## BANGLADESH REPORT ON RESIDENTIAL PROPERTY PRICE INDICES MISSION (DECEMBER 8–12, 2019)

## Source details

**Canonical URL:** [BANGLADESH REPORT ON RESIDENTIAL PROPERTY PRICE INDICES MISSION (DECEMBER 8–12, 2019)](https://www.imf.org/-/media/files/publications/cr/2020/english/1bgdea2020002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2020/english/1bgdea2020002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2020/english/1bgdea2020002.pdf.json)

---

### Summary of Mission Outcomes and Priority Recommendations
- Purpose: Assist Bangladesh Bank (BB) in progressing on the compilation of a residential property price index (RPPI); compile an experimental RPPI; improve data for RPPI compilation.
- Importance: RPPI identified as critical for financial stability policy analysis and macroprudential measures (Second Phase of the G-20 Data Gaps Initiative and guidance on Financial Soundness Indicators). RPPI is also a macro-economic indicator of growth and key for understanding financial market conditions.
- Current status: Bangladesh lacks a reliable official measure of trends in residential property prices.
- Scope: BB aims to compile a quarterly RPPI covering new flats for Dhaka; mission recommended retaining secondary market and other-region data for future expansion.
- Methodology decision: Compile RPPI with the time dummy method with a rolling window of four quarters; index should be compiled with data from 2007 onwards.
- Tools and training: Mission adapted R scripts from the RPPI Practical Compilation Guide to BB data; recommended use of R for the compilation process.
- Data collection improvements: BB plans a new data collection system starting July 2020 to expand geographic coverage, dwelling types, and sample size.
- Dissemination timeline: BB may explore publication of the RPPI with new data from July 2021; internal compilation to continue in the meantime.

Priority Recommendations (selected)
- January/2020: Compile the RPPI with the time dummy with rolling window. — Responsible: BB
- July/2020: New data collection. — Responsible: BB

### Governance
- Team capacity and training:
  - BB has a team of nine staff from different departments compiling the index.
  - Three staff attended IMF courses on RPPI in 2017 and 2019.
  - Recommendation: At least six team members should have some knowledge in R.
- Software recommendation:
  - Use R because it performs econometric regressions and price index calculations in a single script/environment.
- Coverage and data retention:
  - Current focus: quarterly RPPI covering new flats in Dhaka due to insufficient data in other cities/types.
  - Recommendation: continue obtaining all data (including secondary market and rest of country) to store for future expansion.
- Base year alignment:
  - RPPI base year should match CPI and GDP base years to facilitate benchmarking.
  - BBS is aiming at rebasing both CPI and GDP to the fiscal year July 2015–June 2016 during 2020.

### Compilation with the Current Dataset
- Data source and volume:
  - BB receives quarterly data from Delta Brac Housing Finance Corporation, ltd (DBH) on house loans.
  - DBH dataset contains valuation price, size, and location within the city.
  - Data available since 1998 with over 33,000 observations in total.
  - Index work began in 2017.
- Location observations for period 1Q2017 (counts by location code):
  - DHA01 206
  - DHA02 99
  - DHA03 98
  - DHA04 33
  - DHA05 17
  - DHA06 25
  - DHA07 27
  - DHA08 47
  - DHA09 17
  - DHA10 118
  - DHA11 60
  - DHA12 14
  - DHA13 4
  - DHA14 15
  - DHA15 11
  - DHA16 6
  - DHA17 28
  - DHA18 19
  - DHA20 4
  - DHA21 3
- Clustering recommendation:
  - Cluster the 21 locations to achieve at least around 50 observations per quarter per location cluster; clustering should be based on economic criteria (e.g., income level).
- Methodology training and guidance:
  - Mission provided training on hedonic methods, chain linking, and rebasing.
  - Hedonic methods are preferred to address quality changes in the mix of dwellings.
  - BB staff trained on rebasing and chain linking; R scripts adapted for BB data.
- Recommended compilation start date:
  - Start compilation from 2007 using the time dummy method with a rolling window of four quarters (data prior to 2007 have few observations).
- Sample size note:
  - Each quarter has less than 1,000 observations; with a small sample the time dummy yields a more robust and less volatile indicator.
- Data cleaning guidance and descriptive statistics for 1Q2017:
  - R script provided for basic data manipulations: treating outliers, missing values, and duplicates.
  - Descriptive statistics before and after treatment:
    - Before: Minimum 555.581, Q1 3500, Median 4200, Mean 4760.586, Q3 5500, Maximum 55000
    - After: Minimum 826.4463, Q1 3500, Median 4200, Mean 4653.867, Q3 5500, Maximum 14761.9
- RPPI recompiled:
  - Mission recompiled the index from 2007 with base year July 2015/June 2016 using time dummy hedonic method; BB staff trained on script manipulation.

### Improving the Data Collection
- Questionnaire/template guidance:
  - BB prepared a survey template reviewed by the mission and presented to potential data providers.
  - Recommendation: keep the template as short as possible to avoid respondent burden.
  - Avoid open questions; use objective, measurable answers (yes/no or numbers); avoid semantic free text to ensure machine readability.
- Proposed data structure (variables and types):
  - Date — Date of the Approval Date
  - Identification number of Dev. Authority Approval — ID number — Not to be used in regression
  - Address — Categorical
  - Monthly income — Average monthly income — Numerical
  - City — Categorical
  - Thana — Categorical
  - District — Categorical
  - Post code — Categorical
  - Floor level — Categorical
  - Height of the building — Number of stories — Categorical
  - Face Direction — South/North/West/East — Categorical
  - Total size — Square feet — Numerical
  - Purchase price — Net of costs — Numerical
  - Valuation price — Net of costs — Numerical
  - Interest rate — Number — Numerical
  - No. Bedrooms — Number — Categorical
  - No. Bathrooms — Number — Categorical
  - No. lifts — Number — Categorical
  - Parking — Number — Categorical
  - Can two cars easily pass at the same time in the street in front of your house? — Y/N — Categorical
  - Pipeline gas connection — Y/N — Categorical
  - Park — Within 10 min walking — Categorical
  - Hospitals — Within 10 min walking — Categorical
  - Schools — Within 10 min walking — Categorical
  - Public transports — Within 10 min walking — Categorical
  - Brand of Developer — Drop down menu with the main developers — Categorical
  - Construction year — Date
  - Resale — Y/N
  - Government owned plot — Y/N
- Official request and timing:
  - BB should send an official data request to data providers; providers require an official request for internal procedures.
  - Request should be sent with the final data structure no later than March to allow system adaptation time.
- Transmission and storage:
  - Data to be transmitted and stored using the Enterprise Data Warehouse (EDW) from July 2020.
  - Data providers need time to adapt internal systems before EDW automated transmission begins.

### Dissemination
- Publication timing:
  - BB may explore beginning publication of the RPPI with new data from July 2021.
  - Regular compilation will continue for internal use before public dissemination.
- Release schedule:
  - The RPPI is released on a quarterly basis, 45 days after the end of the reference period.
- Guidance:
  - Mission provided recommendations for dissemination as exemplified in the RPPI Guide and supplied a draft technical note (Appendix I).

Priority actions / milestones and target completion dates (excerpt)
- The 21 locations are clustered. — January 2020
- The RPPI is compiled with the time dummy method with a rolling window of four quarters. — January 2020
- The provisional data structure will be sent to the data providers. — January 2020
- An official data request is sent by the BB Governor to the data providers. — March 2020
- New data collection system is implemented. — July 2020

### Appendix I — RPPI Technical Note: Methodology and Operational Details
- Background and coverage:
  - RPPI measures the price evolution for residential new apartments in Dhaka.
  - Coverage is limited to Dhaka city and covers the primary (new dwellings) market for apartments.
- Data sources and periodicity:
  - BB collects data from Delta Brac Housing Finance Corporation, ltd (DBH) on loans for new apartments.
  - The RPPI is disseminated on a quarterly basis.
- Base period:
  - The RPPI is an annually chain-linked price index that uses the year 2015/106 as reference year.
  - The average of the quarterly indexes for 2015/16 equals 100.
- Conceptual methodology:
  - Key characteristics of the dwellings are required to assure a constant quality index.
  - Quality adjustment techniques are required because the same dwelling is sold infrequently and properties are heterogeneous.
- Time Dummy Hedonic Method (specification and implementation):
  - Log-linear specification for each stratum:
    - ln푝푛푡 = 훽0 + ∑훿푡퐷푛푡 + ∑훽푘푧푛푘푡 + 휀푛푡 (퐾 푘=1, 푇 푡=1)
  - Index derivation for each stratum:
    - 퐼푡 = exp(훿̂푡) ∗ 100
  - Rolling window rationale:
    - When a new period is added and the model re-estimated, older period time dummy estimates 훿̂푡 change; therefore use a rolling window.
    - Normally for quarterly indices the “shadow” prices of the characteristics (훽̂) are kept fixed for at least a year.
    - Every quarter a regression is estimated with data from the current quarter and the previous three quarters.
    - The indices from the “new window” and the “previous window” are chained by using the last overlap period between the two windows.
- Chaining example (rolling window quarterly index excerpts):
  - 2018Q1 100.0   100.0
  - 2018Q2 101.2 100.0  101.2
  - 2018Q3 101.1 101.3 100.0 101.1
  - 2018Q4 100.9 101.2 100.8 100.9
  - 2019Q1 101.0 100.3 100.7=100.9/101.2x101.0
  - 2019Q2 101.4 101.8=100.7/100.3x101.4

*Source: 1bgdea2020002 - Section 2*

### Section 1

### BANGLADESH REPORT ON RESIDENTIAL PROPERTY PRICE INDICES MISSION (DECEMBER 8–12, 2019)

### Summary of Mission Outcomes and Priority Recommendations
- Purpose: Assist Bangladesh Bank (BB) in progressing on the compilation of a residential property price index (RPPI); compile an experimental RPPI; improve data for RPPI compilation.
- Importance: RPPI identified as critical for financial stability policy analysis and macroprudential measures (Second Phase of the G-20 Data Gaps Initiative and guidance on Financial Soundness Indicators). RPPI is also a macro-economic indicator of growth and key for understanding financial market conditions.
- Current status: Bangladesh lacks a reliable official measure of trends in residential property prices.
- Scope: BB aims to compile a quarterly RPPI covering new flats for Dhaka; mission recommended retaining secondary market and other-region data for future expansion.
- Methodology decision: Compile RPPI with the time dummy method with a rolling window of four quarters; index should be compiled with data from 2007 onwards.
- Tools and training: Mission adapted R scripts from the RPPI Practical Compilation Guide to BB data; recommended use of R for the compilation process.
- Data collection improvements: BB plans a new data collection system starting July 2020 to expand geographic coverage, dwelling types, and sample size.
- Dissemination timeline: BB may explore publication of the RPPI with new data from July 2021; internal compilation to continue in the meantime.

Priority Recommendations (Table 1)
- January/2020: Compile the RPPI with the time dummy with rolling window. — Responsible: BB
- July/2020: New data collection. — Responsible: BB

### A. Governance
- Team capacity:
  - BB has a team of nine staff from different departments compiling the index.
  - Three staff attended IMF courses on RPPI in 2017 and 2019.
  - At least six team members should have some knowledge in R.
- Software recommendation:
  - Use R because it performs econometric regressions and price index calculations in a single script/environment.
- Coverage and data retention:
  - Current focus: quarterly RPPI covering new flats in Dhaka due to insufficient data in other cities/types.
  - Recommendation: continue obtaining all data (including secondary market and rest of country) to store for future expansion.
- Base year alignment:
  - RPPI base year should match CPI and GDP base years to facilitate benchmarking.
  - BBS is aiming at rebasing both CPI and GDP to the fiscal year July 2015–June 2016 during 2020.

### B. Compilation with the Current Dataset
- Data source and volume:
  - BB receives quarterly data from Delta Brac Housing Finance Corporation, ltd (DBH) on house loans.
  - DBH dataset contains valuation price, size, and location within the city.
  - Data available since 1998 with over 33,000 observations in total.
  - Index work began in 2017.
- Location observations for period 1Q2017 (Table 2):
  - DHA01 206
  - DHA02 99
  - DHA03 98
  - DHA04 33
  - DHA05 17
  - DHA06 25
  - DHA07 27
  - DHA08 47
  - DHA09 17
  - DHA10 118
  - DHA11 60
  - DHA12 14
  - DHA13 4
  - DHA14 15
  - DHA15 11
  - DHA16 6
  - DHA17 28
  - DHA18 19
  - DHA20 4
  - DHA21 3
- Clustering recommendation:
  - Cluster the 21 locations to achieve at least around 50 observations per quarter per location cluster; clustering should be based on economic criteria (e.g., income level).
- Methodological training and guidance:
  - Mission provided training on hedonic methods, chain linking, and rebasing.
  - Hedonic methods are preferred to address quality changes in the mix of dwellings.
  - BB staff trained on rebasing and chain linking; R scripts adapted for BB data.
- Chain linking example (Table 3): preserved chained and unchained index values with bases 1Q2007=100, 2Q2007=100, 3Q2007=100, 4Q2007=100, 1Q2008=100 and chained 1Q2017=100 (values shown in source table).
- Recommended compilation start date:
  - Start compilation from 2007 using the time dummy method with a rolling window of four quarters (data prior to 2007 have few observations).
- Sample size note:
  - Each quarter has less than 1,000 observations; with a small sample the time dummy yields a more robust and less volatile indicator.
- Data cleaning guidance:
  - R script provided for basic data manipulations: treating outliers, missing values, and duplicates.
  - Descriptive statistics for 1Q2017 before and after treatment (Table 4):
    - Before: Minimum 555.581, Q1 3500, Median 4200, Mean 4760.586, Q3 5500, Maximum 55000
    - After: Minimum 826.4463, Q1 3500, Median 4200, Mean 4653.867, Q3 5500, Maximum 14761.9
- RPPI recompiled:
  - Mission recompiled the index from 2007 with base year July 2015/June 2016 using time dummy hedonic method; BB staff trained on script manipulation.

### C. Improving the Data Collection
- Questionnaire/template guidance:
  - BB prepared a survey template reviewed by the mission and presented to potential data providers.
  - Recommendation: keep the template as short as possible to avoid respondent burden.
  - Avoid open questions; use objective, measurable answers (yes/no or numbers); avoid semantic free text to ensure machine readability.
- Proposed data structure (Table 5) — Variable, Description, Type of data:
  - Date — Date of the Approval Date
  - Identification number of Dev. Authority Approval — ID number — Not to be used in regression
  - Address — Categorical
  - Monthly income — Average monthly income — Numerical
  - City — Categorical
  - Thana — Categorical
  - District — Categorical
  - Post code — Categorical
  - Floor level — Categorical
  - Height of the building — Number of stories — Categorical
  - Face Direction — South/North/West/East — Categorical
  - Total size — Square feet — Numerical
  - Purchase price — Net of costs — Numerical
  - Valuation price — Net of costs — Numerical
  - Interest rate — Number — Numerical
  - No. Bedrooms — Number — Categorical
  - No. Bathrooms — Number — Categorical
  - No. lifts — Number — Categorical
  - Parking — Number — Categorical
  - Can two cars easily pass at the same time in the street in front of your house? — Y/N — Categorical
  - Pipeline gas connection — Y/N — Categorical
  - Park — Within 10 min walking — Categorical
  - Hospitals — Within 10 min walking — Categorical
  - Schools — Within 10 min walking — Categorical
  - Public transports — Within 10 min walking — Categorical
  - Brand of Developer — Drop down menu with the main developers — Categorical
  - Construction year — Date
  - Resale — Y/N
  - Government owned plot — Y/N
- Official request and timing:
  - BB should send an official data request to data providers; providers require an official request for internal procedures.
  - Request should be sent with the final data structure no later than March to allow system adaptation time.
- Transmission and storage:
  - Data to be transmitted and stored using the Enterprise Data Warehouse (EDW) from July 2020.
  - Data providers need time to adapt internal systems before EDW automated transmission begins.

### D. Dissemination
- Publication timing:
  - BB may explore beginning publication of the RPPI with new data from July 2021.
  - Regular compilation will continue for internal use before public dissemination.
- Guidance provided:
  - Mission provided recommendations for dissemination as exemplified in the RPPI Guide and supplied a draft technical note (Appendix I).

Priority actions / milestones and target completion dates (excerpt)
- The 21 locations are clustered. — January 2020
- The RPPI is compiled with the time dummy method with a rolling window of four quarters. — January 2020
- The provisional data structure will be sent to the data providers. — January 2020
- An official data request is sent by the BB Governor to the data providers. — March 2020
- New data collection system is implemented. — July 2020

### E. Officials Met During the Mission
- List of officials (selected):
  - Mr. S. M. Moniruzzaman — Deputy Governor, BB
  - Mr. Ahmed Jamal — Deputy Governor, BB
  - Mr. Allah Malik Kazemi — Change Manager Adviser, BB
  - Mr. Asish Kumar Dasgupta — Executive Director (Research), BB
  - Mr. Md. Abdul Kayum — General Manager, Research Department, BB – RPPI team member
  - Ms. Mst. Nur Naher Begum — Deputy General Manager, Research Department, BB, RPPI team member
  - Mr. Mohammad Shahriar Siddiqui — Deputy General Manager, Banking Regulation and Policy Department, BB – RPPI team member
  - (Additional BB staff, DBH representatives, and BBS staff listed in source.)

*Prepared by Vanda Guerreiro; Technical Assistance Report completed in December 2019.*

### Section 2

### Appendix I. RPPI Technical Note

### Background
- The residential property price index (RPPI) measures the price evolution for residential new apartments in Dhaka.
- Reliable property price indexes are essential for the Bangladesh Bank (BB) to assess developments and risks in the real estate market and to understand the linkages between residential real estate markets and financial soundness.

### Coverage
- Coverage is limited to Dhaka city.
- Covers the primary (new dwellings) market for apartments.

### Data Sources
- BB collects data from Delta Brac Housing Finance Corporation, ltd (DBH) on loans for new apartments.

### Periodicity
- The RPPI is disseminated on a quarterly basis.

### Base Period
- The RPPI is an annually chain-linked price index that uses the year 2015/106 as reference year.
- The average of the quarterly indexes for 2015/16 equals 100.

### Dissemination
- The RPPI is released on a quarterly basis, 45 days after the end of the reference period.

### Methodology - Conceptually
- Key characteristics of the dwellings are required to assure a constant quality index.
- Price comparison requires comparable dwellings or the same dwelling in consecutive periods; this is generally not the case for residential properties where the same residential property is only sold every couple of decades.
- Given infrequent sale and heterogeneity of residential properties, quality adjustment techniques are required to derive measures of pure price change.
- High quality RPPI relies heavily on detailed characteristics about each property because a wide range of characteristics can influence the price of a dwelling.

### Methodology – Time Dummy Hedonic Method
- The time dummy method measures the effect of “time” on the price (푝).
- Prices of all dwellings (푛) for several periods (푡) are pooled in the same regression for every stratum, on their characteristics (푧푛푘푡) and on dummy variables for the periods (퐷푛푡).
- Main advantage: simplicity, since the index follows directly from the estimated time dummy parameters.
- It is generally appropriate:
  - with few transactions
  - or aiming at compiling a monthly index
  - or more detailed stratification
- Log-linear specification for each stratum (as presented):
  - ln푝푛푡 = 훽0 + ∑훿푡퐷푛푡 + ∑훽푘푧푛푘푡 + 휀푛푡 (퐾 푘=1, 푇 푡=1)
- The index for each stratum for a period 푡 is derived by exponentiation of the time dummy parameters 훿̂푡:
  - 퐼푡 = exp(훿̂푡) ∗ 100
- For the reference period a dummy variable is not included; the price index for the price reference period is set to equal 100.
- When a new period is added and the model re-estimated, the indices from previous periods will most likely change since the estimated parameters 훿̂푡 of older periods will differ from their previous estimate. Therefore, a rolling window approach is needed.

### Chaining
- To avoid revisions to the indices a rolling window approach is used.
- Normally for quarterly indices the “shadow” prices of the characteristics (훽̂) are kept fixed for at least a year.
- Data from 12 months are used together.
- Every quarter a regression is estimated with data from the current quarter and the previous three quarters.
- The indices from the “new window” and the “previous window” are chained by using the last overlap period between the two windows.
- Example of chaining the indexes from rolling window hedonic regression (quarterly index):
  - Table shows chaining across windows with entries:
    - 2018Q1 100.0   100.0
    - 2018Q2 101.2 100.0  101.2
    - 2018Q3 101.1 101.3 100.0 101.1
    - 2018Q4 100.9 101.2 100.8 100.9
    - 2019Q1  101.0 100.3 100.7=100.9/101.2x101.0
    - 2019Q2   101.4 101.8=100.7/100.3x101.4

*Source: 1bgdea2020002 - Section 2*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2020/english/1bgdea2020002.pdf_