.
- Data quality note: investigate quality of the number-of-bedrooms variable; zero bedrooms may indicate a missing value rather than a true zero for bank submissions before 2022.
- Method for calculating the price index from coefficients on time dummies was demonstrated.

### Robustness and diagnostic checks
- Apply robustness checks to regression results:
  - Ensure estimated coefficients or shadow prices are plausible in terms of sign and magnitude.
  - Conduct stability checks on coefficients for each rolling window and each stratum.
  - Analyze regression residuals to ensure no specific trends in the data.

### Aggregation and weights
- Sub-indices (strata indices) should be aggregated using weights to higher-level indices including the national RPPI.
- Weighting approach:
  - Use annually updated fixed weights.
  - Weight reference period: the previous calendar year (example: indices for the four quarters of 2023 should use 2022 as the weight reference period).
  - Weights are calculated by summing the acquisition cost for transactions in a given stratum and dividing by the total acquisition cost for transactions in all strata during the weight reference period.

### Experimental hedonic indices and findings
- Experimental quality-adjusted hedonic indices were compiled for two strata during the mission.
- Key details:
  - Figure 1: index for condominiums in the NCR calculated using hedonic methods.
  - Due to time constraints, index calculated in a single regression rather than a four-quarter rolling window.
  - Indices were also compiled using data from Q1 2014, whereas the BSP should use loan data from 2016 onwards.
  - Figure 1 also shows the published RREPI and the index computed using the updated stratification method implemented after the 2022 mission recommendations.
  - Both the hedonic index and the updated stratification index display less volatility than the published RREPI.
  - Mission recommended the hedonic approach because it applies more detailed quality adjustment within strata.
  - Figure 2 shows hedonic indices for condominiums in the NCR and houses in AONCR.
  - These two strata account for about 86 percent of the overall weight of the national RPPI.
- Recommendation: BSP should complete calculation of sub-indices using hedonic methods and aggregate to the national RPPI. Intermediate higher-level indices (e.g., national index for condominiums) can also be calculated and compared with currently published indices and those from the updated stratification.

### Commercial property considerations
- Complete further analysis of coverage of loan data for commercial property before computing price indices.
- Issues:
  - Commercial transactions often not financed through bank loans; larger transactions may be financed via special purpose vehicles (SPVs) and real estate investment trusts (REITs).
  - Representativity concerns for loan information.
  - Fewer observations and heterogenous transactions increase complexity and potential volatility/bias in indices.
  - Data requirements for hedonic methods are greater for commercial property; additional characteristics likely needed (e.g., prime v non-prime, existing tenancy, age of building, building material).
- Current data collection template captures appraisal cost of land only; full appraisal value or acquisition cost of the property is not collected—insufficient for a Commercial Property Price Index (CPPI).
- Recommendation: amend data collection form to collect full appraisal value/acquisition cost and additional characteristics; meet with real estate experts and industry stakeholders to understand market operation.
- Prioritization: dissemination of revised RREPI should be prioritized ahead of developing commercial property indices due to complexity and resource requirements.

### Dissemination and communication
- Published RREPI should be revised as soon as possible.
  - Workplan envisages publishing updated price indices computed following the hedonic approach by December 2024.
  - Published RREPI displays significant volatility; hedonic methods ensure compatibility with international standards.
  - Weaknesses of existing published indices include: (i) use of appraisal values rather than acquisition costs, (ii) varying weights from quarter-to-quarter, and (iii) limited quality adjustment.
- BSP should communicate with users and stakeholders during the revision process:
  - Provide advance notice of the revision and request feedback.
  - Draft technical and non-technical communication documents explaining reasons and extent of revisions.
  - Consider renaming the indicator from the RREPI to the RPPI for harmonization with other countries to signal methodological change.
  - Re-design the quarterly press release with a new structure of charts and tables.
- Publish a detailed methodological document on the authorities’ website including: data source, coverage, data quality and cleaning, outlier detection, stratification, weighting information, and methods for index compilation and aggregation.

### Recommended Actions (as stated)
- The BSP should use the time dummy hedonic method with a four-quarter window length to compile quality-adjusted indices.
- A suitable statistical software package should be used for the application of hedonic methods for the RPPI.
- Non-market transactions should be excluded from the loan dataset.
- Some property characteristics should be merged into homogenous groups.
- The authorities should decide on the stratification to be used for the RPPI.
- The authorities should use Cook’s Distance for outlier detection.
- The BSP should specify the regression model for each stratum.
- Robustness checks should be applied to the results of the regressions.
- The sub-indices (i.e., strata indices) should be aggregated using weights to higher-level indices including the national RPPI.
- The BSP should complete the calculation of the sub-indices using hedonic methods and aggregate the results to the national RPPI.

### Officials met during the mission
- Mr. Francisco G. Dakila, Jr., Deputy Governor, Monetary and Economics Sector, BSP
- Ms. Iluminada T. Sicat, Senior Assistant Governor, Monetary Policy Sub-sector, BSP
- Mr. Redentor Paolo M. Alegre, Jr., Senior Director, Department of Economic Statistics, BSP
- Ms. Haydee R. Paulino, Deputy Director, Department of Economic Statistics, BSP
- Ms. Willa Boots J. Tolo, Bank Officer V, Department of Economic Statistics, BSP
- Mr. Jeremy L. De Jesus, Bank Officer V, Department of Economic Statistics, BSP
- Ms. Gloria A. Cubinar, Bank Officer V, Department of Economic Statistics, BSP
- Mr. Richer King R. Supnet, Bank Officer IV, Department of Economic Statistics, BSP
- Mr. Gian Paolo T. De La Torre, Bank Officer II, Department of Economic Statistics, BSP
- Mr. Christian Kenneth R. Collado, Research Analyst II, Department of Economic Statistics, BSP
- Ms. Hazel Anne V. Paguirigan, Junior Central Bank Associate, Department of Economic Statistics, BSP

*IMF | Technical Assistance Report – The Philippines RPPI*

---


_Source: https://www.imf.org/-/media/files/publications/tar/2024/english/tarea2024060.pdf_