The index for shelter, the service that a housing unit provides its occupants, is one of the largest parts of the CPI market basket—the goods and services that American households consume. Owners’ equivalent rent of residences (OER) and rent of primary residence (rent) measure the majority of the change in the shelter cost consumers experience. All three of these series are published each month in each area for which CPI data are published.
The CPI for shelter also includes lodging away from home and tenants’ and household insurance. This factsheet focuses only on rent and OER, however.
The shelter service that a housing unit provides to its occupants is the relevant consumption item for the CPI. Most of the cost of shelter for renter-occupied housing is rent. For an owner-occupied unit, most of the cost of shelter is the implicit rent that owner occupants would have to pay if they were renting their homes, without furnishings or utilities.
Owned housing units themselves are not priced in the CPI Housing Survey. Like most other nations' economic statistics programs, the CPI program views owned housing units as capital (or investment) goods distinct from the shelter service they provide, and therefore not as consumption goods. Spending to purchase and improve houses and other housing units is treated as investment and not consumption in the CPI. Interest costs (such as mortgage interest), property taxes, real estate fees, most maintenance, and all improvement costs are part of the cost of the capital good and are also not treated as consumption items. These non-consumption costs of owned housing are out of scope for the CPI under the cost-of-living framework that guides the index.
The data used as inputs in the construction of the index for shelter, as well as the indexes for rent and OER, are collected in two surveys. The Consumer Expenditure (CE) Survey asks households the share of their budget which goes towards different categories of goods and services, and is subsequently used by the CPI program to create weights for index estimation. The Housing Survey collects price observations of rental housing units across the United States.
The relative importance of an item is its percent of total consumer spending covered by the CPI market basket as of December of the most recent year.
Rent of primary residence
Lodging away from home
Owners’ equivalent rent of residences
Owners’ equivalent rent of primary residence
Unsampled owners’ equivalent rent of secondary residences(1)
Tenants’ and household insurance
The expenditure weight in the CPI market basket for OER is based on the following question that the Consumer Expenditure Survey asks of consumers who own their primary residence:
For more information about the relationship between reported rents and OER, see Reconciling User Costs and Rental Equivalence: Evidence from the U.S. Consumer Expenditure Survey.
The following questions, asked of consumers who rent their primary residence, are the basis of the weight for rent:
From the responses to these questions, the CPI program estimates the total shelter cost to all consumers living in each index area of the urban United States, which is then used to weight the OER index. Note that these responses are not used in estimating price change for the shelter categories, only the weight.
The sample design of the Housing Survey is fairly complex, but essentially housing units are selected in proportion to their share of total spending on shelter. Total spending on shelter includes the total spending on rented and owned housing units, as collected in the CE Survey. Using data from the Decennial Census of Population and Housing, the CPI program defines small geographic areas, called segments, within each of the CPI pricing areas. Segments are one or more Census blocks. The Census provides the number of renter and owner housing units in each segment and the average rent of the renter units in each segment. The missing piece not available from Census is the average owner cost for the owners in the segment. BLS estimates this average implicit rent of the owner units in the segment by modelling Consumer Expenditure Survey renter equivalent values and applying those parameter estimates to each sampled segment, enabling the CPI program to calculate the total spending (rent plus implicit rent) for each segment. The CPI program selects a sample of segments in each pricing area using stratified sampling in proportion to total shelter value. BLS data collectors visit a small number (usually 5) of renter-occupied housing units selected to represent each segment.
The Housing sample design generates a weight for each housing segment in the sample. The segment weights are denoted Ui,α in equations 1 and 2 below. A segment’s weights is its share of the renter and owner spending divided equally among the segment’s sample units. Part of each segment weight represents renter housing and the rest represents owner housing. As noted earlier, owner-occupied units are not priced in the CPI Housing Survey. To make the segment weights specific to the renter universe, a factor, denoted αi,α, which is equal to the renter share of total (owner and renter) in the unit’s segment, is applied to the sampling weight of each unit. To make the weights specific to the owner universe, (1 - αi,α), the complement of the rent factor, is applied to. Thus each segment’s weight is allocated between renter housing and owner housing.
Some units represent only renter units; for these units, αi,α = 1. Others only represent owners, so their αi,α = 0. For example, rental units under rent control do not represent any owner-occupied housing units. In some areas, the sample yields relatively few rental units suitable for measuring owners’ equivalent rent; in these areas the sample has supplemental rental units that represent owners exclusively. In addition, sample rental units in largely owner-occupied neighborhoods mainly represent owner housing, so their αi,α can be near 0.
Owner-occupied housing and rental housing often have a different proportion of detached houses and other structure types. The CPI Housing Survey is designed to be representative of rental housing, so a structure type adjustment factor, Fα in equation 1, is included in OER calculations. It allocates more weight to structure types that the Census's 5-year American Community Survey (ACS) estimates to be more prevalent in the segment's owner-occupied housing than among the segment's sampled units. In a segment, Fα is quotient of detached housing’s estimated share of a segment’s owner-occupied unit and detached housing’s share of responding units for the segment, capped to be no more than 2 or less than 0.5.
The CPI program collects rent data from each sampled unit every 6 months. Many rents change infrequently, being locked in place for a given lease term, and collecting rent data less frequently allows for a larger sample. Most rents included in the sample are continuing rents, and only a minority of observations are rents which have changed since the previous observation period. The CPI program divides each area’s rent sample into six subsamples called panels. The rents for panel 1 are collected in January and July; panel 2, in February and August, etc.
The BLS contacts owner-occupied units only during the Consumer Expenditure Survey, in order to derive item-area expenditure weights. Owner-occupied units are not interviewed in the Housing Survey; the Housing Survey sample contains only rental units. When a rental unit is on panel, CPI data collectors obtain the current rent, what additional services (for example, utilities) are included, and information on any changes to the unit or the rent that has occurred since its previous pricing six months ago.
Using the sample of rental units, the CPI program calculates a measure of price change for each CPI index area for the rent and OER indexes. The first step is standardizing the collected or contract rents, putting them on a monthly basis, and adjusting them for quality changes. These quality changes include adjusting for changes to the structure, such as the number of bedrooms or bathrooms or the type of heating and cooling equipment, or changes in extra charges such as the number of pets. These standardized rents are called normalized rents.
The CPI program also adjusts rents for the effect of aging over time. Rents are collected for the same housing units every 6 months. Consequently, each time the CPI observes the rent of a sample unit, it is 6 months older. To account for this aging, an age-bias factor is applied to the current rent. For additional information about this process, see Updating the Housing Age-Bias Regression Model in the CPI.
For the rent index, the CPI program computes what it calls the economic rent which removes the value of any changes in what is included in the rent between collections. For each sample unit, the CPI program estimates the cost of five types of utilities and makes appropriate adjustments to rents. The types include three energy utilities plus water and sewer service. For water and sewer service, the CPI program has estimates of their average cost in each of the CPI pricing areas, which are updated monthly using the movement of the CPI index for water and sewerage maintenance. For the three energy utilities, the CPI program estimates the quantity (in BTUs) of electricity, natural gas, and fuel oil that each rental unit in the CPI Housing Survey uses. The CPI program then values these quantities using CPI average price data.
To estimate the quantities of electricity, natural gas, and fuel oil, BLS uses data from the Department of Energy’s Residential Energy Consumption Survey (RECS) to estimate formulas for each of the three energy types. The formulas predict a housing unit’s energy consumption from the following characteristics and variables:
Inserting each housing unit’s characteristics into a regression equation for a type of energy yields a predicted quantity of energy consumed.
To value these estimated quantities, the CPI program uses the following average price series: utility (piped) gas per therm, electricity per KWH, and fuel oil #2 per gallon (3.785 liters).
The Housing Survey converts these to price per BTU to make them consistent with the predicted quantities. Each housing unit’s cost for each of the three utilities is its estimated consumption amount times an estimate of its index area’s average utility price. The Housing Survey uses a 12 month moving average of average prices from preceding months (t-1 through t-12) since average prices for the current month are not available.
To calculate price relatives for the OER index, the CPI program calculates what it calls the pure rent from the normalized rent, removing the value of any utilities that might be included in the rent. Owner-occupants pay for their own utilities, and the CPI program accounts for them outside of shelter.
The numerator and the denominator in the formula are weighted averages of the pure rents in month t and t-6. The weights are the sample unit weights Ui,α adjusted by (1 - αi,α), which is the complement of the rent factor. Dividing the average pure rent for the current month (t) by the average pure rent from 6 months earlier (t-6) yields the 6-month relative of price change. The sixth root of this relative is the 1-month change in OER for index area α.
The numerator and the denominator in the formula are weighted averages of the economic rents in month t and t-6. The weights are the sample unit weights Ui,α adjusted by the rent factor αi,α. Dividing the average economic rent for the current month (t) by the average economic rent from 6 months earlier (t-6) yields the 6-month relative of price change. The sixth root of this relative is the 1-month change in rent for index area α.
Construction of the elementary indexes for each area from the monthly relatives of price change is a straight–forward chaining, as shown in equation 3.
The index for period t is computed by multiplying the index for period t-1 by the relative of price change from t-1 to t.
The final step is to combine the elementary indexes into indexes for aggregates, such as the U.S. city average OER index and U.S. city average rent index, using the biennial aggregation weights described above. The CPI program uses the same method for all its aggregate indexes.
Sometimes the CPI program cannot collect data for a sample housing unit. When this occurs, the CPI program imputes price change for the noncollected units. After calculating the current month indexes, the CPI program calculates an economic rent and a pure rent for them for use in the future. These imputed values are the previous month t-6 values multiplied by the relative of price change for the current month t. This has the effect of assuming that the missing observations had the same movement as the nonmissing ones during the period they were missing; the CPI will reflect any additional price movement for them when data collection for these housing units resumes.
The CPI program assumes that sample units reported to be vacant are transitioning to new tenants. Experience with rent data has shown that units tend to experience rent change when the tenant changes. To avoid missing this rent change for vacant units, the CPI program performs a special class-mean imputation to estimate their rents. The estimated current rent of a vacant unit is its previous rent times the average rent change of newly occupied units. For example, if the average rent increase for new tenants in an area was 5 percent, a currently vacant unit that rented 6 months earlier for $1,000 would have an estimated current rent of $1,050. In effect, this assumes that the rent of the vacant unit is increasing at the same rate as occupied units in the same area, the CPI will reflect any additional price movement for these units when they are no longer vacant and data collection resumes.
Last Modified Date: September 8, 2023