## cr18288 - 1. The purpose of the mission was to assist the National Statistics Office (Geostat) in

## Source details

**Canonical URL:** [cr18288 - 1. The purpose of the mission was to assist the National Statistics Office (Geostat) in](https://www.imf.org/-/media/files/publications/cr/2018/cr18288.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2018/cr18288.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2018/cr18288.pdf.json)

---

### Mission purpose and scope
- Mission visited Tbilisi during April 25 to May 4, 2018.
- Purpose: assist the National Statistics Office (Geostat) in developing a residential property price index (RPPI) and expanding coverage of the producer price index (PPI).
- This was the first price statistics mission to Georgia under the three-year Project to Improve National Accounts and Price Statistics in Eastern and Southeastern Europe, funded by the Government of The Netherlands.

### Current PPI coverage and assessment
- Geostat compiles a monthly PPI for industrial products covering:
  - mining and quarrying;
  - manufactured products; and
  - electricity energy, gas, steam, and hot water.
- PPI includes a separate domestic PPI and an export price index.
- Services coverage in PPI is currently limited to freight transport and should be expanded.
- Improvements since the 2011 ROSC:
  - improved index weighting by using the value of output instead of turnover;
  - created electronic questionnaires on the Geostat main website to facilitate electronic data reporting;
  - added metadata for PPI to the Geostat website.
- Coverage statistics and context:
  - Current PPI covers approximately 25 percent of output.
  - Services comprise about 50 percent of output.
  - Largest contributors to services output: wholesale and retail trade, and construction.
- Mission recommendation for PPI expansion (priority services to add):
  - Warehousing and storage services;
  - Telecommunications services;
  - Travel agency and tour operator services.
- Rationale for recommended services:
  - Warehousing and storage builds on Freight transport services measurement.
  - Telecommunications improves deflation of nominal output, especially sales of telecommunications services to enterprises.
- Data collection cost note:
  - Wholesale and retail trade likely very large (approximately 50 percent of active business units engage in this activity).
- Additional methodological recommendation:
  - Develop activity-specific questionnaires that specify product characteristics; e.g., price and quality determining features.

### RPPI rationale and scope
- Purpose and policy use:
  - RPPI will facilitate assessment of developments and risks in property markets and be useful for monetary policy to understand linkages between property asset prices and financial assets.
- Existing indices:
  - National Bank of Georgia compiles a rudimentary index covering residential and commercial property prices in two districts of Tbilisi (one expensive district and one modest district); index is limited and not disseminated.
- Proposed initial RPPI scope:
  - restricted to the capital city (Tbilisi);
  - cover all transactions in new apartments and houses only (existing dwellings excluded initially due to complexity).
- Institutional guidance:
  - RPPI should remain a stand-alone index and should not be included in the CPI (differences in periodicity and initial quality concerns).

### Production schedule, staffing, and dissemination
- Compilation frequency:
  - Geostat should be able to compile the RPPI on a quarterly basis.
- Target dissemination:
  - Target: disseminate the first index for the first quarter of 2021, in mid-May 2021.
- Staffing and resource allocation:
  - RPPI to be developed by the same staff compiling the CPI; production schedule for RPPI can be arranged around CPI schedule to accommodate available staff resources.
  - Based on current CPI production schedule and proposed RPPI development plan, additional staff would not be required.
  - Current CPI staffing: five persons assigned to the CPI.
- CPI release timing and staff allocation:
  - CPI is released on the third day of the month;
  - resources fully allocated to CPI from the twenty-first day of the previous month;
  - price collection for CPI takes place during the tenth and twentieth day;
  - resources can be assigned to the RPPI from the fourth to the ninth of the month and partially allocated up to the twentieth.

### Data sources evaluated for RPPI
- National Agency of Public Registry (NAPR) (most suitable):
  - Registration of all transactions in dwellings with NAPR is compulsory.
  - NAPR may collect transaction value, transactors, dwelling specifications, and location.
  - Mission could not contact NAPR during visit; Geostat should consult NAPR as soon as possible.
  - Possibility of underreporting must be further investigated together with tax scheme.
  - When dwelling bought directly from developer the registered price is the final price as no transaction tax are due; purchases from owner or agency may be underreported.
- Real estate websites (alternative sources):
  - Websites provide asking price and dwelling characteristics.
  - Web scraping and automated data transmission are areas to explore.
  - If websites used, Geostat should inform site owners and request data transmission prior to web scraping.
  - Rules for duplicates when using websites:
    - A dwelling can be advertised on the same site or on different sites by many agencies — duplicates should be excluded.
    - If a dwelling appears in consecutive months with the same price, treat as duplicate entry for index entry purposes.
- National Bank of Georgia index:
  - Websites-based, covers only two districts in Tbilisi with variables confined to district and price; useful for backward consistency analysis and experimentation.

### Data quality assessment and requirements
- Data quality dimensions to assess as soon as data received: reliability, timeliness, completeness.
- Coverage confirmation requirements:
  - new dwellings;
  - apartments and houses;
  - purchases financed with loans and cash.
- Variable verification:
  - Confirm meaning of variables (e.g., size variables).
- Processing requirements:
  - Automatic identification of outliers and missing values every month.
  - Ideal data receipt periodicity: monthly.
- Recommended assessment steps:
  - Assess reliability by cross-check between data sources.
  - Confirm coverage as expected.
  - Select only dwellings located in Tbilisi (initial scope).
  - Confirm meaning of variables/characteristics.
  - Check data for outliers, missing values.
  - Ensure timely receipt, ideally monthly.

### Real estate market characteristics and treatment issues
- Buying process timing and implications:
  - Begin search: Mostly on websites.
  - Agreement on price: can take up to 16 weeks; price is bargained and final price may differ from offered price; a buyer can take up to four months to choose a dwelling.
  - Mortgage approval: depends on day of application; banks have one specific week for approvals; approval faster if dwelling is finished versus in construction.
  - Register: contract register in the notary (10 days) or to NAPR or both; mortgage and contract can be made together when the bank has agreement with developer; 10 days and 1 to 4 weeks timing noted.
- Classification of new dwellings by completion level on websites:
  - black – without interior walls, electricity installation, and communications;
  - white – with interior walls but without windows, communications;
  - green – requiring minor finishing such as painting or flooring.
- Price coverage and quality adjustment:
  - Price to be covered in index: full price of a completed turn-key dwelling.
  - Quality adjustment should account for level of completion and whether appliances (such as kitchen appliances) are included.
- Financing and taxes:
  - New dwellings financed through the developer may include a developer’s financing commission in the price; financing cost should not be included in the index.
  - Value added tax should be included in the price;
  - Income taxes and transaction taxes should be excluded.
- Recommendations (market treatment):
  - Collect additional information regarding the taxation scheme for real estate.
  - Consider a quality adjustment according to the level of completion and whether appliances are included (such as kitchen appliances).
  - Identify possible developer’s financing commission included in the price of new dwellings.

### Methodology options and experiments
- Initial coverage and staging:
  - Initial coverage: capital city only, new apartments and houses.
  - Existing dwellings to be covered later when RPPI methodology stabilized and staff experienced.
- Stratification experimentation (examples to test):
  - Example A: stratify by dwelling type (apartments/houses) and completion level (Black/White/Green).
  - Example B: stratify by dwelling type and district level (High level districts; Medium/Low level districts).
- Two main steps for stratification testing:
  - Check number of observations per strata over time and build basic descriptive statistics (median, mean, standard deviation) by strata for different stratifications.
  - Compile index with the ratios of mean, median and geometric mean and compare results.
- Hedonic methods guidance:
  - Compile index series with different approaches: time dummy; characteristics method; imputations approach; and with different model specifications and forms: linear; log linear.
  - Main steps for hedonic exercise:
    - run the OLS regression;
    - test significance of variables;
    - test multicollinearity (example: no bedrooms and size) and model specification.
  - The lower level of stratification can be included as dummy variables in hedonic regressions.
  - If hedonics not feasible, compile index series with ratios of mean, median and geometric mean and compare results.
- Methodology recommendations:
  - Define the variables.
  - Experiment with different options for stratification.
  - Compile index series with the different approaches: time dummy; characteristics method; imputations approach. Assess results from different model specifications.
  - Run the OLS for different regression forms: linear; log linear.
  - If the hedonics method is not feasible, compile an index series with the ratios of mean, median and geometric mean and compare the results.

### Priority recommendations and target dates (Table 1)
- July 2018
  - Identify potential data sources for the RPPI and receive first data sets.
  - Responsible Institutions: Geostat
- September 2019
  - Compile an experimental RPPI.
  - Responsible Institutions: Geostat
- January 2020
  - Expand PPI coverage to include Warehousing and storage services; Telecommunications; Travel agency and tour operator services.
  - Responsible Institutions: Geostat
- January 2020
  - Develop activity-specific questionnaires that specify product characteristics; e.g., price and quality determining features.
  - Responsible Institutions: Geostat

### Actionable short-term recommendations
- Explore fitness for purpose and quality of web scraping data from the two main websites.
- Obtain first set of data from NAPR and assess its quality.
- Select final data sources and write formal agreements.
- Assess reliability by performing cross checks between data sources.
- Define variables for RPPI compilation and set up automated checks for outliers and missing values.
- Conduct experiments on stratification and hedonic model specifications; compare results across methods.

### Geostat should not consider including the RPPI in the CPI in the short term
- Finding:
  - Geostat should not include the RPPI in the CPI in the short term because of differences in periodicity and lower expected quality of a new index in its first years; for inclusion the net approach is required (only new dwellings covered).
- Recommendation:
  - Develop the RPPI methodology during an experimental period and collect back data.
- Data targets and operational guidance:
  - Aim to collect three years of back data.
  - Start monthly automated transmission as soon as possible; ideally transmission on the third day of each month to align with CPI production routines.
  - Staff in charge of RPPI production should have direct contact with staff responsible for data transmission from project start.
  - Compile an experimental index covering new dwellings for the previous three-year period.
  - Release first index mid-May 2021 for the reference period ending the first quarter of 2021.
- Timeline items from action plan (selected):
  - Assessment of potential data sources:
    - H: Obtain data from NAPR and assess its quality. Target Completion Date: September 2018.
    - H: Assess reliability by crosscheck between data sources. Target Completion Date: September 2018.
    - H: Select final data sources and write formal agreements. Target Completion Date: February 2019.
  - Quality adjustments methods:
    - H: Design clear definition of variables. Target Completion Date: November 2018.
    - H: Use expertise from real estate agents and developers. Target Completion Date: November 2018.
    - H: Experiment with different options for stratification. Target Completion Date: September 2019.
    - H: Compile index series with the different hedonic methods. Target Completion Date: November 2019.
    - H: Assess results from different model specifications. Target Completion Date: January 2020.
    - H: Run the OLS for different regression forms: liner; log linear. Target Completion Date: January 2020.
    - L: If hedonics not feasible compile an index series with the ratios of mean. Target Completion Date: May 2020.
  - Main methodology options:
    - H: Compile experimental index covering new dwellings for the last three years. Target Completion Date: September 2020.
    - H: Release first index mid-May 2021. Target Completion Date: May 2021.

### PPI scope, coverage, and classification
- Current scope:
  - Geostat publishes monthly output PPIs for total production, production for domestic sale, and exports at basic prices (excludes taxes, transport fees, trade margins) collected at ex-producer (factory gate).
  - PPI covers products from CPA 1996 sections C, D, and E (mining and quarrying; manufactured products; electricity energy, gas, steam, and hot water).
  - A price index for material inputs to construction is produced; no PPI for construction activity output.
- Coverage statistics:
  - The PPI covers approximately 25 percent of in-scope output.
  - Services activities excluding construction comprised approximately 50 percent of in-scope output as of 2017.
  - Adding recommended services activities (Warehousing and storage; Telecommunications; Travel agency and tour operator services) would increase coverage of in-scope output by five percent.
- Classification and transition:
  - Geostat uses CPA 1996 for the PPI and NACE Rev. 2 for national accounts and business statistics; national accounts are transitioning to NACE Rev. 2.
  - Recommendation: Update classification basis from CPA 1996 to CPA version 2.1 to align with national accounts and business statistics (note: may result in discontinuities and partial coverage of new products).

### Expanding PPI services coverage — phased implementation
- Phased pilot schedule (recommendation):
  - Phase I — January 2019: pilot survey for warehousing and storage.
  - Phase II — July 2019: pilot survey for telecommunications.
  - Phase III — January 2020: pilot survey for travel agency and tour operators.
- Recommended CPA 2.1 codes to add:
  - 52.10 (Warehousing and storage services);
  - 69 (Telecommunications services);
  - 79.1 (Travel agency and tour operator services).
- Rationale for sequencing:
  - Build on freight transport services index experience for warehousing.
  - Telecommunications index to improve deflation of value added for enterprise sales not covered in CPI.
  - Travel/tour services to improve measurement of tourism sector.

### Sampling, weighting, and index calculation
- Sampling and frame:
  - PPI is a sub-sample of the Business Statistics program drawn from the business register; statistical unit is a kind-of-activity unit (KAU).
  - Geostat is currently unable to identify the principal activity of 26.4 percent of active register units.
  - Product samples aim to achieve coverage of 80 percent of production via cutoff, systematic, and/or judgmental selection.
- Weights and index formula:
  - Aggregate weights at CPA 4-digit level are taken from national accounts (total output including unregistered units and households); allocated to detailed levels based on production value of registered units.
  - Methodology recently improved to use output rather than turnover for weighting.
  - Indices are calculated using a modified Laspeyres formula, with weights and samples updated annually based on Business Statistics survey results using t-2 period data.
- Recommended implementation milestones (selected):
  - Identify sample frame and measure of size determining enterprise selection (Phase completions: Phase 1 September 2018; Phase 2 March 2019; Phase 3 September 2019).
  - Select sample of enterprises (Phase 1 October 2018; Phase 2 April 2019; Phase 3 October 2019).
  - Calculate weights for upper-level indices based on net output of associated NACE activity group (Phase 1 completion – December 2018; Phase 2 June 2019; Phase 3 December 2019).
  - Calculate weights for elementary aggregates based on turnover data collected from enterprises (Phase 1 January 2019; Phase 2 July 2019; Phase 3 January 2020). (Actual: January 2020 listed for this item.)

### Price collection, questionnaires, and data editing
- Field operations and resourcing:
  - Field data collectors support PPI and related price index programs; assigned approximately 150 KAU each.
  - Geostat employed 26 field data collectors in 2018 (not all supported PPI and related programs).
- Questionnaire and reporting issues:
  - A standard questionnaire is used for all products; activity-specific questionnaires would better facilitate specification for constant quality pricing.
  - Respondents are instructed to provide the average price received for all sales in the most recently completed month, but the questionnaire does not instruct how the average price was calculated.
  - The questionnaire instructs respondents to describe reasons for price changes; base price is displayed but respondent cannot edit it.
- Data editing and tolerances:
  - Field data collectors approve all price submissions monthly and notify non-responding enterprises of mandatory obligations.
  - Efficiency suggestion: flag cases for field data collector review only if magnitude of price change is over or below established tolerances; tolerances may need to be set independently for each product group.
- Recommendations:
  - Assign data collectors to administer activity-specific questionnaires for larger respondents and complex areas (services).
  - Develop activity-specific questionnaires that specify product characteristics; price and quality determining features; and transaction terms for constant quality pricing.
  - Remove the base price from the data collection instrument.
  - Add fields for respondents to indicate total turnover collected and units transacted when calculating product average prices.
  - Consider using email to inform respondents when to record monthly price updates (establish secure email protocols).
  - Introduce referral flags based on price change tolerances to limit manual review workload; set tolerances to flag unexpected changes.

### Dissemination and stakeholder relations
- Current dissemination practice:
  - The PPI is published monthly, 20 days after the reference month, meeting SDDS requirements.
  - Time series at the aggregate and 19 sub-group levels are compiled from individual 12-month annual PPI series; longer time-series are derived by chain linking annual index series.
  - The PPI is final upon first release; revisions occur only in rare cases of significant errors.
  - The Geostat website is relatively informative; consistent long index time series are difficult to locate since not disseminated together with monthly data.
- Communication and user engagement:
  - Clear methodology descriptions are provided in Georgian and English; website includes central telephone and email contacts.
  - Communication with key PPI stakeholders occurs on an as-needed basis; recommendation to establish scheduled recurring interactions.
- Dissemination recommendations:
  - Add one-month and twelve-month percentage changes for each data series in PPI releases and interactive data retrieval tool.
  - Establish a users’ advisory group for price statistics meeting regularly and including academic institutions, the business community, national accounts, and the Central Bank.

*International Monetary Fund — cr18288*

### 1.      The purpose of the mission was to assist the National Statistics Office (Geostat) in

### cr18288 - 1.      The purpose of the mission was to assist the National Statistics Office (Geostat) in

### Mission purpose and scope
- Mission visited Tbilisi during April 25 to May 4, 2018.
- Purpose: assist the National Statistics Office (Geostat) in developing a residential property price index (RPPI) and expanding coverage of the producer price index (PPI).
- This was the first price statistics mission to Georgia under the three-year Project to Improve National Accounts and Price Statistics in Eastern and Southeastern Europe, funded by the Government of The Netherlands.

### Current PPI coverage and assessment
- Geostat compiles a monthly PPI for industrial products covering:
  - mining and quarrying;
  - manufactured products; and
  - electricity energy, gas, steam, and hot water.
- PPI includes a separate domestic PPI and an export price index.
- Services coverage in PPI is currently limited to freight transport and should be expanded.
- Improvements made since the 2011 ROSC:
  - improved index weighting by using the value of output instead of turnover;
  - created electronic questionnaires on the Geostat main website to facilitate electronic data reporting;
  - added metadata for PPI to the Geostat website.
- Current PPI covers approximately 25 percent of output; services comprise about 50 percent of output.
- Largest contributors to services output: wholesale and retail trade, and construction.
- Mission recommendation for PPI expansion (priority services to add):
  - Warehousing and storage services;
  - Telecommunications services;
  - Travel agency and tour operator services.
- Rationale for recommended services:
  - Warehousing and storage builds on Freight transport services measurement.
  - Telecommunications improves deflation of nominal output, especially sales of telecommunications services to enterprises.
- Data collection cost for wholesale and retail trade likely very large (approximately 50 percent of active business units engage in this activity).
- Additional PPI methodological recommendation: develop activity-specific questionnaires that specify product characteristics; e.g., price and quality determining features.

### RPPI rationale and scope
- RPPI will facilitate assessment of developments and risks in property markets and be useful for monetary policy to understand linkages between property asset prices and financial assets.
- National Bank of Georgia compiles a rudimentary index covering residential and commercial property prices in two districts of Tbilisi (one expensive district and one modest district); index is limited and not disseminated.
- Mission proposed start scope for RPPI:
  - restricted to the capital city (Tbilisi);
  - cover all transactions in new apartments and houses only (existing dwellings excluded initially due to complexity).
- RPPI should remain a stand-alone index and should not be included in the CPI (differences in periodicity and initial quality concerns).

### Production schedule, staffing, and dissemination
- Geostat should be able to compile the RPPI on a quarterly basis.
- Target: disseminate the first index for the first quarter of 2021, in mid-May 2021.
- RPPI to be developed by the same staff compiling the CPI; production schedule for RPPI can be arranged around CPI schedule to accommodate available staff resources.
- Based on current CPI production schedule and proposed RPPI development plan, additional staff would not be required.
- Current CPI staffing: five persons assigned to the CPI.
- CPI release timing and staff allocation:
  - CPI is released on the third day of the month;
  - resources fully allocated to CPI from the twenty-first day of the previous month;
  - price collection for CPI takes place during the tenth and twentieth day;
  - resources can be assigned to the RPPI from the fourth to the ninth of the month and partially allocated up to the twentieth.

### Data sources evaluated for RPPI
- Most suitable data source may be the National Agency of Public Registry of Ministry of Justice (NAPR).
  - Registration of all transactions in dwellings with NAPR is compulsory.
  - NAPR may collect transaction value, transactors, dwelling specifications, and location.
  - Mission could not contact NAPR during visit; Geostat should consult NAPR as soon as possible.
  - Possibility of underreporting must be further investigated together with tax scheme.
  - When dwelling bought directly from developer the registered price is the final price as no transaction tax are due; purchases from owner or agency may be underreported.
- Two main websites for real estate transactions are alternative data sources.
  - Websites provide asking price and dwelling characteristics.
  - Web scraping and automated data transmission are areas to explore.
  - If websites used, Geostat should inform site owners and request data transmission prior to web scraping.
  - Duplicates handling rules when using websites:
    - A dwelling can be advertised on the same site or on different sites by many agencies — duplicates should be excluded.
    - If a dwelling appears in consecutive months with the same price, treat as duplicate entry for index entry purposes.
- National Bank of Georgia index (websites-based) covers only two districts in Tbilisi with variables confined to district and price; useful for backward consistency analysis and experimentation.

### Data quality assessment and requirements
- Data quality dimensions to assess as soon as data received: reliability, timeliness, completeness.
- Confirm coverage in terms of:
  - new dwellings;
  - apartments and houses;
  - purchases financed with loans and cash.
- Confirm meaning of variables (e.g., size variables).
- Automatic identification of outliers and missing values every month.
- Ideal data receipt periodicity: monthly.
- Recommended steps for assessment:
  - Assess reliability by cross-check between data sources.
  - Confirm coverage as expected.
  - Select only dwellings located in Tbilisi (initial scope).
  - Confirm meaning of variables/characteristics.
  - Check data for outliers, missing values.
  - Ensure timely receipt, ideally monthly.

### Real estate market characteristics and treatment issues
- Buying process timeline highlights:
  - Begin search: Mostly on websites.
  - Agreement on price: can take up to 16 weeks; price is bargained and final price may differ from offered price; a buyer can take up to four months to choose a dwelling.
  - Mortgage approval: depends on day of application; banks have one specific week for approvals; approval faster if dwelling is finished versus in construction.
  - Register: contract register in the notary (10 days) or to NAPR or both; mortgage and contract can be made together when the bank has agreement with developer; 10 days and 1 to 4 weeks timing noted.
- New dwellings classification on websites by level of completion:
  - black – without interior walls, electricity installation, and communications;
  - white – with interior walls but without windows, communications;
  - green – requiring minor finishing such as painting or flooring.
- Price to be covered in index: full price of a completed turn-key dwelling.
  - Quality adjustment should account for level of completion and whether appliances (such as kitchen appliances) are included.
- Financing through developer:
  - New dwellings financed through the developer may include a developer’s financing commission in the price; financing cost should not be included in the index.
- Tax treatment:
  - Value added tax should be included in the price;
  - Income taxes and transaction taxes should be excluded.
- Recommendations (market characteristics):
  - Collect additional information regarding the taxation scheme for real estate.
  - Consider a quality adjustment according to the level of completion and whether appliances are included (such as kitchen appliances).
  - Identify possible developer’s financing commission included in the price of new dwellings.

### Methodology options and experiments
- Initial coverage: capital city only, new apartments and houses.
- Existing dwellings to be covered later when RPPI methodology stabilized and staff experienced.
- Stratification experimentation recommended; example stratifications to be tested:
  - Example A:
    - RPPI
      - 1.1. New dwellings
        - 1.1.1. Apartments
          - 1.1.1.1. Black
          - 1.1.1.2. White
          - 1.1.1.3. Green
        - 1.1.2. Houses
          - 1.1.1.1. Black
          - 1.1.1.2. White
          - 1.1.1.3. Green
  - Example B:
    - 1. RPPI
      - 1.1. New dwellings
        - 1.1.1. Apartments
          - 1.1.2.1. High level districts
          - 1.1.2.2. Medium/Low level districts
        - 1.1.3. Houses
          - 1.1.1.4. High level districts
          - 1.1.1.5. Medium/Low level districts
- Two main steps recommended for stratification testing:
  - Check number of observations per strata over time and build basic descriptive statistics (median, mean, standard deviation) by strata for different stratifications.
  - Compile index with the ratios of mean, median and geometric mean and compare results.
- Hedonic methods guidance:
  - Compile index series with different approaches: time dummy; characteristics method; imputations approach; and with different model specifications and forms: linear; log linear.
  - Main steps for hedonic exercise:
    - run the OLS regression;
    - test significance of variables;
    - test multicollinearity (example: no bedrooms and size) and model specification.
  - The lower level of stratification can be included as dummy variables in hedonic regressions.
  - If hedonics not feasible, compile index series with ratios of mean, median and geometric mean and compare results.
- Recommendations (methodology):
  - Define the variables.
  - Experiment with different options for stratification.
  - Compile index series with the different approaches: time dummy; characteristics method; imputations approach. Assess results from different model specifications.
  - Run the OLS for different regression forms: linear; log linear.
  - If the hedonics method is not feasible, compile an index series with the ratios of mean, median and geometric mean and compare the results.

### Priority recommendations and target dates (Table 1)
- July 2018
  - Identify potential data sources for the RPPI and receive first data sets.
  - Responsible Institutions: Geostat
- September 2019
  - Compile an experimental RPPI.
  - Responsible Institutions: Geostat
- January 2020
  - Expand PPI coverage to include Warehousing and storage services; Telecommunications; Travel agency and tour operator services.
  - Responsible Institutions: Geostat
- January 2020
  - Develop activity-specific questionnaires that specify product characteristics; e.g., price and quality determining features.
  - Responsible Institutions: Geostat

### Actionable short-term recommendations
- Explore fitness for purpose and quality of web scraping data from the two main websites.
- Obtain first set of data from NAPR and assess its quality.
- Select final data sources and write formal agreements.
- Assess reliability by performing cross checks between data sources.
- Define variables for RPPI compilation and set up automated checks for outliers and missing values.
- Conduct experiments on stratification and hedonic model specifications; compare results across methods.

*International Monetary Fund — cr18288*

### 33.      Geostat should not consider including the RPPI in the CPI in the short term.  In

### Geostat should not consider including the RPPI in the CPI in the short term

### RPPI inclusion decision and experimental program
- Finding: Geostat should not include the RPPI in the CPI in the short term because of differences in periodicity and lower expected quality of a new index in its first years; for inclusion the net approach is required (only new dwellings covered).
- Recommendation: Develop the RPPI methodology during an experimental period and collect back data.
- Data targets and operational guidance:
  - Aim to collect three years of back data.
  - Start monthly automated transmission as soon as possible; ideally transmission on the third day of each month to align with CPI production routines.
  - Staff in charge of RPPI production should have direct contact with staff responsible for data transmission from project start.
  - Compile an experimental index covering new dwellings for the previous three-year period.
  - Release first index mid-May 2021 for the reference period ending the first quarter of 2021.
- Timeline items from action plan:
  - Outcome: Assessment of potential data sources
    - H: Obtain data from NAPR and assess its quality. Target Completion Date: September 2018.
    - H: Assess reliability by crosscheck between data sources. Target Completion Date: September 2018.
    - H: Select final data sources and write formal agreements. Target Completion Date: February 2019.
  - Outcome: Quality adjustments methods
    - H: Design clear definition of variables. Target Completion Date: November 2018.
    - H: Use expertise from real estate agents and developers. Target Completion Date: November 2018.
    - H: Experiment with different options for stratification. Target Completion Date: September 2019.
    - H: Compile index series with the different hedonic methods. Target Completion Date: November 2019.
    - H: Assess results from different model specifications. Target Completion Date: January 2020.
    - H: Run the OLS for different regression forms: liner; log linear. Target Completion Date: January 2020.
    - L: If hedonics not feasible compile an index series with the ratios of mean. Target Completion Date: May 2020.
  - Outcome: Main methodology options
    - H: Compile experimental index covering new dwellings for the last three years. Target Completion Date: September 2020.
    - H: Release first index mid-May 2021. Target Completion Date: May 2021.

### PPI scope, coverage, and classification
- Current scope:
  - Geostat publishes monthly output PPIs for total production, production for domestic sale, and exports at basic prices (excludes taxes, transport fees, trade margins) collected at ex-producer (factory gate).
  - PPI covers products from CPA 1996 sections C, D, and E (mining and quarrying; manufactured products; electricity energy, gas, steam, and hot water).
  - A price index for material inputs to construction is produced; no PPI for construction activity output.
- Coverage statistics:
  - The PPI covers approximately 25 percent of in-scope output.
  - Services activities excluding construction comprised approximately 50 percent of in-scope output as of 2017.
  - Adding recommended services activities (Warehousing and storage; Telecommunications; Travel agency and tour operator services) would increase coverage of in-scope output by five percent.
- Classification and transition:
  - Geostat uses CPA 1996 for the PPI and NACE Rev. 2 for national accounts and business statistics; national accounts are transitioning to NACE Rev. 2.
  - Recommendation: Update classification basis from CPA 1996 to CPA version 2.1 to align with national accounts and business statistics (note: may result in discontinuities and partial coverage of new products).

### Expanding PPI services coverage — phased implementation
- Phased pilot schedule (recommendation):
  - Phase I — January 2019: pilot survey for warehousing and storage.
  - Phase II — July 2019: pilot survey for telecommunications.
  - Phase III — January 2020: pilot survey for travel agency and tour operators.
- Recommendations to expand coverage and rationale:
  - Add CPA 2.1 codes 52.10 (Warehousing and storage services); 69 (Telecommunications services); 79.1 (Travel agency and tour operator services).
  - Build on freight transport services index experience for warehousing; telecommunications index to improve deflation of value added for enterprise sales not covered in CPI; travel/tour services to improve measurement of tourism sector.

### Sampling, weighting, and index calculation
- Sampling and frame:
  - PPI is a sub-sample of the Business Statistics program drawn from the business register; statistical unit is a kind-of-activity unit (KAU).
  - Geostat is currently unable to identify the principal activity of 26.4 percent of active register units.
  - Product samples aim to achieve coverage of 80 percent of production via cutoff, systematic, and/or judgmental selection.
- Weights and index formula:
  - Aggregate weights at CPA 4-digit level are taken from national accounts (total output including unregistered units and households); allocated to detailed levels based on production value of registered units.
  - Methodology recently improved to use output rather than turnover for weighting.
  - Indices are calculated using a modified Laspeyres formula, with weights and samples updated annually based on Business Statistics survey results using t-2 period data.
- Recommendations:
  - Identify sample frame and measure of size determining enterprise selection (Phase completions: Phase 1 September 2018; Phase 2 March 2019; Phase 3 September 2019).
  - Select sample of enterprises (Phase 1 October 2018; Phase 2 April 2019; Phase 3 October 2019).
  - Calculate weights for upper-level indices based on net output of associated NACE activity group (Phase 1 completion – December 2018; Phase 2 June 2019; Phase 3 December 2019).
  - Calculate weights for elementary aggregates based on turnover data collected from enterprises (Phase 1 January 2019; Phase 2 July 2019; Phase 3 January 2020). (Actual: January 2020 listed for this item.)

### Price collection, questionnaires, and data editing
- Field operations:
  - Field data collectors support PPI and related price index programs; assigned approximately 150 KAU each.
  - Geostat employed 26 field data collectors in 2018 (not all supported PPI and related programs).
  - Responding units are notified by official letter and instructed to use the website for data reporting; respondents provide specifications, prices, and turnover shares for up to four product varieties.
- Questionnaire and reporting issues:
  - A standard questionnaire is used for all products; activity-specific questionnaires would better facilitate specification for constant quality pricing.
  - Respondents are instructed to provide the average price received for all sales in the most recently completed month, but the questionnaire does not instruct how the average price was calculated.
  - The questionnaire instructs respondents to describe reasons for price changes; base price is displayed but respondent cannot edit it.
- Data editing and tolerances:
  - Field data collectors approve all price submissions monthly and notify non-responding enterprises of mandatory obligations.
  - Efficiency suggestion: flag cases for field data collector review only if magnitude of price change is over or below established tolerances; tolerances may need to be set independently for each product group.
- Recommendations:
  - Assign data collectors to administer activity-specific questionnaires for larger respondents and complex areas (services).
  - Develop activity-specific questionnaires that specify product characteristics; price and quality determining features; and transaction terms for constant quality pricing.
  - Remove the base price from the data collection instrument.
  - Add fields for respondents to indicate total turnover collected and units transacted when calculating product average prices.
  - Consider using email to inform respondents when to record monthly price updates (establish secure email protocols).
  - Introduce referral flags based on price change tolerances to limit manual review workload; set tolerances to flag unexpected changes.

### Dissemination and stakeholder relations
- Current dissemination:
  - The PPI is published monthly, 20 days after the reference month, meeting SDDS requirements.
  - Time series at the aggregate and 19 sub-group levels are compiled from individual 12-month annual PPI series; longer time-series are derived by chain linking annual index series.
  - The PPI is final upon first release; revisions occur only in rare cases of significant errors.
  - The Geostat website is relatively informative; consistent long index time series are difficult to locate since not disseminated together with monthly data.
- Communication and user engagement:
  - Clear methodology descriptions are provided in Georgian and English; website includes central telephone and email contacts.
  - Communication with key PPI stakeholders occurs on an as-needed basis; recommendation to establish scheduled recurring interactions.
- Recommendations:
  - Add one-month and twelve-month percentage changes for each data series in PPI releases and interactive data retrieval tool.
  - Establish a users’ advisory group for price statistics meeting regularly and including academic institutions, the business community, national accounts, and the Central Bank.

_Italic: IMF technical assistance report extract._

---


_Source: https://www.imf.org/-/media/files/publications/cr/2018/cr18288.pdf_