## cr17296

## Source details

**Canonical URL:** [cr17296](https://www.imf.org/-/media/files/publications/cr/2017/cr17296.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2017/cr17296.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2017/cr17296.pdf.json)

---

### Overview and objectives
- Mission: Technical Assistance (TA) mission to Kampala during the period September 15–26, 2014 to assess the concepts and methods used to compile the Consumer Price Index (CPI). Activity: 15ST054. Project: Prices Statistics STA_UGA_2013_05.
- Objectives:
  - Review concepts and methods used to compile the new rebased CPI.
  - Address AFR concerns with the current CPI series: (i) Revisions made to published indices after publication; (ii) published inflation rates inconsistent with inflation rates calculated from published indices; (iii) absence of consistency in aggregation; (iv) the average of the index reference period does not match 100; and (v) inconsistent sampling with some items only in one or two areas.
- Timeline and indexing:
  - First release of the rebased CPI planned for January 2015.
  - Indices to be calculated retrospectively to the weight reference period July 2009–June 2010 (new index reference period).
  - The rebased CPI will be chain linked with the previous series at the All Items CPI level.
- Training and outputs:
  - Training provided on CPI compilation, data collection, data validation, and Excel tools for CPI production.
  - A detailed work program prepared listing main activities to be undertaken before release.

### Methodological soundness and scope
- Classification and structure:
  - Changeover to the Classification of Individual Consumption According to Purpose (COICOP) to improve international comparability and coherence.
  - COICOP’s 12 divisions will be divided into groups and classes; classes further divided into sub-classes reflecting local circumstances, with more detailed subdivisions for food products.
- Scope and exclusions:
  - Explicitly excluded goods and services: narcotics, gambling, prostitution, life insurance and illegal transactions.
  - CPI coverage: purely urban CPI — weights and prices refer only to urban households; institutional households excluded.
- Stratification and geographic coverage:
  - Current CPI: eight main baskets (Jinja, Mbale, Masaka, Mbarara, Gulu, Arua, Kampala high-income, Kampala middle/low-income).
  - Rebased CPI: expand geographical coverage by adding Fort Portal and subdivide Kampala into three baskets (low, middle and high income); total baskets increase from eight to ten.
  - Rationale: Kampala accounts for around half of the CPI basket; finer stratification expected to improve accuracy.
- Conceptual basis (national vs domestic):
  - Current CPI based on the national concept (non-business expenditures of resident population regardless of where expenditure takes place).
  - UBOS initially planned to adopt the domestic concept for the rebased CPI (expenditures made within the domestic territory including non-business expenditure by foreign visitors) to meet regional needs.
  - Mission recommendation: headline CPI should remain based on the national concept for policy-making and international comparability; consistency with past figures is desirable when linking series.
  - Practical compromise agreed: produce two versions via two different weighting schemes using identical elementary aggregates (EAs):
    - National-concept CPI including non-monetary consumption (except imputed rents).
    - Domestic-concept CPI covering Household Final Monetary Consumption Expenditure (HFMCE).
- Treatment of non-monetary and imputed items:
  - UBOS had planned to restrict CPI to HFMCE, excluding non-monetary consumption such as own-account production; mission notes arguments for inclusion (neutralize long-run substitution shifts, affect living standards, coherence with National Accounts deflator).
  - Imputed rents for owner-occupiers in principle should be included; UBOS cites unreliable rent data and plans to improve the rent survey and coordinate on real estate.

### Accuracy and reliability — source data and weights
- Primary weight source:
  - Uganda National Household Survey (UNHS) 2009/2010 (data collection period July 2009–June 2010).
- UNHS sample design and size:
  - Two-stage sampling: enumeration areas sampled first, then households within sampled enumeration areas.
  - UNHS 2009/10 sample size: 6,800 households.
  - Comparison: 7,400 households in the 20005/06 module.
  - UBOS assertion: UNHS 2009/10 sample designed to allow reliable estimation for (i) the whole country, (ii) for rural or urban areas, and (iii) for each of the ten sub-regions.
- Recall periods used for recording expenditures:
  - 7-day recall period for expenditure on food, beverages and tobacco.
  - 30-day recall period for household consumption expenditure on nondurable goods and frequently purchased services such as transport and health services.
  - 365-day recall period for semi-durable and durable goods and services.
- Use of the Mini Household Budget Survey (MHBS):
  - MHBS conducted in 2014 to further disaggregate expenditures at the lowest levels of classification.
  - During the mission, MHBS data entry and editing were underway; final results not yet available.
  - MHBS sampling design: two-stage sampling; enumeration areas sampled in the eight price collection centers; 10 households sampled in each of the 74 sampled enumeration areas.
  - MHBS does not cover 12 months; "the impact of seasonality on the results should be assessed."
  - MHBS intended to decompose expenditures at lower levels, provide product and outlet information for defining EAs, and assist with net valuation of second-hand goods.
- Weight derivation and comparisons:
  - Preliminary higher-level weights analyzed against current CPI (based on UNHS 2005/2006): Kampala high income basket has a very high share compared to other Kampala baskets; unusual weights observed for accommodation and restaurants — further analysis required.
  - UNHS vs Supply and Use Tables (SUT): evidence of underreporting in UNHS for items such as tobacco and beer.
  - Recommendation: adjust weights for under- or over-reporting by comparing UNHS to other sources.
- Domestic principle implementation:
  - Additional sources used to estimate expenditures by non-residents: Tourism Expenditure and Motivation Survey (TEMS) and Informal Cross Border Trade Survey (ICBTS).
  - TEMS carried out in May 2012; contains both business and non-business expenditures and requires adjustments.
  - ICBTS carried out on a continuous basis during two weeks each month; data classified into COICOP categories.
  - If adopting the domestic principle, adjustments should be made to exclude residents’ expenditures abroad and business expenditures of tourists.
- Second-hand goods:
  - Current CPI includes second-hand goods for clothing and vehicles using gross weights; recommendation to use net weights as transactions between households cancel out.
  - Planned improvement: valuation of second-hand goods in the rebased CPI.

### Statistical techniques, aggregation, and benchmarking
- Aggregation inconsistency:
  - AFR flagged absence of consistency in aggregation in the current CPI: aggregation at the national level of group indices does not equal All Items CPI.
  - Cause identified: method of combining regional indices into national indices.
  - Proposed action: change regional aggregation method so inconsistencies are removed; UBOS plans to improve higher-level aggregation in the rebased CPI.
  - Proposed aggregation approach would:
    - Provide identical results for each basket separately;
    - Use the same item weights in each basket and at the national level;
    - Lead to the same national All Items CPI (national indices for groups and sub-groups would differ).
- Prior adjustments to UNHS data and benchmarking:
  - Out-of-scope goods and services removed; only monetary consumption of urban households considered; some UNHS categories reclassified to match COICOP.
  - Observed differences: total weight for Kampala dropped relative to other sub-regions compared to current CPI; within Kampala, high income basket share very high and to be reviewed.
  - Underreporting in UNHS noted as serious; example: significant underreporting for tobacco or beer compared with SUT and household final consumption.
  - Correction coefficients can be established by comparing UNHS with SUT and NA.
  - Some COICOP categories with no expenditures recorded in UNHS (e.g., insurance or financial services) may be included if reliable weights estimated using NA data; for insurance, weights should be based on the implicit service charge.
- Regional grouping and coverage gaps:
  - Two sub-regions have no basket; correction method:
    - All sub-regions grouped into four main regions: Central, Eastern, Western, Northern.
    - Total expenditure of parts not covered added proportionally to other baskets of the same region.
    - Assumes price changes same in price collection center as the whole sub-region.
- Cut-offs, manual review and mapping:
  - Cut-off level of one part per thousand used for including an EA in the CPI basket.
  - Expenditure of EAs below threshold distributed proportionally to other EAs of same COICOP class or assigned to similar EAs.
  - Final exercise concluded by manual review and adjustments as needed.
  - Once weighting scheme validated, list of EAs covered in rebased CPI to be finalized and current price data mapped into new classification.

### Index compilation, elementary aggregation, and handling missing prices
- Reference period, base prices, and backward revision:
  - Rebased CPI retrospectively calculated up to weight reference period July 2009–June 2010, which will become the new index reference period.
  - A base price will be estimated corresponding to the arithmetic average of the prices observed during the 12 months’ index reference period.
  - Prices must be imputed if not available up to this period; UBOS intends to use ICP price data for backward imputation if needed.
  - Backward revision links weight and price reference periods and enables compilation of a Laspeyres price index; this reverses the previous approach that derived a Lowe price index by price-updating weight-period expenditures.
  - No monthly indices will be published for the 12-month index reference period; the index for the reference period as a whole will be 100 by definition.
- Elementary aggregation and index formulae:
  - EAs are the lowest level for reliable weight information.
  - Default method at elementary level: unweighted geometric mean of individual price quotes (Jevons price index).
  - Within an EA there may be one or several representative items; representative items implicitly weighted by number of price quotes collected for each.
  - Using appropriate weights, EAs should be progressively aggregated within each basket using the Two-stage (Modified) Laspeyres index formula.
- Temporarily missing prices:
  - Temporarily missing prices are imputed for a maximum of three months.
  - Imputation process is bottom-up: use average price change within same EA before moving up classification (within same basket) if no prices available at level below.
  - No indicators produced on number of missing prices per month; CPI staff confirmed their number was low.

### Price collection, quality adjustment, and production system
- Organization and coverage:
  - Price collection program coordinated by CPI staff.
  - In each of the seven urban price collection centers (eight centers in the rebased CPI), price collectors report to one of the supervisors at headquarters.
  - National pricing policies collected only once and same price included in all regional baskets.
  - Second-hand clothing prices collected only in Kampala and used as proxy for other baskets.
- Frequency, timing, and volume:
  - Frequency depends on product type; some prices collected quarterly or less frequently (rents, health services, electricity, water charges, education).
  - Telecommunication prices collected whenever tariffs change.
  - Prices are generally collected during the week including the 15th of the month.
  - Fresh food: prices collected twice per month (first week and mid-week).
  - Kampala: some products collected every week for the weekly CPI, but only first week and mid-week prices used for official CPI.
  - "There are more than 7,000 prices which enter the CPI every month."
- Market collection and outlet selection:
  - Large share of prices collected in markets by purchasing products; products weighted at the office and a unit price calculated approximating purchasers’ prices after negotiation.
  - Selection of outlets and products purposive by the supervisor; replacements used if a product is missing for three months or outlet closed.
- Recording, validation, and training:
  - Prices recorded on paper questionnaires; previous month price not included to avoid copying.
  - Price collectors encouraged to report events and explanations; non-standardized comments allowed.
  - Data validation uses thresholds to detect outliers; supervisors follow up and may confirm prices with respondents; rule: do not edit a reported price unless confirmed an error.
  - Regular training organized for price collectors; written instructions and item specifications provided.
- Product comparability and quality change:
  - Price data sufficiently detailed to ensure period-to-period comparisons within selected outlets refer to the same variety.
  - Clothing items use specifications partly from the International Comparison Program (ICP); field identification of quality changes may be difficult.
  - Quality adjustments: replacements use the overlap method when new and old product prices are available in the same month to implicitly adjust for quality differences.
- Production system:
  - CPI compiled in a range of Excel files; UBOS considers Excel should be further developed before switching.
  - Identified improvements: automate repetitive tasks, avoid manual transfers between files.
  - System should permit different weighting schemes and be flexible for future adjustments.

### Chain linking, serviceability, dissemination, and metadata
- Chain linking:
  - Plan: chain link the rebased CPI with the previous CPI series at the All Items level using the new index reference period July 2009–June 2010 as the overlapping period (annual overlap method).
  - Implication: price developments during overlap consistent with old CPI; implicit price change between June 2010 and July 2010 may be flawed.
  - UBOS may recalculate and link rebased CPI series at COICOP division level backwards on user request.
- Frequency, timeliness, and publication:
  - Ugandan CPI compiled monthly and released on the last day of the reference month; timeliness exceeds GDDS and SDDS standards.
  - Pre-release communication: users to be informed in advance; UBOS plans presentations to consultation committees, media and public.
  - Publication breakdown: indices and weights broken down to COICOP class level for food and non-alcoholic beverages and to COICOP division level for remaining products; results at national level and for the 10 baskets separately.
  - In future, the weight of each basket in the national CPI should be published.
- Rounding and precision:
  - Monthly and annual changes compiled using indices with full precision.
  - For publication: weights rounded to four digits, indices rounded to two digits and rates of change rounded to one digit.
- Weight review and metadata:
  - UBOS plans an annual review of weights and adjustments at lower levels as needed.
  - Recommended to update weights at least every five years; more frequent complete revisions if resources permit.
  - GDSS metadata should be up to date and reviewed regularly; a comprehensive methodological document should be published and updated regularly.
- Training and work program milestones:
  - Training covered chain-linking, adjusting weights at lower levels, and advanced Excel tools.
  - Work program objective: Release the new CPI in January 2015.
  - Selected milestones:
    - October: Compare UNHS with other sources and correct under/over reporting; validate regional weights; finalize weight derivation; analyze MHBS.
    - November: Define EAs and item specifications; map existing prices; estimate base prices; initiate new price collection; train price section staff.
    - December: Finalize and test compilation system; calculate price indices from reference period to now; chain-link old CPI; pre-announce rebasing.
    - January (2015): Stop price collection for items removed; prepare dissemination; First release of the rebased CPI January 30.
    - March (2015): Update metadata and prepare methodology documentation.

### Current CPI series — identified issues and recommendations
- Main concerns:
  - Revisions made to published indices after publication.
  - Published inflation rates inconsistent with inflation rates calculated from published indices.
  - Absence of consistency in aggregation.
  - The average of the index-reference-period indices does not = 100.
  - Inconsistent sampling, with some items only sampled in one or two areas.
- Revisions practice and recommendations:
  - Current practice: revise the previous month's CPI when late data arrive, especially from education and health providers; provisional status is indicated in press release and revised results published next month.
  - UBOS view: revising past results improves CPI quality and revisions tend to be marginal.
  - Counterpoint: CPI commonly used for indexation and therefore should not be revised routinely.
  - Strategies to avoid revisions:
    - Consult users about publication schedule relaxation.
    - Accelerate price collection and follow-up procedures.
    - Review and replace uncooperative respondents.
    - Systematic oversampling to ensure sufficient price coverage.
  - UBOS considers revision policy a strategic decision requiring stakeholder consultation, especially with the Bank of Uganda.
- Inconsistency in aggregation:
  - Current practice aggregates items per basket and then aggregates subgroup, group and All Items indices of baskets into national indices using same basket shares, causing inconsistency.
  - Recommendation: adjust aggregation method so results independent of aggregation sequence; UBOS plans improvements in rebased CPI.
- Base period index inconsistency:
  - Arithmetic average of the indices of the index reference period must equal 100, but was found not to hold for the overall CPI.
  - Top-down analysis traced inconsistency to Kampala low and middle income basket and within that to the group “health, entertainment and other services” and further to an item in the health sub-group with an erroneous base price.
  - Recommendation: make consistency checks by averaging index reference period indices at all aggregation levels and compile results with sufficient precision.
- Regional sampling of items:
  - Basket consists of 276 items.
  - Sampling distribution:
    - 14 percent of items sampled in only one basket.
    - 7 percent of items sampled in only two baskets.
  - Cause: items entered a basket only if expenditure found significant in that basket; items sampled in a limited number of baskets are mainly in Kampala or the Northern region.

### Key numerical and administrative details
- Mission dates: September 15–26, 2014.
- Activity identifier: 15ST054.
- Project: Prices Statistics STA_UGA_2013_05.
- Weight reference period / new index reference period: July 2009–June 2010.
- Planned first release of rebased CPI: January 2015.
- Change in number of baskets: increase from eight to ten.
- UNHS 2009/10 sample size: 6,800 households.
- UNHS 20005/06 module sample size (comparison): 7,400 households.
- Recall periods used in UNHS: 7-day; 30-day; 365-day.
- MHBS sampling: 74 sampled enumeration areas with 10 households sampled in each area.
- TEMS carried out in May 2012.
- Temporarily missing prices imputed for a maximum of three months.
- Publication rounding: weights to four digits, indices to two digits, rates of change to one digit.
- "There are more than 7,000 prices which enter the CPI every month."

*Executive Summary of the IMF TA report on the rebasing of the Uganda CPI (mission September 15–26, 2014).*

### Executive Summary ......................................................................................................

### Executive Summary

### Overview
- Mission: Technical Assistance (TA) mission to Kampala during the period September 15–26, 2014 to assess the concepts and methods used to compile the Consumer Price Index (CPI). Activity: 15ST054. Project: Prices Statistics STA_UGA_2013_05.
- Objectives:
  - Review concepts and methods used to compile the new rebased CPI.
  - Address AFR concerns with the current CPI series: (i) Revisions made to published indices after publication; (ii) published inflation rates inconsistent with inflation rates calculated from published indices; (iii) absence of consistency in aggregation; (iv) the average of the index reference period does not match 100; and (v) inconsistent sampling with some items only in one or two areas.
- Timeline and indexing:
  - First release of the rebased CPI planned for January 2015.
  - Indices to be calculated retrospectively to the weight reference period July 2009–June 2010 (new index reference period).
  - The rebased CPI will be chain linked with the previous series at the All Items CPI level.
- Training and outputs:
  - Training provided on CPI compilation, data collection, data validation, and Excel tools for CPI production.
  - A detailed work program prepared listing main activities to be undertaken before release.

### Methodological Soundness
- Classification and structure:
  - Changeover to the Classification of Individual Consumption According to Purpose (COICOP) to improve international comparability and coherence.
  - COICOP’s 12 divisions will be divided into groups and classes; classes further divided into sub-classes reflecting local circumstances, with more detailed subdivisions for food products.
- Scope and exclusions:
  - Explicitly excluded goods and services: narcotics, gambling, prostitution, life insurance and illegal transactions.
  - CPI coverage: purely urban CPI — weights and prices refer only to urban households; institutional households excluded.
- Stratification and geographic coverage:
  - Current CPI: eight main baskets (Jinja, Mbale, Masaka, Mbarara, Gulu, Arua, Kampala high-income, Kampala middle/low-income).
  - Rebased CPI: expand geographical coverage by adding Fort Portal and subdivide Kampala into three baskets (low, middle and high income); total baskets increase from eight to ten.
  - Rationale: Kampala accounts for around half of the CPI basket; finer stratification expected to improve accuracy.
- Conceptual basis (national vs domestic):
  - Current CPI based on the national concept (non-business expenditures of resident population regardless of where expenditure takes place).
  - UBOS initially planned to adopt the domestic concept for the rebased CPI (expenditures made within the domestic territory including non-business expenditure by foreign visitors) to meet regional needs.
  - Mission recommendation: headline CPI should remain based on the national concept for policy-making and international comparability; consistency with past figures is desirable when linking series.
  - Practical compromise agreed: produce two versions via two different weighting schemes using identical elementary aggregates (EAs):
    - National-concept CPI including non-monetary consumption (except imputed rents).
    - Domestic-concept CPI covering Household Final Monetary Consumption Expenditure (HFMCE).
- Treatment of non-monetary and imputed items:
  - UBOS had planned to restrict CPI to HFMCE, excluding non-monetary consumption such as own-account production; mission notes arguments for inclusion (neutralize long-run substitution shifts, affect living standards, coherence with National Accounts deflator).
  - Imputed rents for owner-occupiers in principle should be included; UBOS cites unreliable rent data and plans to improve the rent survey and coordinate on real estate.

### Accuracy and Reliability — Source Data and Weights
- Primary weight source:
  - Uganda National Household Survey (UNHS) 2009/2010 (data collection period July 2009–June 2010).
  - UNHS purpose: collect data on demographic, social and economic characteristics of household population.
- UNHS sample design and size:
  - Two-stage sampling: enumeration areas sampled first, then households within sampled enumeration areas.
  - UNHS 2009/10 sample size: 6,800 households.
  - Comparison: 7,400 households in the 20005/06 module.
  - UBOS assertion: UNHS 2009/10 sample designed to allow reliable estimation for (i) the whole country, (ii) for rural or urban areas, and (iii) for each of the ten sub-regions.
- Recall periods used for recording expenditures:
  - 7-day recall period for expenditure on food, beverages and tobacco.
  - 30-day recall period for household consumption expenditure on nondurable goods and frequently purchased services such as transport and health services.
  - 365-day recall period for semi-durable and durable goods and services.
- Use of the Mini Household Budget Survey (MHBS):
  - MHBS conducted in 2014 to further disaggregate expenditures at the lowest levels of classification.
  - During the mission, MHBS data entry and editing were underway; final results not yet available.
- Weight derivation and comparisons:
  - Preliminary higher-level weights analyzed against current CPI (based on UNHS 2005/2006): Kampala high income basket has a very high share compared to other Kampala baskets; unusual weights observed for accommodation and restaurants — further analysis required.
  - UNHS vs Supply and Use Tables (SUT): evidence of underreporting in UNHS for items such as tobacco and beer.
  - Recommendation: adjust weights for under- or over-reporting by comparing UNHS to other sources.
- Domestic principle implementation:
  - Additional sources used to estimate expenditures by non-residents in Uganda contain inconsistencies: (i) expenditures of resident households abroad are included; and (ii) some business expenditures of tourists in Uganda are included.
  - If adopting the domestic principle, adjustments should be made to address these inconsistencies.
- Second-hand goods:
  - Current CPI includes second-hand goods for clothing and vehicles using gross weights; recommendation to use net weights as transactions between households cancel out.
  - Planned improvement: valuation of second-hand goods in the rebased CPI.

### Accuracy and Reliability — Statistical Techniques and Aggregation
- Aggregation inconsistency:
  - AFR flagged absence of consistency in aggregation in the current CPI: aggregation at the national level of group indices does not equal All Items CPI.
  - Cause identified: method of combining regional indices into national indices.
  - Proposed action: change regional aggregation method so inconsistencies are removed; UBOS plans to improve higher-level aggregation in the rebased CPI.
- Revisions practice:
  - Current practice to revise the CPI figure of the previous month due to late price data entry, especially from education and health service providers.
  - General best practice recommended: CPI should not be revised routinely.
  - UBOS view: revision policy is a strategic decision requiring further consultation with stakeholders, especially the Bank of Uganda.

### Serviceability and Implementation
- Implementation plan:
  - Detailed work program prepared including main activities to be undertaken before the rebased CPI can be released.
  - Once weighting scheme finalized, the list of elementary aggregates (EAs) for the rebased CPI can be defined.
- Publication details:
  - For publication, indices and weights will be broken down to the COICOP class level for food and non-alcoholic beverages and to the COICOP division level for remaining products.
- Training and capacity:
  - CPI staff sufficiently experienced and trained to carry out the rebasing; preparatory work started in 2012.
  - Census activities delayed finalization of weighting scheme and release.
  - Training provided on CPI compilation, data collection, data validation; additional Excel tools training delivered on request.

### Key Numerical and Administrative Details
- Mission dates: September 15–26, 2014.
- Activity identifier: 15ST054.
- Project: Prices Statistics STA_UGA_2013_05.
- Weight reference period / new index reference period: July 2009–June 2010.
- Planned first release of rebased CPI: January 2015.
- Change in number of baskets: increase from eight to ten.
- UNHS 2009/10 sample size: 6,800 households.
- UNHS 20005/06 module sample size (comparison): 7,400 households.
- Recall periods used in UNHS: 7-day; 30-day; 365-day.

*Executive Summary of the IMF TA report on the rebasing of the Uganda CPI (mission September 15–26, 2014).*

### 29. In order to further disaggregate the expenditures obtained from the UNHS, an

### cr17296 - 29. In order to further disaggregate the expenditures obtained from the UNHS, an

### MHBS survey and disaggregation of UNHS expenditures
- An additional survey, called MHBS, was carried out during the third quarter of 2014 to further disaggregate the expenditures obtained from the UNHS.
- The MHBS does not cover 12 months; "the impact of seasonality on the results should be assessed."
- Sampling design:
  - Two-stage sampling.
  - Enumeration areas sampled in the eight price collection centers covered in the rebased CPI.
  - 10 households sampled in each of the 74 sampled enumeration areas.
- Purposes and potential uses:
  - Decompose expenditures at lower levels of the CPI.
  - Provide additional information on products and outlets to consult when defining and reviewing the EA.
- Status during mission:
  - Data entry and editing took place but final results were not available yet.
  - Consequently, the quality of the results obtained with the MHBS could not be assessed.
- Once results from the MHBS are available:
  - It will be possible to further split the expenditures derived from the UNHS and define weights for the EAs.
  - In general, the EAs will be defined at COICOP sub-class level or below.
  - The MHBS may provide information needed to adjust weights for second-hand goods (net principle), since the UNHS only contains gross weights.

### Other sources for non-resident and informal expenditures
- Two additional sources are used to estimate expenditures of non-residents in Uganda:
  - (i) The Tourism Expenditure and Motivation Survey (TEMS).
  - (ii) The Informal Cross Border Trade Survey (ICBTS).
- TEMS:
  - Carried out in May 2012.
  - Objective: determine the value of tourism activities and current demand for tourism.
  - Questionnaires handed out to tourists at the main entry points.
  - Provided information on total expenditure of tourists for broad product categories.
  - TEMS amounts contain both business and non-business expenditures; adjustments needed because CPI is restricted to household final consumption.
- ICBTS:
  - Carried out on a continuous basis during two weeks each month.
  - Collects information on informal exports and imports from different border points.
  - Data have been classified as far as possible into their respective COICOP categories.
- Allocation rules discussed:
  - Informal exports assigned to the sub-region where the border point is located.
  - Tourist expenditure assignment:
    - Selected tourist sites assigned to the geographically closest basket; expenditures allocated proportionally to number of visits to these sites.
    - For Kampala (divided into three baskets based on income levels), an ad-hoc allocation was proposed; subsequently proposed that tourism expenditure be assigned to the three baskets proportionally to the total weight of each basket.
- Conceptual issues and adjustments:
  - UNHS covers expenditures of residents both in Uganda and abroad; expenditures of residents abroad should be excluded under the domestic principle—estimation needed.
  - TEMS includes business expenditures which should be excluded from CPI weights; TEMS contains visit-purpose information enabling additional adjustments.
  - Future data structuring recommended to derive two parallel weighting schemes (national and domestic), distinguishing at higher levels:
    - (i) consumption by residents in Uganda;
    - (ii) consumption by residents abroad;
    - (iii) consumption by non-residents in Uganda;
    - (iv) own account consumption.
  - These breakdowns should be available at higher levels; at lower levels same breakdowns derived from the MHBS may be used.

### Price collection: program, practices, and data quality
- Organization and coverage:
  - Current price collection program directly coordinated by CPI staff.
  - In each of the seven urban price collection centers (eight centers in the rebased CPI), price collectors report to one of the supervisors at headquarters.
  - In general, prices are collected locally in different price collection centers.
  - National pricing policies (water, electricity, telecom) collected only once and same price included in all regional baskets.
  - Second-hand clothing prices collected only in Kampala and serve as proxy for other baskets.
- Frequency and timing:
  - Frequency depends on product type; some prices collected only quarterly or less frequently (rents, health services, electricity, water charges, education).
  - Telecommunication prices collected whenever there is a change in tariffs.
  - In general, prices are collected during the week which includes the 15th of the month.
  - Fresh food: prices collected twice per month (first week and mid-week).
  - Kampala: prices for some products collected every week for the weekly CPI, but only first week and mid-week prices used for official CPI.
  - "There are more than 7,000 prices which enter the CPI every month."
- Market collection and unit pricing:
  - Large share of prices collected in markets by purchasing the products.
  - Products are weighted at the office and a unit price calculated which then enters the price index, approximating purchasers’ prices after negotiation.
- Outlet and product selection:
  - Selection of outlets and particular products is purposive by the supervisor.
  - Strategy: select a well-sold product that matches item specifications, in cooperation with respondents.
  - If a product is missing for three months, price collector directed toward a replacement product as close as possible; may be in another outlet.
  - Outlet closures replaced by another outlet of the same type in the same area; targeted replacements made if needed.
- Recording, validation, and training:
  - Prices recorded on paper questionnaires containing item name, additional item specifications, outlet details; the price of the previous month is not included to avoid copying.
  - Price collectors encouraged to report events and provide explanations for price changes; non-standardized comments allowed.
  - Data validation techniques detect possible outliers or errors by testing if price changes lie within thresholds; supervisors follow up and may confirm prices with respondents; a second price collection sometimes necessary.
  - Rule: do not edit a reported price unless it has been confirmed that there is an error.
  - Regular training is organized for price collectors at head office; written instructions provided with item specifications and outlet information and frequency of collection.
  - Example product file referenced in Appendix VI.
- Product comparability and quality change:
  - Price data are sufficiently detailed by product and specifications to ensure period-to-period comparisons within selected outlets refer to the same variety.
  - For clothing, suspected volatile price changes may be related to changes in quality.
  - Clothing items described using specifications partly taken from the International Comparison Program (ICP); field identification of quality changes may be difficult and can be overcome through additional training and experience.
  - If items are not comparable, prices should not be directly compared and quality adjustments should be made.

### Statistical techniques: derivation of weighting scheme and benchmarking
- Preliminary weighting scheme:
  - Developed by UBOS using data from the UNHS 2009/10.
  - Analysis carried out for the 10 baskets separately; UNHS data disaggregated into sub-regions representing these baskets.
  - For Kampala, data further disaggregated according to household income level.
  - Two sub-regions left out because no price collection center is located in them.
  - Relationship between Regions, Sub-regions, Price Collection Centers and Baskets explained in Appendix IV.
- Prior adjustments to UNHS data:
  - Out-of-scope goods and services removed.
  - Only monetary consumption of urban households considered.
  - Some UNHS categories in health, education and rent reclassified to match COICOP.
- Observed differences and concerns:
  - Total weight for Kampala relative to total weight of other sub-regions dropped significantly compared to current CPI basket shares.
  - Within Kampala, high income basket had a very high share compared to middle income and low income baskets; criterion for defining high income households to be reviewed.
  - Underreporting in household expenditure surveys noted as serious/common problem; items with social stigma or small purchases may be unreported.
  - Example: comparison with SUT and household final consumption revealed significant underreporting in UNHS for tobacco or beer.
  - Comparing data sources challenged by conceptual differences (e.g., institutional households included in NA but not in UNHS; treatment of income in-kind or gifts; scope differences between national level comparisons and CPI weights that exclude rural households and two sub-regions).
  - Benchmarking against NA household consumption revealed large discrepancy for accommodation services and restaurants; shifts to be further investigated.
- Use of SUT and NA for corrections:
  - Correction coefficients can be established by comparing UNHS with other sources.
  - SUT provides framework to allocate supplies between intermediate or final uses including household final consumption.
  - Some COICOP categories with no expenditures recorded in UNHS (e.g., insurance or financial services) may be included in the rebased CPI if reliable weights can be estimated using NA data.
  - For insurance, weights in CPI should be based on the implicit service charge: gross premiums plus income from investment of insurance reserves less amounts payable to policy holders in settlement of claims.
- Domestic vs national principle and allocation of non-resident spending:
  - When implementing domestic principle, weights computed as sum of resident expenditure (UNHS) + tourist expenditure (TEMS) + informal exports (ICBTS).
  - Conceptual inconsistencies:
    - UNHS includes residents’ expenditures abroad which should be excluded under domestic principle—estimation required.
    - TEMS contains business expenditures that should be excluded.
  - Two parallel weighting schemes recommended in future: national principle (including non-monetary consumption) and domestic principle.
- Regional grouping and coverage gaps:
  - Two sub-regions have no basket; correction method:
    - All sub-regions grouped into four main regions: Central, Eastern, Western, Northern (see Appendix IV).
    - Total expenditure of parts not covered added proportionally to other baskets of the same region.
    - Assumes that for each region, sub-regions included in CPI represent the parts not covered.
  - This approach relies on regional expenditure data rather than population figures and preserves within-basket structure.
  - Although weights refer to whole sub-region, prices are only collected in one urban center per sub-region; assumes price changes same in price collection center as whole sub-region.
- Cut-offs, manual review, and mapping:
  - Cut-off level of one part per thousand will be used for including an EA in the CPI basket.
  - Expenditure of EAs below threshold distributed proportionally to other EAs of same COICOP class or assigned to similar EAs.
  - Cut-off levels defined for each basket separately; implicit national-level cut-offs may be lower.
  - Final exercise concluded by manually reviewing and adjusting results if needed.
  - Weights for different aggregation levels provided in relative terms at different levels.
  - Once weighting scheme validated, list of EAs covered in rebased CPI can be finalized.
  - Within each EA, items selected and item specifications defined; current price data must be mapped into new classification.
  - UBOS already started collecting prices for products not yet covered; prices for new basket in Fort Portal have been collected since 2010, but additional price collection may still be needed.

### Reference period, base prices, and backward revision
- Reference period and base price:
  - Rebased CPI will be retrospectively calculated up to the weight reference period July 2009–June 2010, which will become the new index reference period.
  - A base price will be estimated corresponding to the arithmetic average of the prices observed during the 12 months’ index reference period.
  - Prices must be imputed if not available up to this period.
  - UBOS intends to use prices data from the ICP for backward imputation if needed.
- Index type and methodological implications:
  - Backward revision links weight and price reference periods and enables compilation of a Laspeyres price index.
  - This reverses the previous approach where expenditures of weight reference period were price-updated to a later price reference period to derive a Lowe price index.
  - Recalculation of past figures includes new classification and additional prices collected at new centers or from new products; substitution bias can be evaluated by comparing published and recalculated indices.
  - Extensive use of imputations may impact results and complicate interpretation.
- Publication and presentation:
  - No monthly indices will be published for the 12-month index reference period.
  - The index for the reference period as a whole will be 100 by definition.
  - This removes visibility of the inconsistency described in paragraph 108, but correct estimation of base prices remains necessary to avoid biased results.

### Index compilation, elementary aggregation, and missing prices
- Elementary aggregation:
  - EAs are the lowest level for reliable weight information.
  - Default method at elementary level: unweighted geometric mean of individual price quotes (Jevons price index).
  - Within an EA there may be one or several representative items; representative items implicitly weighted by number of price quotes collected for each.
  - Due to resource intensity of obtaining detailed weighting information, an unweighted approach is favored for transparency.
- Aggregation:
  - Using appropriate weights, EAs should be progressively aggregated within each basket using the Two-stage (Modified) Laspeyres index formula.
  - Aggregation of baskets to national level should be done consistently (see paragraphs 103–105).
- Temporarily missing prices:
  - Temporarily missing prices are imputed for a maximum of three months.
  - Imputation process is bottom-up: use average price change within same EA before moving up classification (within same basket) if no prices available at level below.
  - No indicators produced on number of missing prices per month; CPI staff confirmed their number was low.

*IMF country report content unit cr17296 — extracted chapter text*

### 80. There are no items in the current CPI which are explicitly labelled as strongly

### cr17296 - 80. There are no items in the current CPI which are explicitly labelled as strongly

### Quality adjustment
- In case of replacements, quality adjustments are made to capture only pure price changes.
- The UBOS uses the overlap method in general.
  - To apply the overlap method, the price of the new and the old product must be available during the same month.
  - Linking with overlapping prices makes an implicit adjustment for the difference in quality between the two products, as it assumes that the relative prices of the new and old product reflect their relative qualities.

### Production system
- The CPI is compiled in a range of Excel files.
- UBOS considers that the Excel solution should be further developed before switching to an alternative solution.
- Identified improvements:
  - Automate repetitive tasks.
  - Avoid manually transferring intermediate results from one file to another.
- Extensive modifications are required because a different classification will be used in the rebased CPI.
  - The system should permit the use of different weighting schemes and be flexible to account for future adjustments in the samples.

### Chain linking
- Plan: chain link the rebased CPI with the previous CPI series at the All Items level.
  - The new index reference period July 2009–June 2010 will be used as the overlapping period (annual overlap method).
- In linked series:
  - Price developments during the overlapping period will be consistent with the old CPI.
  - The comparison between July 2010 and the overlapping period on average will be consistent with the rebased CPI.
  - The implicit price change between June 2010 and July 2010 may be flawed.
- UBOS may recalculate and link rebased CPI series at the COICOP division level backwards if there is a user request.

### Serviceability: timeliness, dissemination, and precision
- Frequency and timeliness:
  - The Ugandan CPI is compiled with a monthly frequency and released on the last day of the reference month.
  - This timeliness exceeds both General Data Dissemination Service (GDDS) and Special Data Dissemination Service (SDDS) standards.
- Pre-release communication:
  - Prior to the first release of the rebased CPI, users should be informed in advance about the upcoming change.
  - UBOS plans presentations to CPI consultation committees and to inform the media and the general public.
- Publication breakdown:
  - Indices and weights are broken down to the COICOP class level for food and non-alcoholic beverages and to the COICOP division level for the remaining products.
  - Results will be provided both at the national level and for the 10 baskets separately.
  - In the future, the weight of each basket in the national CPI should also be published.
- Rounding and precision for publication:
  - Monthly and annual changes are compiled using indices with full precision.
  - For publication: weights are rounded to four digits, indices are rounded to two digits and rates of change are rounded to one digit.
- Weight review and updates:
  - UBOS plans to carry out every year a review of the weights and adjust lower levels as needed.
  - New products can be introduced into the samples, and the weights at the lower level re-established on the basis of more recent information.
  - It is recommended to update weights at least every five years.
  - A more frequent complete weight revision should be considered if resources permit, especially as data sources such as the UNHS may be available at a higher frequency.
- Metadata and documentation:
  - With significant methodological changes, GDSS metadata should contain up to date information.
  - The content of the metadata should be reviewed on a regular basis and updated as needed.
  - A comprehensive methodological document should be published in a readily accessible form that should include the website of the UBOS, and updated regularly.

### Training and work program
- Training provided to CPI staff covered:
  - Chain-linking and adjusting weights at lower levels.
  - Advanced Excel tools applicable to CPI production.
- Work Program objective:
  - Release the new CPI in January 2015.
- Selected milestones and dates from the Work Plan:
  - October: Compare the UNHS with other sources and correct for under/over reporting; Validate the regional weights and adjust the Kampala baskets as needed; Finalize the weight derivation; Analyze the data from the MHBS; Compare the new and the old weighting scheme including rationalization of major shifts from the old weighting pattern.
  - November: Define the EA, identify the items within each EA and define the item specifications; Map existing prices data into the new structure; Estimate base prices; Review the prices and outlets in each EA and adjust the EA as needed; Initiate new price collection as needed; Organize training of price section staff about the rebased CPI.
  - December: Finalize and test the compilation system for the rebased CPI; Calculate price indices from the reference period to now; Chain-link the old CPI using the reference period as the overlap period; Pre-announce the upcoming rebasing of the CPI; Run internal discussions on the results.
  - January (2015): Stop price collection for items removed from the rebased CPI; Prepare the dissemination support for the rebased CPI; First release of the rebased CPI January 30.
  - March (2015): Update metadata and prepare documentation of the methodology of the rebased CPI.

### Current CPI series — identified issues and recommendations
- Main concerns discussed with UBOS:
  - Revisions made to published indices after publication.
  - Published inflation rates inconsistent with inflation rates calculated from published indices.
  - Absence of consistency in aggregation.
  - The average of index-reference-period indices does not = 100.
  - Inconsistent sampling, with some items only sampled in one or two areas.

- Revisions:
  - UBOS policy: in case of late entry of data, especially from education and health service providers, revise the CPI figure of the previous month.
  - Delays in price collection are caused by lack of cooperation from respondents.
  - If price collection for these services is incomplete at publication, the press release indicates the provisional status of the results; the following month the press release indicates that the results from the previous month have been revised.
  - UBOS argues revising past results improves CPI quality; revisions are restricted to the previous month and their impacts tend to be marginal.
  - Counterpoint: A CPI is commonly used for indexation purposes in business or legal contracts. Therefore, unlike other statistical indicators, a CPI should not be revised.
  - Strategies to avoid revisions:
    - Consult users about relaxing the tight publication schedule of the official CPI as early estimates are already published with the weekly CPI.
    - Accelerate price collection; make respondents aware of reporting deadlines and implement follow-up procedures.
    - Review samples and replace uncooperative respondents.
    - Systematic oversampling to ensure sufficient number of prices for reliable monthly price changes.
  - UBOS considers the revision policy to be a strategic decision requiring further consultation with stakeholders, especially the Bank of Uganda.

- Inconsistency in aggregation:
  - Current practice:
    - Items are first aggregated separately for each basket.
    - Sub-group, group and All Items indices of the baskets are aggregated into corresponding national indices using the same basket shares.
  - Problem:
    - Aggregation of groups (or sub-groups) at the national level does not equal the All Items CPI obtained by aggregating the All items level of the different baskets.
  - Recommendation:
    - Adjust the aggregation method so that consistent results are obtained independently of the sequence of aggregation.
    - Proposed approach would:
      - Provide identical results for each basket separately;
      - Use the same item weights in each basket and at the national level;
      - Lead to the same national All Items CPI (national indices for groups and sub-groups would differ).
  - UBOS plans to improve higher-level aggregation in the rebased CPI.
  - Appendix III provides a numerical example illustrating inconsistent aggregation and the alternative weighting scheme that yields consistency.

- Base period indices:
  - Requirement: The arithmetic average of the indices of the index reference period has to be equal to 100.
  - Consistency check found this does not hold for the overall CPI.
  - Top-down analysis traced the inconsistency to:
    - Kampala low and middle income basket where the All Items level average was not equal to 100.
    - Within that basket, the group “health, entertainment and other services” had inconsistency.
    - Within that group, the health sub-group had an inconsistent index reference period average.
    - For one item in the health sub-group, the index did not average to 100 due to an erroneous base price different from the arithmetic average of monthly prices observed during fiscal year 2005/06, as required.
  - Recommendation:
    - Make consistency checks by averaging the index reference period indices at all aggregation levels and compile results with sufficient precision.
    - Follow up even minor inconsistencies.

- Regional sampling of items:
  - The basket consists of 276 items.
  - Sampling distribution:
    - 14 percent of items sampled in only one basket.
    - 7 percent of items sampled in only two baskets.
  - Cause: Items entered a basket only if its expenditure was found to be significant in that basket.
  - Rationale: This approach guarantees the accuracy of each basket and facilitates price collection in each area.
  - Items sampled in a limited number of baskets are almost exclusively sampled either only in Kampala (high income and middle and low income baskets) or only in the Northern region (Arua and Gulu baskets).
  - Explanation: Kampala and the Northern region are the most different from the rest of Uganda.

*cr17296 - 80*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2017/cr17296.pdf_