Estimates of imputed rent

Note: this paper was formerly known as 6525.0 - Experimental Estimates of Imputed Rent, Australia, 2013-14.

This information paper presents a detailed explanation of the final methodologies implemented for estimates of imputed rent in the 2015–16 Survey of Income and Housing (SIH) and Household Expenditure Survey (HES). These include imputed rent estimates for owner-occupied dwellings and the imputed benefit to tenants not paying market rents.

The ABS first released experimental household level estimates of imputed rent for owner-occupied dwellings and other tenure types not paying market rents in May 2008, derived from data reported in the 2003–04 and 2005–06 Surveys of Income and Housing (SIH). The same methodologies were used for each subsequent SIH up to 2011–12. For the 2013–14 SIH, new experimental methodologies for estimating gross imputed rent were developed to address limitations of the previous methodologies. For the 2015–16 SIH, further refinements to the experimental methodologies were applied.

Estimates of imputed rent for 2003–04 to 2015–16 are presented in Household Income and Wealth, Australia, 2015–16 (cat. no. 6523.0). Estimates between 2003–04 and 2013–14 have been revised to be consistent with the new method outlined in this paper.

Household income statistics compiled for individual households are critical to analysis and modelling that supports understanding the socio-economic circumstances of different household types. They are also important in developing and evaluating policies on income support, income distribution and income taxation.

The ABS regularly collects detailed information on household income, expenditure and wealth in SIH and HES. SIH is conducted every two years, with the latest results published with respect to 2015–16. The HES is conducted every six years, with the latest survey relating to 2015–16. The SIH and HES were conducted on an integrated basis in 2003–04, 2009–10 and 2015–16.

The ABS releases summary statistics on household income in Household Income and Wealth, Australia (cat. no. 6523.0). Microdata from the surveys, such as the Confidentialised Unit Record Files (CURFs) and access to files via the DataLab (restricted to authorised users), are also available to users to support comprehensive and detailed analyses.

The most restricted concept of income used in income analysis is the gross private income of individuals. While this measure is useful for certain purposes, it is generally of limited use when trying to understand people’s broader economic wellbeing. Published ABS household income analysis, in accordance with international statistical standards, extends this income measure by:

• adding government transfers (both cash and in kind),
• deducting direct and indirect taxes,
• equivalising household income to adjust for household size, and
• adding an imputed rent value for the net benefits of home ownership and subsidised rent payments.
   

Imputed rent is included in both household income and expenditure. This conceptually treats owner-occupiers as if they were renting their home from themselves, thus simultaneously incurring rental expenditure and earning rental income.

Imputed rent is included in income on a net basis i.e. the imputed value of the services received less the value of the housing costs generally incurred by the household in their role as a landlord. Gross imputed rent is added to the expenditure of owner-occupiers, and any housing costs generally borne by a landlord are deducted (e.g. general and water rates, building insurance).

Including imputed rent in income for people living in different tenure types enables more meaningful analysis to be undertaken on their economic wellbeing and enables better analysis of changes over time in income levels and income distribution if tenures change. International standards for household income and expenditure statistics also recommend the inclusion of imputed rent to support different types of analyses.

Different methodologies are used to produce household level estimates of gross and net imputed rent for:

• owner-occupied dwellings, and
• other tenure types.
   

This section provides a step by step explanation of how household level estimates of gross and net imputed rent have been created for these tenure types.

The market value of the rental equivalent for owner-occupied dwellings can be estimated in a number of ways (e.g. self-report, stratification and regression approaches). The statistical office of the European Union, Eurostat, has reviewed rental equivalence methods and recommended regression or stratification techniques in countries where representative market rates can be obtained (Eurostat 2006). Australia has a well-established private rental market, and this data is the basis for both the previous and new methodologies. A non-parametric stratification technique is used to estimate the market value of the rental equivalent in the new methodology for owner-occupied dwellings. The stratification method replaces the previously used hedonic regression method. A stratification method was first adopted for the 2013-14 Survey of Income and Housing (SIH) outputs and has undergone further refinement for the 2015–16 SIH outputs to improve these methodologies (see the 'Changes to methodology over time' section).

The net imputed rent for owner-occupied dwellings has been estimated as:

• the market value of the rental equivalent (referred to as gross imputed rent); less
• the housing costs normally paid by landlords i.e. general rates, water and sewerage rates, mortgage interest, building insurance, repairs and maintenance.
   

The imputation of gross and net imputed rent includes other housing tenures in order to value the in-kind benefit conferred to households paying subsidised rent (e.g. tenants of an employer or of a state/territory housing authority) and households occupying their dwelling rent-free. The gross imputed rent for these housing tenures has been estimated using the same stratified rental data from the Census of Population and Housing as used for owner-occupiers. In deriving net imputed rent for these tenure types, actual rent paid is the major housing cost deducted.

The methodology for estimating gross imputed rent for individual owner-occupied dwellings uses the relationship between dwelling price (available from Valuers General data) and the market rent that a dwelling would receive (available from Census data).

By merging aggregated Valuers General (VG) data and Census dwelling data for the smallest available geographic area, rental yields based on average dwelling prices and reported rents have been calculated for each rental dwelling in the Census. These rental yields (termed base rental yields) underpin the new methodology.

The methodology takes account of differences in the value of dwellings and market rents due to location, type of dwelling and Socio-Economic Indexes for Areas Index of Relative Socio-Economic Advantage and Disadvantage (SEIFA IRSAD), to produce an estimate of a market rent for each owner-occupied dwelling in the SIH.

Appendix 2 summarises the detailed steps described below to produce gross imputed rent estimates for owner-occupied dwellings.

Census data was used to provide market rents paid by households by type of dwelling for each SA1 in Australia in scope of the SIH.

VG data was used to provide an average (mean) dwelling price for each SA1 in Census 2011. To ensure there were sufficient dwelling sales to produce reliable estimates, VG dwelling sales prices for the financial year before and after the Census were used i.e. 2010–11 and 2011–12 for the 2011 Census. To ensure the dwelling sales prices were representative of the region, an average price was only calculated if there were at least five dwelling sales in the SA1 in the two year period. For the 2011 Census, approximately 16% of SA1’s were excluded due to an insufficient number of dwelling sales. This has negligible impact due to the broader stratifications applied to the data, as explained in Step 1.2 below.

To produce historical imputed rent estimates using the new methodology, CD Census 2006 data and VG dwelling sales data from 2005–06 and 2006–07 were also used. Estimates for 2015–16 will be reviewed based on 2016 Census data and 2015–16 and 2016-17 VG dwelling sales data in due course.

The following three steps were used to calculate a base rental yield to produce gross imputed rent estimates for individual dwellings in the SIH.

The first step was to pool VG dwelling sales and Census rental data. A preliminary rental yield was allocated to each Census rental dwelling by dividing its reported rent by the mean VG dwelling price from sales in the SA1 or CD where the Census dwelling was located. If no mean dwelling price was available because there were less than five dwelling sales in the relevant SA1/CD, the rental records for that region were excluded.

Census rental records, with the preliminary rental yield added to each record, were then stratified using the following variables:

• State/territory (eight categories)
• Section of state (four categories)
• Dwelling type (up to two categories)
• SEIFA IRSAD quintiles (five categories).
   

The final strata were determined based on analysis that there were sufficient dwellings to support the estimates. Strata created for each of the states and territories can be found in the 'Data access' section of this product.

Table 2 in the 'Data access' section shows the 25 strata created in each of the six states. Five categories based on different combinations of section of state and dwelling type were created for each of the five SEIFA IRSAD quintiles.

The territories have fewer strata. In the Northern Territory there were 15 strata because section of state was reduced to two categories (major/other urban and bounded locality/ rural balance) (see Table 3 in 'Data access'). In the Australian Capital Territory, there were 10 strata because section of state was not used (see Table 4 in 'Data access').

The total number of strata for the entire SIH population was therefore 175.

The third step was to create a final base rental yield for each of the 175 strata. The preliminary rental yields for the Census rental records in each stratum (from Step 1.2), were ranked from highest to lowest. The final base rental yield for each stratum was the median preliminary rental yield for that stratum.

To ensure rental yields were representative for each stratum a median was only calculated if there were at least five Census rental records in the stratum. If there were insufficient records, the rental yield was imputed using a rental yield from an adjacent SEIFA IRSAD quintile with the same dwelling characteristics.

Base SIH cycles are those enumerated at around the same time as the Census i.e. SIH 2005–06 (2006 Census) and SIH 2011–12 (2011 Census).

As the SIH contains all of the stratification variables listed in Step 1.2, each owner-occupied dwelling in the sample can be matched to a unique stratum. The gross imputed rent for each owner-occupied dwelling in the base SIH cycles was calculated as the final base rental yield for the relevant stratum of the dwelling, multiplied by the estimated sale price of the dwelling reported in the SIH.

Dwellings with an extremely low estimated sale price reported in the SIH, resulted in an unreasonably low estimate of gross imputed rent. A minimum value was therefore applied, equal to the market rent at the top of the first percentile for the relevant state in the Census. This ensured the gross imputed rent estimate reflected reasonable market costs associated with renting a property.

The relationship between rents and dwelling prices can vary over time as prices change at different rates. To reflect the relative differences between the change in dwelling prices compared to rents, an adjustment factor was created using data reported in the intercensal SIH cycles for which gross imputed rent was to be calculated (SIHs 2003–04, 2007–08, 2009–10 and 2013–14). An intercensal adjustment factor was also used for SIH 2015–16 due to updated VG data not being available at the time of compilation (this will be updated in due course).

As there is significant variation in the change in rents and house prices between states and territories, separate adjustment factors were created for greater capital city area and rest of state for each of the six states. Adjustment factors were created for each territory at the territory level. Appendix 2 and the 'Changes to methodology over time' chapter provide further analysis underpinning the development of the intercensal adjustment factors.

There were three steps to produce the intercensal rental yields.

For each SIH from 2003–04 to 2015-16, a mean rental yield was calculated for each state/territory as the mean rent reported by market renters divided by the mean value of owner-occupied dwellings.

A rental yield adjustment factor was calculated for each of the non-base SIH cycles as its mean rental yield divided by the mean rental yield for the relevant base SIH cycle (2005–06 or 2011–12) (from Step 2.1).

SIH 2005–06 is the base cycle for SIHs 2003–04, 2007–08 and 2009–10. For SIH 2013–14 and 2015–16, the base cycle was SIH 2011–12. Updated VG data was not available at the time of SIH 2015–16 compilation therefore the SIH 2011–12 base cycle was used. It is expected that new base rental yields will be implemented for SIH 2017–18 to include updated Census and VG data.

An adjusted rental yield was calculated for each stratum in each intercensal SIH, as the final base rental yield of the stratum (from the process outlined in step 1) multiplied by the rental yield adjustment factor (step 2.2) for the relevant state/territory of the stratum.

Gross imputed rent for the intercensal SIH cycles was calculated for each owner-occupied dwelling as the adjusted rental yield (Step 2.3) for the relevant stratum of the dwelling, multiplied by the estimated sale price of the dwelling reported in the SIH.

To calculate the net imputed rent for owner-occupied dwellings, the following housing costs normally paid by landlords were subtracted from the gross imputed rent:

• body corporate payments
• general and water rates
• the interest component of repayments of loans that were obtained for the purposes of purchasing or building the dwelling
• house insurance, and
• repair and maintenance costs.
   

All housing costs were net of refunds or subsidies received from outside the household.

All of the relevant housing costs are collected in the SIH except for expenditure information on house insurance and repairs and maintenance. Household Expenditure Survey (HES) 2003–04, 2009–10 and 2015–16 data were used to estimate these expenditures.

Average repair and maintenance costs were calculated for owner-occupiers, stratified by the number of bedrooms. The relevant average expenditure was allocated to each owner-occupied dwelling in the SIH. In non-HES years, these costs were extrapolated using the published ABS Consumer Price Index for 'House repairs and maintenance'.

Stratification by number of bedrooms was also used to calculate an average cost of house insurance using HES data. Average house insurance costs for the relevant number of bedrooms were allocated to all owner-occupiers in the SIH based on HES data. Where house insurance costs were combined with home contents insurance costs, a factor was applied to the total expenditure to estimate the amount for excluding home contents insurance costs. In non-HES years, expenditure was estimated by inflating the most recent HES data using the published Consumer Price Index for 'Insurance services'.

Methodologies for calculating gross imputed rent and net imputed rent for other tenure types remain unchanged. The methodologies implemented in SIH 2013–14 were used in SIH 2015–16 and will continue to be used in future cycles. The current methodologies still in use are detailed below.

Some renters do not pay a market rent, effectively receiving a subsidy for their living costs. Typically, the subsidised rent is made available by government state and territory housing authorities, employers or a family or friend, collectively termed, ‘Other tenure types’.

The value of this subsidy can be estimated by calculating the gross imputed rent (i.e. market rent) for the property and then deducting the actual rent paid by the tenant (reported in the SIH).

The methodology for calculating gross imputed rent for owner-occupied dwellings is not suitable for other tenure types as there is no estimate in the SIH of the value of these dwellings. Therefore, a different approach has been developed that uses the strata developed for imputing the rent for owner-occupied dwellings.

Appendix 2 summarises the steps described below to produce gross imputed rent estimates for other tenure types.

An imputed market rent value was calculated for other tenure types using the median Census rent for the relevant stratum of the subsidised rental dwelling.

For intercensal SIHs, the estimates have been indexed to account for changes in rent over time. The indexation method is similar to that used for owner-occupied dwellings i.e. the percentage difference between the mean rent from the SIH conducted at the time of the last Census and the mean rent for the SIH cycle in question.

Not all households identified in the SIH as potentially living in a subsidised rental dwelling actually receive any discount on their rent. Therefore, if the estimated gross imputed rent was lower than the actual reported rent, the reported rent was used for the gross imputed rent estimate.

This methodology has been applied to SIH 2013–14 and SIH 2015–16 and will be used for subsequent surveys. For all cycles up to SIH 2011–12, the gross imputed rent for other tenure types remains unchanged from previously published estimates that used the hedonic regression model. For these other tenure types, the current methodology has minimal impact on the estimates (2% in 2011–12).

For other housing tenure types, the housing costs subtracted from gross imputed rent to derive net imputed rent are outlined in Table 1.

For each of the housing tenures described in Table 1, any refunds or subsidies received for rent payments were implicitly accounted for in the estimation of net imputed rent. For consistency across all housing tenures, the reported values of any rental refunds or subsidies received by private market renters have been included in the estimates of net imputed rent.

For tenants of state/territory housing authorities, the mean difference between the initial gross imputed rent estimates and the reported rent paid were compared with the mean weekly rental subsidy published in the Report on Government Services (RoGS) for each state. Where initial net imputed rent estimates underestimate the mean benefit for public tenants, net imputed rents are benchmarked to the RoGS published state mean weekly rental subsidies using a multiplicative adjustment.

The ABS developed and implemented new experimental methodologies of gross imputed rent for owner-occupied dwellings and other tenure types for the Survey of Income and Housing (SIH) 2013–14.

Following peer review and coherence investigations with National Accounts imputed rent estimates and Census data, further refinements have been made to the gross imputed rent methodology for owner-occupied dwellings. Methodologies have now been finalised and implemented for the 2015–16 SIH and Household Expenditure Survey (HES).

Gross imputed rent methodology for other tenure types remains unchanged and is consistent with the experimental methodology implemented in SIH 2013–14.

Investigations suggested that an upward bias may be present in the experimental estimates for owner-occupied dwellings, particularly for dwellings with four or more bedrooms. For strata of a similar location and dwelling type (e.g. separate houses in Sydney), experimental rental yields were observed to increase with the number of bedrooms. This observation does not align with the general expectation of higher rental yields for lower value dwellings, and vice versa. Investigations also suggested that imputed rents were slightly underestimated for one and two bedroom dwellings.

For example, the rental yield for a two bedroom house in a given region was 0.00059113 under the experimental method. For houses with additional bedrooms in the same areas, yields increased to 0.00089476 for a five bedroom house. Continuing on with this example, a two bedroom house valued at $500,000 and a five bedroom house valued at $900,000 would receive $296 and $805 in gross imputed rent respectively. Note that the latter amount is likely to exceed market rents for a five bedroom in the same area. Additionally, depending on market conditions, rent for a two bedroom house could easily exceed the $296 calculated.

The inclusion of number of bedrooms as a stratification variable was the main cause for this trend. Within the experimental method strata, the composition of rental properties compared to owner-occupied dwellings appeared very different. There are far more owner-occupied properties that were separate dwellings with three or more bedrooms when compared to rental properties (particularly in higher SEIFA quintile areas). Therefore in calculating final base rental yields, the rent paid for rental dwellings (the numerator) tended to be for a smaller number of bedrooms in each strata, while dwelling value based on the Valuers General (VG) sales data (the denominator) related to all dwellings sold in the relevant SA1. As a result, the experimental estimates are likely to have overestimated the rental yields for owner-occupied dwellings, such as in the example above, particularly those with four or more bedrooms, and underestimate the rental yield for one or two bedroom units.

Removing the number of bedrooms as a stratification variable, reduces cases of very high imputed rents for dwellings with a higher number of bedrooms and increases imputed rents for dwellings with fewer bedrooms. Using the same houses from the example above, the rental yield for both dwellings is 0.0007598 under the revised method, which for our two and five bedroom houses would estimate gross imputed rents of $380 and $684 respectively. While the latter imputed rent value may be higher than current market rents for a similar property for the area (based on type and number of bedrooms), it is still plausible. It is often difficult to find a rental market equivalent which reflects the additional features of owner-occupied dwellings. Aspects such as solar panels, insulation and improved heating/cooling and other appliances more prevalent in owner-occupied dwellings can contribute to higher gross imputed rents for these dwellings compared to market alternatives.

This limitation may be addressed if the VG dataset provided dwelling characteristics information such as number of bedrooms or dwelling type, which are currently not available on the dataset. The review did not identify the absence of dwelling type from the VG dataset as a significant limitation. Investigations into Census data revealed that most strata were relatively homogenous for type of dwelling. Similarities between yields for semi-detached dwellings and flats, units and apartments were noted.

The final methodologies remove the number of bedrooms as a stratification variable, allowing type of dwelling and rental yield for each region to determine the gross imputed rent for the dwelling. Strata for semi-detached dwellings were also combined with flats, units and apartments due to very similar rental yields.

The overall impact of these changes served to reduce derived rental yield for dwellings in major urban areas with four or more bedrooms, and increase the yield for one and two bedroom units, apartments and semi-detached dwellings.

The distribution of gross imputed rent has remained mostly unchanged between the experimental and final methodologies. Graph 1 below outlines the change in the distribution for owner-occupied dwellings between the experimental methodology implemented in SIH 2013–14 and the finalised methodology, using data for 2013–14 as a comparison. Graph 1 shows that the new final methodology has removed the upward bias for larger dwellings (higher end of the distribution) and the downward bias for one bedroom units (lower end of the distribution).

As outlined in the Final Methodology section of this publication, an adjustment factor is used to improve the estimation of changes in gross imputed rent over time. The relative change in mean market rents and the mean value of owner-occupied dwellings between each SIH cycle are used for the adjustment factor. The experimental method adjusted estimates at a state level only. Further analysis revealed that the survey data can support an adjustment that includes a Capital City and Rest of state split for the six states. The final methodology implements this change to intercensal adjustment factors to better account for different patterns of change in house prices and rents across regions.

The revised adjustment factors were validated by comparing the gross imputed rent estimates for the SIH 2011–12 when they were calculated using 2011 Census data and those calculated using Census 2006 data extrapolated forward. As shown in Graph 2, the two results for the 2011–12 period remain consistent in the current method.

For the Survey of Income and Housing (SIH) 2015–16 imputed rent estimates have been produced using only the new final methodologies.

Revised household level historical estimates for each SIH between 2003–04 and 2013–14 have also been produced. There are three ways by which data analysts can access these new estimates of imputed rent for owner-occupied dwellings.

Imputed rent estimates for each Survey of Income and Housing (SIH) and Household Expenditure Survey (HES) between 2003–04 and 2013–14 using the methodology outlined in this paper are available in following ways:

• Aggregate estimates using the new methodology have been published in Data cube 13 - Imputed Rent, available from the Data downloads section of Household Income and Wealth, Australia, 2015–16 (cat. no. 6523.0).
• Estimates are available via customised data requests from the ABS for each SIH and HES between 2003–04 and 2013–14. This is a charged service.
• The ABS publishes Confidentialised Unit Record Files (CURFs) for each SIH and HES. For more information on the SIH and HES CURF files see Microdata: Income and Housing, Australia 2013–14 (cat. no. 6541.0.30.001) and Microdata: Household Expenditure, Income and Housing 2015–16 (cat. no. 6540.0).
• SIH and HES data may also be accessed via the DataLab.
   

For each SIH and HES from 2003–04 to 2013–14, CURF and DataLab users can derive a modified estimate of gross imputed rent for individual owner-occupied dwellings using the methodology outlined in this paper.

A data cube is available under the Data downloads section of this publication that provides rental yields for strata created only using data items available on the relevant CURFs. It also provides a minimum market rent estimate for each state for each SIH year.

For detailed microdata files, for each of the six Australian states, 25 strata were used (Table 2) using SEIFA IRSAD, Section of State and dwelling type. For the two Australian territories, the detailed Section of State classification was not used. Two groupings were used for the Northern Territory resulting in 15 strata (Table 3). Only one grouping was used for the Australian Capital Territory resulting in 10 strata (Table 4).

Detailed geographical items are not included on historical CURFs, therefore gross imputed rent estimates can only be calculated using available data items. To calculate gross imputed rent estimates for owner-occupied dwellings using the the current method for historical SIHs, CURF users must use the stratification variables available (outlined in Tables 5 and 6 below) from the relevant CURF to identify the rental yield, which is published in the data cube in the Data downloads section of this product.

The gross imputed rent estimate is calculated as the estimated sale price of the individual owner-occupied dwelling multiplied by the rental yield for that SIH cycle. Estimates produced that are below the state minimum should be imputed with the minimum market rent for the relevant state/territory (also available in the data cube described above).

For example, to calculate CURF estimates of gross imputed rent for separate houses in Sydney in 2003–04 using the final methodology, users can:

• stratify records by:
  • state/territory of usual residence - to get New South Wales (STATEHBC)
  • area of usual residence - to get Capital City (AREAHCF)
  • dwelling structure - to get Separate House (DWELTCF)
• locate the rental yield for Sydney in the data cube published in the Data downloads section of this product (i.e. 0.0005493)
• multiply the Estimated sale price of dwelling item (HVALUECH) by rental yield provided to get the gross imputed rent estimates.
   

Gross imputed rent estimates for CURFs have been aligned with estimates derived from more detailed microdata files for the relevant state and territory in each SIH by adjusting the rental yields for each stratum by the amount necessary to achieve alignment. This adjustment helps minimise the effect of using limited data items on the CURF to stratify and calculate the estimates, when compared to those calculated from the detailed microdata files. The resulting difference between estimates produced using detailed microdata and CURF microdata is small (less than 10% difference).

The housing costs deducted from gross imputed rent to calculate net imputed rent estimates for owner-occupied dwellings have not changed. Therefore, revised estimates of net imputed rent using the new methodology can be calculated by deducting the original housing costs for each dwelling from the revised gross imputed rent estimate for that dwelling.

Gross and net imputed rent estimates for other tenure types will not be revised for SIH 2003–04 to 2011–12 because the impact of the new final methodology is minimal on these estimates e.g. 2% in 2011–12. Methodologies implemented in 2013–14 for other tenure types remain unchanged and are consistent with the methodologies used in SIH 2015–16.

This release previously used catalogue number 6525.0.