1. Introduction

The Somalia National Bureau of Statistics (SNBS), with technical support from the World Bank, has completed a comprehensive modernization of the Consumer Price Index (CPI) system. This reform represents a major step towards building a harmonized, nationally representative inflation measure, fully aligned with international standards.

Key innovations include:

• Updating the weighting system using data from the Somalia Integrated Household Budget Survey (SIHBS) 2022;
• Transitioning from COICOP 1999 to the latest COICOP 2018 classification.
• Expanding geographic coverage to all Federal Member States (FMS) and Somaliland;
• Strengthening institutional capacity at both national and state levels for data collection, validation, and publication.

This methodological note describes the changes introduced, the processes undertaken, and the implications for the newly released CPI.

2. Background

Prior to the modernization, CPI production in Somalia was fragmented and limited:

• Separate CPIs were compiled for Banaadir, Puntland, and Somaliland using the COICOP 1999 classification systems, and weights drawn from the 2016 Somalia High Frequency Survey for Banaadir and Puntland, and the 2013 Household Income and Expenditure Survey (HIES) for Somaliland.
• In the absence of a unified national CPI, the index for Mogadishu/Banaadir was used as a proxy for the entire country in policy and economic analysis.

Since January 2022, price data collection activities have been extended to all Federal Member States (FMS), with data collected primarily in the capital cities. In Puntland, collection is more extensive and also covers the cities of Bosaso and Galkayo, in addition to Garowe.

However, until the recent reform, these data were not integrated into the national CPI framework. Furthermore, there was no harmonization of the CPI basket structure across States, with varying levels of detail and item coverage. In terms of data collection methods, Puntland and Somaliland continued to use paper-based (PAPI) systems, while the remaining States employed tablet-based Computer-Assisted Personal Interviewing (CAPI) systems.

3. Transition to COICOP 2018

A central feature of the modernization process is the adoption of the Classification of Individual Consumption According to Purpose (COICOP) 2018, which replaces the previously used COICOP 1999 classification system. This transition introduces more detailed consumption categories, including the addition of a fourth level (subclass) that allows for more granular analysis. It also expands the number of divisions from twelve to thirteen by splitting Insurance and Financial Services and Personal Care and Miscellaneous Goods. Furthermore, the classification of goods and services has been improved, particularly in key sectors like health, communication, and recreation.

Implementing this transition required a comprehensive mapping process. A total of 537 items from the Somalia Integrated Household Budget Survey (SIHBS) were classified according to the 9-digit COICOP 2018 structure. Additionally, over 1,300 items in the CPI basket were realigned with the new COICOP codes. Throughout this process, special attention was paid to ensuring consistency across states, allowing the classification framework to fully support both state-level and national aggregation of the CPI.

4. Estimation of CPI Weights based on SIHBS 2022

The updated CPI weights are derived from the SIHBS 2022, the most comprehensive and recent source of household consumption data in Somalia. The survey covered more than 7,200 households across urban, rural, and nomadic areas in all Federal Member States and Somaliland, allowing for reliable state-specific and national expenditure estimates.

Given the heterogeneity of consumption patterns and potential survey reporting issues:

• For food products, unit prices were calculated by combining quantity and expenditure data, with outlier detection based on the regional distribution of unit prices;
• For non-food products, the analysis focused on the distribution of expenditure values across households;
• Conversion to USD was applied using official exchange rates to standardize values across states.

Outliers were not excluded but adjusted by imputing regional median values, ensuring data robustness without discarding plausible household behaviours.

The estimation process for food expenditures relied on data regarding quantities consumed over the previous seven days, as well as information on purchases made within the last thirty days. When purchases had occurred within one week of the interview, expenditures were annualized under the assumption of stable weekly consumption patterns. For purchases made between eight and thirty days prior, the reported expenditures were appropriately scaled to estimate annual spending. In cases where purchase information was outdated or missing, estimations were derived using declared consumption quantities in combination with regional average prices. Overall, the process involved the treatment of about 150,000 individual records, resulting in a detailed and robust estimate of household food expenditures across the country.

For non-food items, the estimation process considered different recall periods, ranging from 7 days to 1 month, 6 months or 12 months, depending on the type of goods or services involved. Annual expenditures were calculated by applying the appropriate extrapolation factors to the reported figures. Durable goods, including vehicles and household appliances, were treated separately, with estimates based on information about asset ownership and acquisition provided by households. Additionally, education expenses, energy consumption, and transport services were carefully annualized using data collected through the specific modules of the SIHBS survey.

As regards imputed rents, the SIHBS questionnaire records actual rents for tenants, as well as self-reported imputed rent for owners and non-market tenants. However, 43.1% (2,265 out of 5,255) of households that report to be non-renters do not provide a self-reported estimated rent, and an additional 4.8% declare “0” as an estimate of rent. Therefore, the housing expenditure component of the consumption aggregate is constructed by annualizing actual rent and computing imputed rent for owners and non-market tenants, based on a hedonic regression model that accounted for key housing characteristics such as location, size, and available facilities.

Following the estimation of household expenditures, SIHBS items were classified according to COICOP 2018 at the 9-digit level (defined coherently with the 5-digit UNSD COICOP codes, six-digit for food items) providing a detailed and standardized framework for categorizing consumption. Each item in the CPI basket was then aligned with these classifications and assigned an 11-digit COICOP code to support accurate aggregation and reporting. In cases where a direct correspondence between SIHBS items and CPI basket entries was not available, the related expenditure shares were redistributed to appropriate higher-level categories within the COICOP structure—such as subclasses, classes, or divisions. This process ensured comprehensive coverage of household consumption in the CPI, even where limitations in the original survey design were present.

Table 1 reports the final set of CPI weights was normalized to sum to 100% and disaggregated by State (Banaadir, Puntland, Galmudug, Hirshabelle, Jubaland, South West, Somaliland) COICOP 2018 Division.

Table 1 Somalia CPI weights by COICOP 2018 division, including imputed rentals. 2022

Note: Somaliland temporarily paused the participation in the CPI innovation process due to local elections and change management in their statistics office

Table 1 shows that Food and Non-Alcoholic Beverages account for approximately 45 percent of total household expenditure at the national level, with variations ranging from 41 percent in Banaadir to over 50 percent in Hirshabelle and Jubaland.

Table 2 presents a comparative overview of the structure of household consumption across two reference periods, highlighting the shift from the older COICOP 1999 classification used in 2016 to the updated COICOP 2018.  It highlights significant increases in the relative importance of divisions such as Clothing and Footwear, Furnishings, and Transport compared to previous weighting structures. Moreover, the updated system includes new divisions such as Insurance and Financial Services, and Personal Care and Social Protection, reflecting broader and more modern consumption categories.

Table 2. Somalia CPI weights by COICOP 2018 and COICOP 1999 division. 2016 and 2022

5. Geographic Coverage

An essential achievement of the innovation project is the expansion of the CPI’s geographic coverage, with price data now being systematically collected across all Federal Member States and Somaliland. In most states, data collection activities are concentrated in the capital city, whereas in Puntland, the exercise extends to three cities—Garowe, Bosaso, and Galkayo—ensuring broader regional representation. Data collection is primarily conducted through Computer-Assisted Personal Interviewing (CAPI) technology, although in Puntland and Somaliland, paper-based methods (PAPI) are still in use, with plans underway to transition to CAPI.

To support the new structure, Data Delivery Agreements (DDAs) have been established between SNBS and the State statistical offices, clearly defining roles, responsibilities, and timelines. This expansion strengthens the ability of the national CPI to accurately capture and reflect the diverse economic conditions across Somalia.

6. Compilation Methodology

The CPI compilation follows a three-stage process:

Stage 1 – Calculation of elementary Price Indices

Elementary indices (at the 9-digit COICOP level) are calculated for each city/state using geometric means of monthly price relatives classified by 11-digit COICOP level codes.

Stage 2 – Aggregation at State and National Level

State-level indices are aggregated using the updated state-specific weights. In all the FMS (with the exception of Puntland), given that the data collection is carried out only in the capital city (that is assumed as representative of the entire State), the elementary aggregate price indices are considered as representative of the State. Only in Puntland, where price collection in carried out in three different cities, to arrive at State level elementary aggregate price indices, it is necessary to aggregate the indices calculated by using the prices collected in three towns (Bossaso, Galkayo, and Garowe): in this case the town level elementary aggregate indices are grouped by weighted arithmetic mean by using the population information as proxy.

National elementary indices are compiled through horizontal weighting, reflecting each State’s contribution to the national expenditure on each item.

Stage 3 – Vertical Aggregation

Higher-level indices (class, group, division) are compiled using Laspeyres-type formulas, respecting the COICOP 2018 hierarchy.

All indices are produced with a fixed base year (2022=100).