SAMPLE DESIGN, SCOPE AND COVERAGE

This section provides more information on the following aspects of the HECS sample:

Sample design
Scope
Coverage
Selected dwellings, sample loss and selected households
Responding households and final sample


SAMPLE DESIGN

Dwellings enumerated between January and June 2012 were drawn from the same sample of dwelling selected for the second half of the 2011-12 Survey of Income and Housing (SIH). A separate HECS-only sample of dwellings was selected for enumeration between July and December 2012.

Both HECS and SIH follow the same sample design principles. More information on the SIH sample design can be located in Part 2.2 of the Survey of Income and Housing, User Guide, Australia, 2011-12 (cat. no. 6553.0).

The HECS dwelling based sample is designed to produce reliable estimates for broad aggregates of total income and total energy expenditure for households resident in private dwellings in Australia, for each state and for the capital cities in each state and territory. More detailed estimates should be used with caution, especially for Tasmania, the Northern Territory and the Australian Capital Territory where estimates may have high relative standard errors.

Dwellings were selected through a stratified, multistage cluster design from the private dwelling framework of the ABS Population Survey Master Sample. Selections were distributed across the twelve months of 2012 so that the survey results are representative of energy consumption and income patterns across the year.

Self-complete paper form

Households interviewed were also asked to complete a self-complete paper form, which was issued to households at the time of interview. The form could be completed either during interview, between visits for personal interviews, or after the interview and returned to the ABS via post.

There is likely to be some non-response bias which may affect the representativeness of the information collected from the paper form as not all forms were completed and returned. Non-response of self-complete paper is discussed below in the responding and final sample section.

Longitudinal participation

Households were asked if they were willing to participate in additional questions on their household energy usage and costs every three months until the January - March calendar quarter of 2013. Therefore households could choose to participate in up to four periods of data collection, commencing in the April - June calendar quarter of 2012.

More information on how households participated in the longitudinal component is available in the 'Data collection and item description' section of this User Guide.

As participation in the longitudinal component was voluntary, there are likely to be non-response biases which affect the representativeness of the information collected from this group. Non-response for the longitudinal component is discussed below in the responding and final sample section.

SCOPE

HECS collected information from usual residents of private dwellings in urban and rural areas of Australia (excluding very remote areas), covering about 97% of the people living in Australia. Households provided information by personal interview and a self-complete paper form, and could participate in the longitudinal component via a web form or telephone interview.

Private dwellings are houses, flats, home units, caravans, garages, tents and other structures that were used as places of residence at the time of interview. Long-stay caravan parks are also included. These are distinct from non-private dwellings which include hotels, boarding schools, boarding houses and institutions. Residents of non-private dwellings are excluded.

Usual residents excludes:

• households that contain members of non-Australian defence forces stationed in Australia, and
• households that contain diplomatic personnel of overseas governments
• households in collection districts defined as very remote - this has only minor impact on aggregate estimates except in the Northern Territory where such households account for about 23% of the population.

For most states and territories the exclusion of people in very remote areas has only a minor impact on any aggregate estimates that are produced because they constitute just a small proportion of the population. Very remote and remote areas are defined by the assignment of an Accessibility/Remoteness Index of Australia (ARIA) score. ARIA is a remoteness value (a continuous variable between 0 and 15) that measures the physical distance which separates people in a particular area and where their goods, services and opportunities for social interaction may be accessed. The range of ARIA scores have been categorised as follows:

• Least Remote: defined as having an ARIA score less then 5.95
• Remote: defined as having an ARIA score greater than or equal to 5.95 but less than 10.5
• Very Remote: defined as having an ARIA score greater than or equal to 10.5.
• The ARIA categories and how ARIA scores are calculated are further explained in the Australian Standard Geographical Classification (ASGC) (cat. no. 1216.0).

COVERAGE

Information was collected only from usual residents. Usual residents were residents who regarded the dwelling as their own or main home. Others present were considered to be visitors and were not asked to participate in the survey.

The longitudinal component was only collected from households who provided consent to participate, who supplied contact information and who could complete future questionnaires via the internet or telephone.

SELECTED DWELLINGS, SAMPLE LOSS AND SELECTED HOUSEHOLDS

18,665 dwellings were initially selected for the HECS sample. When fieldwork commenced some dwellings selected for inclusion were found to be out of scope units. Collectively these are referred to as sample loss, and are composed of the following groups:

• dwellings that are out of scope of the survey, under construction, demolished, or converted to non-private dwellings or non-dwellings
• vacant private dwellings
• private dwellings that contain only either out of scope residents (e.g. dwellings occupied by foreign diplomats and their dependants), or visitors.
  

For HECS, sample loss was 2,868 dwellings, 15.4% of the selected combined sample.

Sometimes dwellings that have been selected for inclusion in a survey were found to comprise more than one actual dwelling because, for example, an additional residence such as a 'granny flat' had been added to the original dwelling. In such cases, each actual dwelling becomes a separate household. Occasionally the residents of a selected dwelling request that their details be provided separately from other dwelling residents, for privacy reasons. A separate household is then created for each such group of residents. For HECS, 31 selected dwellings were split into two households, nine were split into three households, and one was split into four or more households.

The net result was that 15,797 households were approached to complete the HECS.
RESPONDING HOUSEHOLDS AND FINAL SAMPLE

In scope households selected for inclusion in the survey can be categorised as responding or non-responding households. Responding households are either fully responding or partially responding. In the HECS, some information missing from partially responding households is imputed, as described in the 'Data processing methods' section of this User Guide.

Non-responding households included:

• households affected by death or illness of a household member
• households in which the significant person(s) in the household did not respond because they could not be contacted, had language problems or refused to participate
• households in which the significant person(s) did not respond to key questions.

The final sample on which most estimates are based, is composed of persons for which all necessary information is available. The information may have been wholly provided at the interview (fully-responding) or may have been completed through imputation for partially responding households. However some items for household energy (including all non-response to the self-complete paper form and the longitudinal component) were not able to be imputed. These are further outlined in the "Data processing methods" section of this user guide.

Of the selected dwellings, there were 15,797 in the scope of the survey, of which 11,978 (75.8%) were included as part of the final estimates. The final sample consists of those 11,978 households, comprising 23,402 persons aged 15 years old and over. The final sample includes 545 households which had at least one imputed value in household energy expenditure.

The following table shows the distribution of the final samples between states and territories and between capital cities and the balance of state.

HECS FINAL SAMPLE, 2012

Capital city

Balance of state

Total

---

	Households	Persons (a)	Households	Persons (a)	Households	Persons (a)
	no.	no.	no.	no.	no.	no.
NSW	1,248	2,684	804	1,481	2,052	4,165
Vic.	1,136	2,370	909	1,677	2,045	4,047
Qld	850	1,692	849	1,637	1,699	3,329
SA	1,023	1,982	1,001	1,821	2,024	3,803
WA	892	1,729	911	1,746	1,803	3,475
Tas.	440	822	827	1,534	1,267	2,356
NT	347	693	62	124	409	817
ACT	679	1,414	0	0	679	1,414
Aust.	6,615	13,386	5,363	10,020	11,978	23,406

---

- nil or rounded to zero
(a) Number of persons aged 15 years and over.

Of the 11,978 households who were interviewed for the HECS, there were 4,051 (34%) households who responded to at least one collection period in the longitudinal component. Non-responding households in the longitudinal component included households who did not respond to either the online web questionnaire or telephone interview. 83 households had one or more longitudinal records removed as the respondents had moved dwellings since either their household interview or their last longitudinal collection. All subsequent longitudinal collection records for these households are also excluded from the final sample.

A summary of longitudinal participation by quarter of interview is provided in the table below. Note that the quarter of interview is based on pension indexation dates, which means these do not align exactly with calendar dates. For instance, the January - March 2012 quarter of interview excludes dates from the 20 - 31 March (for more information, see the quarter of interview glossary term).

NUMBER OF HOUSEHOLDS PARTICIPATING LONGITUDINAL PARTICIPATION BY QUARTER OF INTERVIEW

	Households interviewed	Longitudinal consent	Number of responses to longitudinal (a)				Partial longitudinal	Complete longitudinal
			April - June 2012	July - September 2012	October - December 2012	January - March 2013

---

Quarter of interview
1 January - 19 March 2012	2,872	1,425	973	920	804	802	561	639
20 March - 30 June 2012	3,909	1,828	98	1054	929	931	600	656
1 July - 19 September 2012	2,276	1,029	-	1	652	644	222	537
20 Sept - 30 December 2012	2,921	1,311	-	-	70	896	na	896
Total	11,978	5,593	1,071	1,975	2,455	3,273

---

na not applicable
(a) A small number of longitudinal follow up interviews took place in the same quarter of household interview. This is mainly due to differences due to quarter of interview reflecting pension indexation quarters and longitudinal follow-up periods reflecting calendar quarters.

Non-responding households for the self-complete paper form include those households who did not return their paper form to the interviewer (during interview) or to the ABS via post (after interview). Of the 11,978 households who were interviewed for the HECS, 9,191 (77%) completed a paper form.

There are likely to be elements of non-response bias for data obtained from the paper form or through the longitudinal component. These characteristics should be considered when analysing and interpreting data from these records. The risk of biased estimates increases where significant percentages of missing values exist for estimation classes. This is particularly true when the characteristics of the respondents with missing values and those with known values are significantly different. This also implies that the uncertainty in these estimates may be greater than is indicated by RSEs.

The following table presents selected characteristics of households who provided data from the self-complete paper form and/ or who responded to the longitudinal component.. Also presented, for comparison, are the characteristics of the HECS final sample.

It highlights particular groups of the population where there may be some risk of biased estimation, as the responding population differs from the fully responding HECS sample (which itself is a representative sample). For instance, most classes of household tenure, when compared to the fully responding HECS sample, are more highly represented by owner households than renter households, for paper form and longitudinal components. On the other hand, known estimates for age of dwelling have similar representation in each of the three components, when compared to the HECS fully responding sample.

CHARACTERISTICS OF RESPONDING HOUSEHOLDS TO PAPER FORM AND LONGITUDINAL COMPONENTS

	Weighted estimates (a)
	Completed paper form	Provided longitudinal data	Fully responding HECS sample

---

Household characteristics (during household interview)			%
Source(s) of energy used in dwelling (b)
Electricity only	31	30	32
Electricity and mains gas only	46	44	45
Electricity, mains gas and other sources of household energy	4	5^	4
Electricity and LPG/ bottled gas only	8	9	8
Electricity, LPG/ bottled gas and other sources of household energy	4	5	4
Electricity and other sources of energy only	6	7	6
Total	100	100	100
Gross household income quintile
Lowest	20	15^	20
Second	21	21	20
Third	20	21	20
Fourth	20	20	20
Highest	20	23^	20
Total	100	100	100
Main source of household income
Wages and salaries	60^	62	61
Own unincorporated business	4	5	4
Government pensions and allowances	25	22^	25
Other income	11^	11^	9
Total	100	100	100
Household net worth quintile
Lowest	18^	12^	20
Second	20	19	20
Third	20	21	20
Fourth	21	22^	20
Highest	21	26^	20
Total	100	100	100
Tenure and landlord type
Owner households	69^	77^	67
Owner without a mortgage	34^	36^	31
Owner with a mortgage	35	41^	36
Renters (c)	29^	22^	31
State/territory housing authority	4	3^	4
Private landlord	24^	18^	25
Total (d)	100	100	100
Dwelling structure
Separate house	79	80^	78
Semi detached, row or terrace house or townhouse	10	10	11
Flat, unit or apartment	11	9^	11
Total (e)	100	100	100
Estimated age of dwelling
Less than 5 years old	7	7	7
5 years to less than 20 years old	27	27	27
20 to less than 30 years old	17	16	17
30 or more years old	48	49	48
Don't know	1^	1^	2
Total	100	100	100
Family composition of household
Couple family with dependent children	25	28	26
One parent family with dependent children	6	5^	6
Couple only	28^	31^	26
Other one family households	12	11	12
Multiple family households	2	2	2
Lone person	25	22^	25
Group households	3	2^	4
Total	100	100	100
State or territory of usual residence
New South Wales	30^	34	32
Victoria	26^	25	25
Queensland	20	19	20
South Australia	9^	7	8
Western Australia	10^	10	11
Tasmania	3^	3	2
Northern Territory	1	1	1
Australian Capital Territory	2	2	2
Total	100	100	100
Climate Zone
High humid summer, warm winter (Zone 1)	2	2^	3
Warm humid summer, mild winter (Zone 2)	18	18	18
Hot dry summer, warm winter (Zone 3)	1	0^	1
Hot dry summer, cool winter (Zone 4)	4	6	5
Warm temperate (Zone 5)	33	33	33
Mild temperate (Zone 6)	35^	31^	33
Cool temperate (Zone 7)	8	10	7
Total	100	100	100
Total sample (n)	9,191	4,051	11,978
Number of households ('000) (a)	6,560.8	2,272.5	8,733.6

---

^ Proportion of is significantly different to the fully responding HECS sample.
(a) Fully responding household weights.
(b) Excludes households who do not use electricity or who use mains gas and LPG/ bottled gas.
(c) Includes other landlord type.
(d) Includes other tenure type.
(e) Includes other dwelling type.

A summary of the characteristics of households who participated in all four collection periods of the longitudinal component is available in an Appendix in the Household Energy Consumption Survey, Australia: Summary of Results, 2012 (cat. no. 4670.0) publication. This appendix demonstrates one method for accounting for non-response bias in the longitudinal data, used for the Feature Article appearing in the same product.