## 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/tarea2024064-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tar/2024/english/tarea2024064-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/tar/2024/english/tarea2024064-print-pdf.pdf.json)

---

### A. Introduction
- TA mission conducted from April 8–12, 2024 to assist the State Statistical Service of Ukraine (SSSU) with a methodological review of their House Price Index (HPI).
- Mission objectives completed:
  - (i) review listings data collected by the SSSU and the data preparation being applied,
  - (ii) assess the stratification and hedonic methods used for the HPI,
  - (iii) review weights and aggregation procedures used to compile the national index,
  - (iv) provide guidance on dissemination of the HPI,
  - (v) provide practical training to SSSU staff.
- Mission context and timeline:
  - Mission conducted under the Data for Decisions (D4D) Trust Fund; pre-mission meeting held on March 20, 2023.
  - SSSU published HPI in January 2024 for reference quarter Q4 2023 after approx. two years interruption.
  - SSSU planned to publish improved HPI for reference quarter Q1 2024 on May 20, 2024.
  - Main changes for Q1 2024 publication: (i) listings data as new source, (ii) use of offer prices instead of transaction prices, (iii) regression-based (hedonic) quality adjustment replacing stratified means.
- Mission provided practical training on hedonic methods and will continue supporting SSSU as the workplan is implemented.

### B. Data Sources — findings and recommended short-term actions
Findings:
- New residential properties (primary market):
  - Data manually collected by SSSU regional offices from real estate agents and property developers.
  - About 900 observations on new residential properties every quarter.
  - Sampling strategy based on regional populations.
  - Recommendation to expand collection within regions with low number of observations.
- Existing residential properties (secondary market):
  - SSSU uses online listings from OLX (www.olx.ua).
  - In 2023, SSSU received about 20,000-25,000 listings each quarter.
  - Web-scraping began in 2022; an MOU in Q1 2023 allowed direct receipt of OLX data.
- Data quality and preparation:
  - SSSU supplied four quarters of data for 2023; R was used for data analysis and visualization (histograms, scatter plots, descriptive statistics).
  - OLX data received by SSSU was already filtered (e.g., no properties < 30 square meters or with > three rooms).
  - Duplicates found in 2023: 75 for existing properties and 17 for new properties.
  - Offer prices can be quoted in Ukrainian hryvnia, US dollars or Euros; all offers should be converted to Ukrainian hryvnia using appropriate exchange rates. The RPPI should be compiled using Ukrainian hryvnia.
- Property characteristics available in listings data:
  - (i) offer prices,
  - (ii) location details i.e., region and city,
  - (iii) property attributes i.e., year of construction, total area in sqm, kitchen area in sqm, material type, number of rooms, floor number (existing properties only), total floors in the building (existing properties only).
  - Some variables have incomplete information (e.g., material type and year of construction) or may not significantly impact price; not all will be used in compilation.
- Geographic granularity:
  - Current collection: region and city only.
  - Ukraine has 24 regions; significant intra-regional price variation suggests collecting more granular location details (administrative districts or postcodes) would improve quality adjustment.
- Coverage:
  - Current listings data covers apartments only.
  - Recommendation: extend data collection to include houses via OLX and regional SSSU offices.

Recommended Actions (preserve phrasing):
- Expand the data collection for new properties within regions with low number of observations.
- Request all available data from OLX including the errors and outliers.
- Duplicates should be excluded for existing properties and included for new properties.
- Investigate whether it is possible to collect more detailed information on the location of the properties for sale.
- Extend the data collection to include houses i.e., via the OLX platform and via the regional SSSU offices.

### C. Compilation Methods — Q1 2024 Publication in May 2024
Findings and short-term recommendations:
- SSSU plan to implement hedonic methods for Q1 2024 publication on May 20, 2024. The hedonic method planned is a non-standard form of the characteristics approach.
- The new method is an improvement over stratified means by applying more refined quality adjustment, but more optimal hedonic methods exist.
- Due to the May publication deadline, mission recommended only short-term implementable changes.
- Priority recommendations with target date May 2024:
  - Use the available regional weights and apply a weighted mean rather than an unweighted geometric mean to the regional estimated prices. — Responsible: SSSU
  - Exclude regions with zero or very low number of observations. — Responsible: SSSU
  - Increase data collection for one specific region i.e., Kyrovohrad region — Responsible: SSSU
- Additional Q1 2024 guidance:
  - Continue to use weights from the year 2021 for the price indices in 2023 and 2024. — Target: May 2024
- Experimental work completed:
  - Experimental quality-adjusted hedonic indices for regional strata were compiled during the mission using a more standard application of hedonic adjustment.
  - Time coverage of experimental indices: only the four quarters of 2023 due to lack of time series data; can be extended when more quarterly data for 2024 becomes available.
  - Result: residential property prices generally increased in Ukraine during 2023, with significant regional variation — South and East not increasing as quickly or falling (regions most impacted by the war).

Updated method details (hedonic characteristics approach applied for Q1 2024):
- Method: single model using a four-quarter rolling window across all data (no stratification).
- Model specification:
  - Dependent variable: log of price per square meter.
  - Independent variables (dummy variables): (i) region, (ii) number of bedrooms, (iii) new or existing apartment, and (iv) time i.e., quarter.
  - A new regression is run every quarter.
- Predicted prices from successive quarterly regressions are used to re-price predefined property types (e.g., new apartment with two rooms in Kyiv) every quarter.
- Number of predefined property types: 150, held constant over time; price indices computed from updated estimated (predicted) prices and weights applied to compute higher-level indices.
- Replication attempt: Using 2023 data and the R statistical package, the mission calculated very similar coefficients on the explanatory variables, providing reassurance that regressions were being run correctly by the SSSU.

Short-term recommendations (implementable before May 20, 2024):
- Make better use of available weight information:
  - Currently a geometric mean of the regional estimated (predicted) prices is being applied.
  - Recommendation: apply a weighted mean of the regional prices so regions with greater expenditure on residential property have more impact on higher-level aggregates.
- Exclude regions with zero or very low numbers of observations.
- Increase data collection for one specific region: Kyrovohrad region.
- Continue to use weights from the year 2021 for the price indices in 2023 and 2024.
  - Rationale: Land registry weights showed unexplained fluctuations: in 2021 the split between new and existing properties was approximately one third for new properties and two-thirds for existing properties; in 2022 and 2023 the split reversed to approximately two-thirds new properties and one-third existing properties. This reversal is not thought to be real and may reflect recording changes since the start of the war. The SSSU will monitor land registry data and investigate further. SSSU use a combination of registry data and internal survey information for weights.

Recommended Actions (short-term):
- Use the available regional weights and apply a weighted mean rather than an unweighted geometric mean to the estimated (predicted) prices.
- Exclude regions with zero or very low numbers of observations.
- Increase data collection for one specific region i.e., Kyrovohrad region.
- Continue to use weights from the year 2021 for the price indices in 2023 and 2024.

### D. Compilation Methods — Beyond May 2024 (workplan items and target dates)
Priority actions and target completion dates:
- Data sources for the RPPI:
  - Expand the data collection for new properties within regions with low number of observations. — Nov 2024
  - Request all available data from OLX including the errors and outliers. — Dec 2024
  - Duplicates should be excluded for existing properties and included for new properties. — Dec 2024
  - Investigate whether it is possible to collect more detailed information on the location of the properties for sale. — Dec 2024
  - Extend the data collection to include houses i.e., via the OLX platform and via the regional SSSU offices. — Jun 2025
- Compilation methods – beyond May 2024:
  - Use a suitable statistical package such as R for the application of hedonic methods. — Dec 2025
  - Consult both internally and externally to decide on a suitable stratification of the data. — Dec 2024
  - Consolidate relevant categorical variables into homogenous groups. — Dec 2024
  - Experiment with other hedonic methods such as the time-dummy hedonic (TDH) method with a rolling window. — Dec 2024
  - Use a statistical method, such as Cook’s Distance, for outlier detection. — Dec 2024
  - Consider using the OLX data to compute the weights for existing properties. — May 2025
  - Become familiar with the requirements under relevant EU statistical legislation. — Dec 2024

Additional method recommendations and rationale:
- Stratification:
  - Consult internally and externally to decide on suitable stratification; mission tested stratification based on six aggregated regions (Central, East, Kyiv, North, South, and West) and two property types (new and existing) producing twelve experimental strata.
- Hedonic methods:
  - Hedonic methods include all regression-based methods; some use predicted prices while others use time-dummy coefficients to estimate price change with respect to time.
  - Mission suggests more standard hedonic applications and experimentation with time-dummy hedonic (TDH) methods with a rolling window.
- Outlier detection:
  - Use statistical methods such as Cook’s Distance for outlier detection.
- Weights and aggregation:
  - Apply available regional weights and use a weighted mean rather than an unweighted geometric mean for regional estimated prices (priority for May 2024).
  - Exclude regions with zero or very low observations for the Q1 2024 publication.

Recommended Actions (compilation):
- Use a suitable statistical package such as R for the application of hedonic methods.
- Consult both internally and externally to decide on a suitable stratification of the data.
- Consolidate relevant categorical variables into homogenous groups.
- Experiment with other hedonic methods such as the time-dummy hedonic (TDH) method with a rolling window.
- Use a statistical method, such as Cook’s Distance, for outlier detection.
- Consider using the OLX data to compute the weights for existing properties.
- Become familiar with the requirements under relevant EU statistical legislation.

Experimental indices and results:
- Experimental quality-adjusted hedonic indices for the regional strata were compiled during the mission using the hedonic (time-dummy) method over the four quarters of 2023.
- Model used for each stratum: dependent variable = log of price (in local currency); independent variables = (i) floor area, (ii) number of rooms, (iii) region. Region variable excluded from Kyiv model due to lack of additional location information within Kyiv.
- Result: Residential property prices were generally increasing in Ukraine during 2023, with significant variation across regions; prices in the South and East were not increasing as quickly or were falling (regions most impacted by the ongoing war).

Additional technical details:
- Categorical variable consolidation example — Floor number (existing properties only) consolidated into four categories:
  - (i) floor 1–5
  - (ii) floor 6–9
  - (iii) floor 10–15
  - (iv) floor 16+
- Table 3 — Number of Observations for Categories of Floor Number by Quarter, 2023Q1–2023Q4:
  - Floor 1-5: Q1 2023 = 14,380; Q2 2023 = 17,302; Q3 2023 = 19,704; Q4 2023 = 19,592
  - Floor 6-9: Q1 2023 = 5,717; Q2 2023 = 6,975; Q3 2023 = 7,950; Q4 2023 = 7,770
  - Floor 10-15: Q1 2023 = 2,601; Q2 2023 = 3,456; Q3 2023 = 3,869; Q4 2023 = 3,886
  - Floor 16+: Q1 2023 = 1,580; Q2 2023 = 2,363; Q3 2023 = 2,264; Q4 2023 = 2,379

### E. Dissemination
Findings and recommended medium/long-term actions:
- Current SSSU policy: published price indices cannot be revised.
- Recommended changes at the conclusion of the war in Ukraine (target dates: To be decided):
  - Review the revisions policy for the HPI.
  - Update the structure of the published HPI data.
  - Update the methods document and publish on the authorities’ website.
- Users would benefit from:
  - Revising past indices if data sources and methods are improved, to provide more accurate retrospective data.
  - Access to a longer time-series, e.g., a chained HPI with a single index reference period.
- Recommended dissemination structure:
  - (i) a long time-series for the overall HPI and sub-indices referenced to a specific year (e.g., 2023 = 100),
  - (ii) quarterly growth rates,
  - (iii) annual (year-on-year) growth rates,
  - (iv) annual average growth rates.
- Methods documentation:
  - SSSU published a methods document for the updated approach to be used for Q1 2024 (May 2024).
  - Recommendation: update the methods document after post-May 2024 improvements and publish on authorities’ website. Document should include data source, coverage, data quality and cleaning, outlier detection, stratification, weighting information, and compilation and aggregation methods.

Recommended Actions (dissemination):
- Review the revisions policy for the HPI at the conclusion of the war in Ukraine.
- Update the structure of the published HPI data at the conclusion of the war in Ukraine.
- Update the methods document and publish on the authorities’ website at the conclusion of the war in Ukraine.

### F. Officials Met During the Mission
- A team of three staff from the SSSU attended the mission meetings each day.
- Resident Representative office in Vienna under the leadership of Mr. Vahram Stepanyan provided logistics, planning and interpretation services.
- Specific officials:
  - Nataliia Lysenko — Head of Division for Housing and Construction Price Statistics, Department for Price Statistics, SSSU
  - Victoriia Isaiuk — Senior Specialist of the Division for Construction and Housing Price Statistics, Department for Price Statistics, SSSU
  - Alla Levchenko — Head of the Division for International Comparison, Department for Price Statistics, SSSU

*Source: IMF | Technical Assistance Report – Ukraine Residential Property Price Index Mission – Section I. Detailed Technical Assessment and Recommendations*

### Section I. Detailed Technical Assessment and Recommendations ...................................................... 5

### Section I. Detailed Technical Assessment and Recommendations

### A. Introduction
- A technical assistance (TA) mission was conducted from April 8–12, 2024, to assist the State Statistical Service of Ukraine (SSSU) with a methodological review of their House Price Index (HPI).
- Mission objectives completed:
  - (i) review listings data collected by the SSSU and the data preparation being applied,
  - (ii) assess the stratification and hedonic methods used for the HPI,
  - (iii) review weights and aggregation procedures used to compile the national index,
  - (iv) provide guidance on dissemination of the HPI,
  - (v) provide practical training to SSSU staff.
- The mission was conducted under the Data for Decisions (D4D) Trust Fund; a pre-mission meeting was held on March 20, 2023.
- Publication context:
  - SSSU published HPI in January 2024 for reference quarter Q4 2023 after approx. two years interruption.
  - SSSU plan to publish improved HPI for reference quarter Q1 2024 on May 20, 2024.
  - Main changes for Q1 2024 publication: (i) listings data as new source, (ii) use of offer prices instead of transaction prices, (iii) regression-based (hedonic) quality adjustment replacing stratified means.
- The mission provided practical training on hedonic methods and will continue supporting SSSU as the workplan is implemented.

### B. Data Sources — findings and recommended short-term actions
Findings:
- New residential properties (primary market):
  - Data manually collected by SSSU regional offices from real estate agents and property developers.
  - About 900 observations on new residential properties every quarter.
  - Sampling strategy based on regional populations.
  - Recommendation to expand collection within regions with low number of observations.
- Existing residential properties (secondary market):
  - SSSU uses online listings from OLX (www.olx.ua).
  - In 2023, SSSU received about 20,000-25,000 listings each quarter.
  - Web-scraping began in 2022; an MOU in Q1 2023 allowed direct receipt of OLX data.
- Data quality and preparation:
  - SSSU supplied four quarters of data for 2023; R was used for data analysis and visualization (histograms, scatter plots, descriptive statistics).
  - OLX data received by SSSU was already filtered (e.g., no properties < 30 square meters or with > three rooms).
  - Duplicates found in 2023: 75 for existing properties and 17 for new properties.
  - Offer prices can be quoted in Ukrainian hryvnia, US dollars or Euros; all offers should be converted to Ukrainian hryvnia using appropriate exchange rates. The RPPI should be compiled using Ukrainian hryvnia.
- Property characteristics available in listings data:
  - (i) offer prices,
  - (ii) location details i.e., region and city,
  - (iii) property attributes i.e., year of construction, total area in sqm, kitchen area in sqm, material type, number of rooms, floor number (existing properties only), total floors in the building (existing properties only).
  - Some variables have incomplete information (e.g., material type and year of construction) or may not significantly impact price; not all will be used in compilation.
- Geographic granularity:
  - Current collection: region and city only.
  - Ukraine has 24 regions; significant intra-regional price variation suggests collecting more granular location details (administrative districts or postcodes) would improve quality adjustment.
- Coverage:
  - Current listings data covers apartments only.
  - Recommendation: extend data collection to include houses via OLX and regional SSSU offices.

Recommended Actions (preserve phrasing):
- Expand the data collection for new properties within regions with low number of observations.
- Request all available data from OLX including the errors and outliers.
- Duplicates should be excluded for existing properties and included for new properties.
- Investigate whether it is possible to collect more detailed information on the location of the properties for sale.
- Extend the data collection to include houses i.e., via the OLX platform and via the regional SSSU offices.

### C. Compilation Methods — Q1 2024 Publication in May 2024
Findings and short-term recommendations:
- SSSU plan to implement hedonic methods for Q1 2024 publication on May 20, 2024. The hedonic method planned is a non-standard form of the characteristics approach.
- The new method is an improvement over stratified means by applying more refined quality adjustment, but more optimal hedonic methods exist.
- Due to the May publication deadline, mission recommended only short-term implementable changes.
- Priority recommendations (Table 1) with target date May 2024:
  - Use the available regional weights and apply a weighted mean rather than an unweighted geometric mean to the regional estimated prices. — Responsible: SSSU
  - Exclude regions with zero or very low number of observations. — Responsible: SSSU
  - Increase data collection for one specific region i.e., Kyrovohrad region — Responsible: SSSU

Additional Q1 2024 compilation guidance:
- Continue to use weights from the year 2021 for the price indices in 2023 and 2024. — Target: May 2024

Experimental work completed:
- Experimental quality-adjusted hedonic indices for regional strata were compiled during the mission using a more standard application of hedonic adjustment.
- Time coverage of experimental indices: only the four quarters of 2023 due to lack of time series data; can be extended when more quarterly data for 2024 becomes available.
- Result: residential property prices generally increased in Ukraine during 2023, with significant regional variation — South and East not increasing as quickly or falling (regions most impacted by the war).

### D. Compilation Methods — Beyond May 2024 (workplan items, target dates preserved)
Priority actions and target completion dates (from TABLE 2):
- Topic: Data sources for the RPPI
  - Expand the data collection for new properties within regions with low number of observations. — Nov 2024
  - Request all available data from OLX including the errors and outliers. — Dec 2024
  - Duplicates should be excluded for existing properties and included for new properties. — Dec 2024
  - Investigate whether it is possible to collect more detailed information on the location of the properties for sale. — Dec 2024
  - Extend the data collection to include houses i.e., via the OLX platform and via the regional SSSU offices. — Jun 2025
- Topic: Compilation methods – beyond May 2024
  - Use a suitable statistical package such as R for the application of hedonic methods. — Dec 2025
  - Consult both internally and externally to decide on a suitable stratification of the data. — Dec 2024
  - Consolidate relevant categorical variables into homogenous groups. — Dec 2024
  - Experiment with other hedonic methods such as the time-dummy hedonic (TDH) method with a rolling window. — Dec 2024
  - Use a statistical method, such as Cook’s Distance, for outlier detection. — Dec 2024
  - Consider using the OLX data to compute the weights for existing properties. — May 2025
  - Become familiar with the requirements under relevant EU statistical legislation. — Dec 2024

Additional method recommendations and rationale:
- Stratification:
  - SSSU should consult internally and externally to decide on suitable stratification of the data; mission tested a stratification based on six aggregated regions (Central, East, Kyiv, North, South, and West) and two property types (new and existing) producing twelve experimental strata.
- Hedonic methods:
  - Hedonic methods include all regression-based methods; some use predicted prices while others use time-dummy coefficients to estimate price change with respect to time.
  - The mission suggests more standard hedonic applications and experimentation with time-dummy hedonic (TDH) methods with a rolling window.
- Outlier detection:
  - Use statistical methods such as Cook’s Distance for outlier detection.
- Weights and aggregation:
  - Apply available regional weights and use a weighted mean rather than an unweighted geometric mean for regional estimated prices (priority for May 2024).
  - Exclude regions with zero or very low observations for the Q1 2024 publication.

### E. Dissemination
Findings and recommended medium/long-term actions:
- Current SSSU policy: published price indices cannot be revised.
- Recommended changes at the conclusion of the war in Ukraine (target dates: To be decided):
  - Review the revisions policy for the HPI.
  - Update the structure of the published HPI data.
  - Update the methods document and publish on the authorities’ website.
- Users would benefit from:
  - Revising past indices if data sources and methods are improved, to provide more accurate retrospective data.
  - Access to a longer time-series, e.g., a chained HPI with a single index reference period.

### F. Officials Met During the Mission
- A team of three staff from the SSSU attended the mission meetings each day.
- The Resident Representative office in Vienna under the leadership of Mr. Vahram Stepanyan provided logistics, planning and interpretation services.

*Source: IMF | Technical Assistance Report – Ukraine Residential Property Price Index Mission – Section I. Detailed Technical Assessment and Recommendations*

### 25. The SSSU first published the HPI in 2016. E ight HPIs were disseminated, across primary and

### tarea2024064-print-pdf - 25. The SSSU first published the HPI in 2016. E ight HPIs were disseminated, across primary and secondary markets for residential apartments, overall and for each room size i.e., one-room, two-room, and three-room apartments.

### Publication history and immediate context
- The SSSU first published the HPI in 2016.
- Eight HPIs were disseminated, across primary and secondary markets for residential apartments, overall and for each room size i.e., one-room, two-room, and three-room apartments.
- The authorities were not able to publish the HPI during 2022 and 2023 due to constraints imposed by the war in Ukraine.
- Publication resumed in January 2024.
- The existing method follows the stratified mean approach; the authorities plan to implement hedonic methods for the Q1 2024 publication which will be released on May 20, 2024. As this publication date was pre-announced, the SSSU are committed to meeting the deadline.
- In advance of the mission, the SSSU provided SPSS code and MS Excel spreadsheets to assist assessment of the updated method.

### Updated method applied for Q1 2024 (hedonic characteristics approach)
- The updated method is a hedonic characteristics approach applied in a non-standard format.
- The new method applies a single model, using a four-quarter rolling window, across all the data (no stratification).
- Model specification:
  - Dependent variable: log of price per square meter.
  - Independent variables (dummy variables): (i) region, (ii) number of bedrooms, (iii) new or existing apartment, and (iv) time i.e., quarter.
  - A new regression is run every quarter.
- Predicted prices from successive quarterly regressions are used to re-price predefined property types (e.g., new apartment with two rooms in Kyiv) every quarter.
- There are 150 predefined property types, held constant over time; price indices are computed from updated estimated (predicted) prices and weights are applied to compute higher-level indices.
- Replication attempt: Using 2023 data and the R statistical package, the mission calculated very similar coefficients on the explanatory variables, providing reassurance that regressions were being run correctly by the SSSU.

### Short-term recommendations (implementable before May 20, 2024)
- Make better use of available weight information:
  - Currently a geometric mean of the regional estimated (predicted) prices is being applied.
  - Recommendation: apply a weighted mean of the regional prices so regions with greater expenditure on residential property have more impact on higher-level aggregates.
- Exclude regions with zero or very low numbers of observations.
- Increase data collection for one specific region: Kyrovohrad region.
- Continue to use weights from the year 2021 for the price indices in 2023 and 2024.
  - Rationale: Land registry weights showed unexplained fluctuations: in 2021 the split between new and existing properties was approximately one third for new properties and two-thirds for existing properties; in 2022 and 2023 the split reversed to approximately two-thirds new properties and one-third existing properties. This reversal is not thought to be real and may reflect recording changes since the start of the war. The SSSU will monitor land registry data and investigate further. SSSU use a combination of registry data and internal survey information for weights.

Recommended Actions (short-term):
- Use the available regional weights and apply a weighted mean rather than an unweighted geometric mean to the estimated (predicted) prices.
- Exclude regions with zero or very low numbers of observations.
- Increase data collection for one specific region i.e., Kyrovohrad region.
- Continue to use weights from the year 2021 for the price indices in 2023 and 2024.

### Compilation methods — beyond May 2024
- Software and coding:
  - Transfer processes currently in MS Excel and SPSS to the R statistical package (data cleaning/filtering, outlier detection, index calculation, aggregation).
  - Authorities agreed and intend to upskill coding capacity in R; all R scripts written during the mission were provided to the authorities.
- Stratification:
  - The SSSU should consult internally and externally to decide on a suitable stratification; once implemented it is not recommended to change the stratification frequently.
  - Each stratum should contain sufficient observations each quarter for stable indices.
  - Mission-tested stratification: six aggregated regions (Central, East, Kyiv, North, South, and West) × two types of properties (new, existing) = twelve strata.
- Categorical variable consolidation:
  - Merge property characteristics into homogenous groups to ensure each category has enough observations for accurate coefficient estimation (shadow prices).
  - Example: floor number (existing properties only) consolidated into four categories:
    - (i) floor 1–5
    - (ii) floor 6–9
    - (iii) floor 10–15
    - (iv) floor 16+
  - Table 3 — Number of Observations for Categories of Floor Number by Quarter, 2023Q1–2023Q4:
    - Floor 1-5: Q1 2023 = 14,380; Q2 2023 = 17,302; Q3 2023 = 19,704; Q4 2023 = 19,592
    - Floor 6-9: Q1 2023 = 5,717; Q2 2023 = 6,975; Q3 2023 = 7,950; Q4 2023 = 7,770
    - Floor 10-15: Q1 2023 = 2,601; Q2 2023 = 3,456; Q3 2023 = 3,869; Q4 2023 = 3,886
    - Floor 16+: Q1 2023 = 1,580; Q2 2023 = 2,363; Q3 2023 = 2,264; Q4 2023 = 2,379
- Hedonic methods to experiment with:
  - Time-dummy hedonic (TDH) method with a rolling window (demonstrated by the mission using R). Advantages:
    - Price indices can be calculated directly from coefficients on time-dummies, simplifying calculations.
    - Tailored models for each stratum allow more refined quality adjustment (e.g., Kyiv-specific model if more granular location data collected).
- Outlier detection:
  - Use statistical, regression-based methods such as Cook’s Distance applied within each stratum for every rolling window quarter.
  - Suggested threshold for outlier determination: four divided by the number of observations used in the regression.
- Weights for existing properties:
  - Consider using OLX data to compute weights for existing properties; OLX is considered representative of regional expenditures on residential property.
  - Authorities should compare weights from the SSSU survey of transacted apartments with weights computed using OLX offer prices.
  - Initial comparison: OLX data would give a smaller weight to one-room apartments and a larger weight to three-bedroom apartments across many regions.
- Experimental indices:
  - Experimental quality-adjusted hedonic indices for the regional strata were compiled during the mission using the hedonic (time-dummy) method over the four quarters of 2023.
  - Model used for each stratum: dependent variable = log of price (in local currency); independent variables = (i) floor area, (ii) number of rooms, (iii) region. Region variable excluded from Kyiv model due to lack of additional location information within Kyiv.
  - Result: Residential property prices were generally increasing in Ukraine during 2023, with significant variation across regions; prices in the South and East were not increasing as quickly or were falling (regions most impacted by the ongoing war).

Recommended Actions (compilation):
- Use a suitable statistical package such as R for the application of hedonic methods.
- Consult both internally and externally to decide on a suitable stratification of the data.
- Consolidate relevant categorical variables into homogenous groups.
- Experiment with other hedonic methods such as the time-dummy hedonic (TDH) method with a rolling window.
- Use a statistical method, such as Cook’s Distance, for outlier detection.
- Consider using the OLX data to compute the weights for existing properties.
- Become familiar with the requirements under relevant EU statistical legislation.

### EU compliance
- Relevant EU legislation: EU regulation 2016/792 on harmonized indices of consumer prices and the house price index and EU regulation 2023/1470 laying down methodological and technical specifications as regards the house price index and the owner-occupied housing price index.
- If recommendations are implemented, the authorities will largely be compliant with this legislation.
- Future TA will assist familiarization with requirements and further harmonization of indices.

### Dissemination — beyond May 2024
- Revisions policy:
  - Current SSSU policy: published price indices cannot be revised.
  - Recommendation: review revisions policy for the HPI at the conclusion of the war in Ukraine. If data sources and compilation methods are developed as proposed, it would be preferable to update existing price indices with new price indices to provide users more accurate retrospective data.
  - Revisions should be accompanied by explanatory documents detailing data source and method changes and why new indices are higher quality.
- Structure of published HPI data:
  - Current dissemination presents rates of price change according to four different measures.
  - Recommendation: publish a longer time-series (e.g., a chained HPI with a single index reference period), specifically:
    - (i) a long time-series for the overall HPI and sub-indices referenced to a specific year (e.g., 2023 = 100),
    - (ii) quarterly growth rates,
    - (iii) annual (year-on-year) growth rates,
    - (iv) annual average growth rates.
  - Authorities will examine data structure at the conclusion of the war in Ukraine.
- Methods documentation:
  - SSSU have published a methods document for the updated approach to be used for Q1 2024 (May 2024).
  - Recommendation: update the methods document after post-May 2024 improvements and publish on authorities’ website. Document should include data source, coverage, data quality and cleaning, outlier detection, stratification, weighting information, and compilation and aggregation methods.

Recommended Actions (dissemination):
- Review the revisions policy for the HPI at the conclusion of the war in Ukraine.
- Update the structure of the published HPI data at the conclusion of the war in Ukraine.
- Update the methods document and publish on the authorities’ website at the conclusion of the war in Ukraine.

### Officials met during the mission
- Nataliia Lysenko — Head of Division for Housing and Construction Price Statistics, Department for Price Statistics, SSSU
- Victoriia Isaiuk — Senior Specialist of the Division for Construction and Housing Price Statistics, Department for Price Statistics, SSSU
- Alla Levchenko — Head of the Division for International Comparison, Department for Price Statistics, SSSU

*IMF | Technical Assistance Report – Ukraine Residential Property Price Index Mission*

---


_Source: https://www.imf.org/-/media/files/publications/tar/2024/english/tarea2024064-print-pdf.pdf_