Housing Tenure, Energy Consumption and the Split-Incentive Issue in Australia

Abstract

In recent years, there has been growing global recognition of the need to reduce carbon emissions in response to climate change concerns. It is generally acknowledged that the energy efficiency of existing homes can be improved, but there are significant barriers to its uptake. In particular, improving the energy efficiency of private rental housing presents unique policy challenges due to a split-incentive problem. This has prompted some governments to introduce programmes that encourage landlords to improve the energy efficiency of their properties. While landlords are responsible for the purchase of many energy-consuming household appliances, tenants are responsible for energy bills. Since the landlord does not reap the immediate benefits of investment in energy-efficient equipment, the incentives motivating such investment are weaker than for homeowners. This paper aims to quantify the magnitude of the split-incentive problem in the Australian private rental housing market by invoking a modelling approach where energy expenditure is estimated as a function of housing tenure, dwelling type, location, climate and other socio-demographic variables. We find no evidence in support of the split-incentive hypothesis in Australia. The paper concludes that differences in housing policy arrangements could be critical to the presence and importance of split incentives.

Key Words:

• Housing tenure
• energy consumption
• split incentives
• principal-agent problem

Acknowledgement

The research reported in this paper was funded by grant number 40560 from the Australian Housing and Urban Research Institute (AHURI). The authors would like to thank Michelle Gabriel, Phillipa Watson and Maryann Wulff for their comments on the paper's research. The authors would also like to thank Richard Seymour, who coded up the 2006 rules governing eligibility for energy rebates in Australia, which has enabled us to distinguish between households that are eligible and ineligible for these rebates. We are grateful to two anonymous referees for helpful comments on an earlier version of this paper. This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and is managed by the Melbourne Institute of Applied Economic and Social Research (MIAESR). The findings and views reported in this paper, however, are those of the authors and should not be attributed to either FaHCSIA or the MIAESR.

Notes

1. Other end-use sectors include the commercial, industrial and transportation sectors.

2. Changes in relative prices are unlikely to immediately prompt changes in space and water heating systems because of the capital costs.

3. About 62 per cent of residents in nonprivate or mobile homes don't actually report their energy expenditure.

4. Over half are living in rented property or are living rent free, and most (66 per cent) live in detached housing.

5. Group households (unrelated persons who reside together in the same dwelling) are retained as we are interested in overall household energy consumption, that is the total energy consumption of all persons living in the same dwelling regardless of whether they are related or not.

6. The household member in charge of paying household bills is simply asked to provide his/her best estimate of the annual amount spent on electricity, gas and other heating fuel bills such as firewood and heating oil, without specific reference to whether the expenditure amount should relate to the primary home. There are 1,234 households that own properties in addition to their primary place of residence.

7. Estimates are available from the authors on request.

8. The findings are not sensitive to choice of reference group.

9. At the time this study was completed, cross-section modelling was the only option, but subsequent waves of the HILDA Survey have since been released that contain the energy expenditure variable. Hence, longitudinal analysis is a potential future direction for research. Energy expenditure variables were first made available in the HILDA Survey in its 5th wave (2005), and have since been reported in every subsequent wave up to the most recent wave 10.

10. Another study sharing the same limitation and using a region control include Rehdanz (2007). A report by the Australian Competition and Consumer Commission (2007) found that electricity prices ranged from 9.56 cents/kWh in the state of New South Wales to 16.04 cents/kWh in the Northern Territory in 2003–2004.

11. Each state government has energy rebate programmes that offer a fixed dollar rebate to eligible households in all states other than Victoria, where rebates are tied to the dollar value of energy bills. Details on energy rebates from State and Commonwealth government websites indicate that most energy rebates are for the use of electricity. The fixed rebate varies from a low of $118 for households without dependent children in Western Australia to a high of $365 in Northern Territory in 2009. In Victoria and the Australian Capital Territory, the value of the rebates varies with respect to whether the concession is received in summer or winter. For example, in the Australian Capital Territory, the summer rate of concession is 25.15 cents per day while the winter rate is 92.2 cents per day. The Commonwealth Government also offers a Utilities Allowance for energy, rates, water and sewerage costs. This is an amount that does not vary with the size of bills or household type and that eligibility is restricted to some income support payment recipients. Typically households eligible for energy rebates are low-income concession card holders such as holders of government-issued healthcare cards, pensioner concession cards, and Department of Veterans’ Affairs concession cards for war widows and incapacitated war veterans. Households that are eligible for these concession cards have to first qualify for receipt of a government benefit. For example, to be eligible for the government-issued healthcare cards, households have to first be in receipt of one of the following government benefits: Newstart Allowance, Youth Allowance, Partner Allowance, Special Benefit, Widow Allowance, Parenting Payment, or the maximum rate of a family payment called Family Tax Benefit Part A.

12. AHURI-3M, a microsimulation model of the Australian Housing Market that contains a tax-benefit module, has been employed to identify households eligible for state government energy rebates. We apply government energy rebate parameters from the 2009 (the latest information available from State and Commonwealth government websites at the time the measurement of energy rebates was conducted) to ascertain rebate eligibility for households within our sample.

13. The ARIA measures remoteness of a point based on the physical road distance between that point and the nearest urban centre. For example, major cities comprise collection districts (the smallest spatial unit used by the Australian Bureau of Statistics when collecting Census data) with an ARIA index of 0 to 0.2, while inner regions comprise collection districts with an average ARIA index greater than 0.2 but less than or equal to 2.4 (for further details, refer to Australian Bureau of Statistics, 2001).

14. BoM also calculates HDDs and CDDs using alternative temperature thresholds of 120°C and 180°C respectively.

15. The energy bills for communal areas in multi-family housing (apartment blocks) could be met through strata fees that are not reported by owners and tenants as part of their annual energy expenditure. A strata fee is an amount payable by residents for the maintenance of an area shared by two or more properties that are linked together. For example, apartment residents tend to share common areas such as lifts, corridors or gardens. The reported expenditures of residents of flats, apartments and units may then under-estimate energy consumption in these dwellings.

16. We are grateful to an anonymous referee for pointing this out.

17. Where a continuous variable is negative or zero, the log of the variable is set equal to zero as it is not possible to take the log of negative or zero values. Under the Box–Cox transformation procedure, the energy expenditure variable y is transformed to (yλ – 1)/λ. Likelihood ratio tests are employed to test if y should appear in linear (λ = 1) or log (λ = 0) form (Kennedy, 2008). The likelihood ratio test statistic for the null hypothesis that λ = 0 is 120.69 and permits rejection of the null at the 1 per cent level of significance. The likelihood ratio test statistic for the null hypothesis that λ = 1 is 4,098.1 and also permits rejection of the null at the 1 per cent level of significance. A linear model using levels of the continuous variables (including the dependent) yields (in the all household model) an R² = 0.147. On computing the antilog of the predicted values from a log-linear model a very similar R² = 0.145 is obtained when regressing actual expenditure values and these predicted values.

18. Residents of Hobart are not separately identified and so all Tasmanian residents are included in the sample.

19. Unless stated otherwise all differences in energy expenditures reported in this section are statistically significant at 5%.

20. However, this difference in energy expenditure is only statistically significant at the 10 per cent level.

21. Households with energy bills less than $100 per annum have been omitted (see ‘Method’ section). The exclusion does not affect the key implications of our modelling. On re-estimating the regression model with these households added back into the sample, the homeowner coefficient increases from 0.143 in the restricted sample to 0.177 in the expanded sample (and again significant at the 1 per cent level). A Chi² test cannot reject the null (at the 5 per cent level) that these coefficients are equal.

22. The percentage impact estimates for binary variables are calculated from (e^β – 1), where β is the estimated coefficient (see Halvorsen & Palmquist, 1980).

23. Our estimate is at the low end of this range. Nesbakken (1999) found that energy consumption was more price sensitive among high-income than low-income households; the absence of a satisfactory price variable could depress the income elasticity estimate.

24. In the all household estimates the homeowner coefficient is 0.142 and is significant at the 1 per cent level. When estimated on a restricted sample that omits those eligible for energy rebates, the homeowner coefficient is 0.160 and is again significant at the 1 per cent level. A Chi² test is unable to reject the null hypothesis of homeowner coefficient equality (a test statistic of 0.15 is highly insignificant).

25. Linear models with the addition of interaction variables involving property type, education and number of bedrooms have also been estimated, but results fail to detect the presence of split-incentive effects among any of the owner subgroups identified by the interaction variables. Results are available from the authors upon request.

26. At the mean age of 50.3 years energy expenditure is given by 353 – 4.04*50.3 = 150. The age at which the interaction effect offsets the impact of the binary owner variable is found by solving for age in the equality 353 – 4.04*age = 0.

27. According to the Australian 2006 Census, persons aged 87 years or over made up 1.1 per cent of the Australian population in 2006 (Australian Bureau of Statistics, 2007).

28. Such rent premiums could also be capitalised into house prices.

29. This mobility of properties can also have negative impacts on tenants’ wellbeing, particularly ontological security, but this is outside the scope of this study.