The purpose of this survey was to provide Natural Resources Canada data on the energy use characteristics of private dwellings in Canada and on household use of energy resources. The data will allow Natural Resources Canada to assess the effectiveness of existing energy efficiency programs and to develop new ones.
Data release – April 4, 2006 (No public use microdata file was produced by Statistics Canada and data will not be made available through the Data Liberation Initiative (DLI).)
The objective of this survey is to provide Natural Resources Canada (NRCan) information on energy use for Canadian households residing in single family dwellings and residential buildings less than five stories. The survey profiles use of selected energy consuming equipment and appliances, energy related characteristics of dwellings, household demographics, identified patterns of behaviour related to consumption, and amounts of energy consumed during the reference period. Data on the age and size of the dwelling, dwelling conditions, improvements, types of heating and cooling equipment, and demographic information, including family composition and income, have been collected in this survey.
The survey data will allow the Office of Energy Efficiency (OEE) of Natural Resources Canada (NRCan), to determine average household energy intensity levels (energy usage vs. area) and compare the intensity levels of different types of houses, house area, age of houses, storeys of houses, population centers, etc. These data can also be broken down into energy source sub-categories, for efficiency comparisons. As a result, the data will allow NRCan to improve its understanding of household energy use (trends); assess energy efficiency potential in the residential sector; analyze and better understand the effects of energy efficiency measures being considered for the future and, as a result, review existing energy efficiency programs and assist with the implementation of energy-consumption regulations for household appliances. The regulations set minimum performance standards to eliminate less energy-efficient products from the market. As well, the SHEU data will be used to decide where future policy should be directed to improve Canada's energy efficiency, in an effort to reduce greenhouse gas emissions and meet our Kyoto objectives.
Besides NRCan, a network of energy end-use centres created by NRCan, provincial governments and utility companies will also examine the trends of energy use, using published SHEU reports.
The "2003 Survey of Household Energy Use (SHEU): Summary and Detailed Reports" are available free of charge by contacting Natural Resources Canada (fax: 613-947-4120, firstname.lastname@example.org) or under "Energy Use Statistics" at the following NRCan Web site (http:// oee.nrcan.gc.ca/publications).
The target population is comprised of dwellings that were occupied as primary residences in the ten Canadian provinces and that fit into one of the following categories: single detached, semi-detached, row, mobile, duplex or dwelling in a building with no more than four storeys (the last two categories were not covered in the 1998 edition of the SHEU). Specifically excluded from the survey's coverage were dwellings not mentioned above, dwellings located in a First Nation community or military base, businesses, institutions, demolished dwellings, dwellings under construction, seasonal or secondary residences, and dwellings occupied by individuals who work full-time within the Canadian Armed Forces.
The questionnaire was designed as two separate and distinct components. The first component, the more detailed of the two, applied to the occupants of all the dwellings selected in the sample. The second, shorter, component required for rented dwellings and condominiums, applied to landlords or the condominium managers. It was designed to gather information that tenants and condominium occupants did not know, or were unlikely to know.
The questions in the questionnaire were primarily pooled from the previous SHEU surveys. Additional revisions to the content of the questionnaire were inspired by the needs of various subject matter areas at the Office of Energy Efficiency (OEE) of Natural Resources Canada (NRCan).
After undergoing focus group testing in 6 cities covering 4 regions of Canada, the finalized questionnaire was sent to the Operations and Research Development Division (ORDD) to be programmed as a "Computer Assisted Personal Interview" (CAPI) application. The focus group testing was coordinated by Statistic Canada's Questionnaire Design Resource Centre (QDRC).
Before being deployed to regional interviewers, the CAPI application underwent a series of qualitative tests by the design and development teams to ensure proper functioning of question flows and edit checks. Additional tests were conducted by regional delegates to test the portability of the electronic questionnaires across interviewer laptops and across secured transmission lines to head office.
Permissions to collect energy consumption data from suppliers (electricity, gas, oil), and the release of this data by suppliers to Statistics Canada and Natural Resources Canada (under a data sharing agreement), necessitated the design of a special form to address both the legal and privacy issues associated with protecting the parties concerned. The "Energy Supplier Release Form" was expressly designed for that purpose and was only applied when the form was duly signed by the account holder.
This is a sample survey with a cross-sectional design.
The SHEU sample was composed of respondents from the area frame of the Canadian Community Health Survey (CCHS) who were interviewed for the CCHS between January and October 2003. The auxiliary information available for SHEU purposes -- type of dwelling, telephone number and occupant type (tenant or owner) -- was therefore relatively recent when we established the sample in January 2004; the information had been gathered during the previous three to twelve months and was six months old on average.
The SHEU sample is a stratified simple random subsample of CCHS respondents. The SHEU strata definitions are based on the provincial health regions, with some slight modifications to avoid overlap between strata of the Labour Force Survey (LFS) and those of the health regions (LFS strata not assigned to a single health region were assigned to one). The sample allocation to SHEU strata was proportional to population estimates. This approach to sample design, pseudo-proportional to the weight, has the advantage of reducing design effects, resulting in estimates of better quality for given sample sizes.
The sample of dwellings for the survey was designed to produce reliable estimates for each of the five large Canadian regions (Atlantic provinces, Quebec, Ontario, Prairie provinces, British Columbia) as well as national estimates for certain aggregate variables such as groupings by type of dwelling or year built, and by rural or urban location.
Keeping in mind the proposed survey design and the expected response rate, the sample size required to meet the analytical needs was 6,433 dwellings.
Please refer to section 2.3 (Sample design) of APPENDIX A. METHODOLOGY in the "Additional documentation" link that follows for details on the distribution of these dwellings.
Data collection for this reference period: 2004-03-15 – 2004-09-30
Responding to this survey is voluntary.
Data are collected directly from survey respondents and extracted from administrative files.
The first phase of data collection for SHEU included the work of obtaining information from the individuals residing in the sampled dwellings and third parties such as landlords, property or condominium managers. The second phase involved collection of consumption data from suppliers of energy to these selected dwellings.
Shortly before the data was to be collected, a letter of introduction summarizing the purpose of the survey was sent out by the regional offices of Statistics Canada to the occupants of the sampled dwellings.
The first phase of collection, conducted from March 15th to June 11th of 2004, necessitated that interviews be done in person at the dwellings selected in the sample, notably to help respondents answer questions about the dwelling characteristics and to obtain signed consents to contact energy providers. Interviewers were asked to contact the occupant of the selected dwelling by telephone to arrange a Computer-Assisted Personal Interview (CAPI), preferably with the household energy account holder(s) or person(s) most knowledgeable about home heating, major appliances and the physical features of the dwelling.
If the selected dwelling was rented or part of a condominium, the landlord, property or condominium manager was contacted to obtain additional information on physical features of the dwelling and on heating/cooling equipment which may have been unknown to the occupant. Personal interviews were conducted when landlords, property or condominium managers were responsible for payment of utility bills and a signed authorization to contact the energy suppliers was required. Otherwise the interviews were mostly conducted by telephone.
A response rate of 70.7%, which represented 4,551 cases out of the 6,433 sampled dwellings, was attained for the first phase of collection. These cases were judged to contain sufficient data and had satisfied the requirement for data sharing between Statistics Canada and Natural Resources Canada. After factoring 309 cases classified as "out of scope", and readjusting the number of eligible sampled dwellings to 6,124, the adjusted response rate was 74.3%.
The second phase of data collection was coordinated by Statistics Canada head office. It involved gathering electricity, gas and heating oil consumption for the 2003 reference year from suppliers in cases where the account holders had given their signed consent.
A total of 338 energy suppliers had to be contacted to obtain consumption data for approximately 4,000 electricity, 1,930 natural gas, 540 heating oil and 80 propane accounts. The data return rates by energy type were: electricity 76%, natural gas 64%, heating oil 48% and propane 41%. This phase of collection was conducted from July through September, 2004.
The use of the "Blaise" software in programming the "Computer Assisted Personal Interviewing" (CAPI) application allowed for several edits to be built directly into the electronic questionnaire. Flow edits ensured logical sequencing of questions, range edits ensured interviewers entered values within specified limits and a limited number of consistency edits ensured a logical interface between responses.
At the post collection phase the data was subjected to further processes of validation through a "Generalized Processing System" (GPS). (The GPS was built to convert edit specifications, prepared in Excel spreadsheets and LogiPlus (decision tables), into SAS programs. Additional custom SAS programs for validation, weighting and imputation were developed by the survey methodologists.) For example, at the "Clean-up" step, measures of non response to key data elements essential in calculating energy consumption were used to determine whether a record should be kept or rejected. Additional sets of rules (decision tables) were constructed during the "Edit" and "Derivation" steps to ensure that derived values from various data components were consistent and plausible. For example, that the types of fuels used correlated with equipments used in the dwelling.
The rejected records were considered non-responses. Approximately 200 records were rejected because of poor quality or insufficient data.
Derived values for heated areas (living space, basement, and garage) were screened using the Sigma-gap method (this method sorts the data and selects all of the records from the record where the distance from the previous record is greater than one standard deviation) whereas values for energy consumption were validated using a ratio interval method. If the calculated energy intensity (dwelling heated area / total consumption in GJ) was outside the determined interval, the data were either fixed manually or rejected and flagged for imputation. Missing values for heated areas and missing amounts of energy consumed were imputed.
Responses to questions that applied to the respondent but had missing or invalid data were either imputed or assigned the "not stated" code. For questions where it was unknown whether they applied to the respondent, the responses were also assigned the "not stated" code.
Variables with quantitative characteristics, whether collected or derived, were subjected to a screening of outlier values and imputation. The screening of outlier values was most often done by the use of the sigma-gap method and imputation by the "hot deck" method.
The rates of imputation range from 1% to 77%, the median of the rates for the 20 imputed variables being 25%. Recommendations were made to assess the quality of estimates based on imputed variables.
Please refer to Section 4.1 (Processing the data) of APPENDIX A. METHODOLOGY for a summary on the rates of imputation for the SHEU variables subject to imputation.
The weighting of the SHEU sample involved 11 steps:
1. Adjustment of the CCHS subweight to account for the undercoverage at the dwelling level of the file used to draw the SHEU sample (the CCHS file was missing cases where the household responded but the selected member did not respond)
2. Selection of the SHEU sample
3. Adjustment to account for multiple dwellings
4. Adjustment to account for type-0.1 full non-responses (dwellings whose address was not confirmed)
5. Adjustment to account for type-0.2 full non-responses (dwellings whose occupancy was not confirmed)
6. Adjustment to account for type-1 full non-responses (no data gathered)
7. Adjustment to account for type-2 full non-responses (households who moved into the selected dwellings in 2004)
8. Adjustment to account for type-1 partial non-responses (data on the heating system and hot water tank, used for determining the energy sources used by the dwelling, were missing)
9. Adjustment to account for type-2 partial non-responses (other cases of partial non-response)
10. Initial calibration of weights to produce estimates consistent with CCHS estimates, with two margins: first, the number of dwellings in scope, by province, and, second, the number of dwellings based on certain groupings of dwelling types and based on a selected geographic level (national or regional)
11. Second calibration of weights to produce estimates consistent with LFS estimates in 10 domains: grouped type of dwelling (single detached dwellings vs. all others) for each of the usual five infranational regions (Atlantic provinces, Quebec, Ontario, Prairie provinces, British Columbia)
The bootstrap method was used for estimating the SHEU variance. The 500 bootstrap weights of the CCHS were subjected to the 11 steps of the SHEU weighting procedure.
Overall, the data gathered for the SHEU are of good quality. The reports of the interviewers and the internal consistency of the data attest to this. However, as described in Section 4.1 (Processing the data) of APPENDIX A. METHODOLOGY, some aspects of the quality of the data and some reservations emerging during the collection and processing phases are worthy of note.
The guidelines described in Table 14 of section 4.3 (Variance Estimation) APPENDIX A METHODOLOGY determine the quality of the estimates produced from non-imputed variables. In the case of imputed variables, the same procedure applies but the concept of coefficient of variation is replaced by the concept of adjusted coefficient of variation to account for the rate of imputation. The Bootvar program is used to calculate the coefficient of variation of an estimate.
Statistics Canada is prohibited by law from releasing any data which would divulge information obtained under the Statistics Act that relates to any identifiable person, business or organization without the prior knowledge or the consent in writing of that person, business or organization. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.
Respondents, including non-occupant owners where applicable, were informed of a data-sharing agreement under which Natural Resources Canada would have access to a complete microdata file. Respondents were asked the following question for their authorization to share the gathered information:
Statistics Canada has entered into an agreement with the Office of Energy Efficiency of Natural Resources Canada (NRCan) to share information from this survey.
NRCan will not be given your name or other identifiers. The information shared with them will contain sufficient geographic detail to allow them to analyse the data for small areas. NRCan has agreed to keep all the information provided confidential and to use it only for statistical purposes.
Do you agree to allow Statistics Canada to share your information with NRCan?
An answer in the affirmative to that question authorized Statistics Canada to share the information gathered with Natural Resources Canada.
The coverage of the target population for the SHEU was excellent, given that the sample consisted of responding households from the CCHS area frame (the household response rate to the CCHS was 89% for the ten provinces as a whole), and interviews could be conducted with households that did not own a telephone.
Dwelling occupants who had moved into their dwellings during 2004 were excluded from the SHEU and treated as non respondents at the weighting stage. The main justification stemmed from the fact that data on energy consumption for the 2003 reference year would not be available. Respondents who took up residence in their dwelling in 2003 were interviewed but, as expected, their rates of imputation for the energy use variables were higher than the rest of the sample.
Given that an average of 12 months separated the collection of SHEU and CCHS, the classification of dwelling types at the time of SHEU could differ from the dwelling classification at the time of CCHS. This could be the result of alterations to the dwelling between the two collection periods or to a classification error in one of the surveys. For the SHEU respondents, 8% (weighted and unweighted) of the dwelling classifications had changed between the two surveys. This rate is markedly higher in the case of duplex and semi-detached residences, which suggests that these two types of dwelling may have been mistakenly switched during data collection.
The raw response rate for the SHEU was 70.7% (number of respondents divided by the initial sample size) and the operational response rate was 75.5% (sum of the number of respondents and the number of out-of-scope units divided by the initial sample size).
For a proportion estimated at 10%, the SHEU was aiming for a coefficient of variation of 6.6% at the national level and 14.5% for each of the five infranational regions. These targets were surpassed, with a CV of 5.5% at the national level and CVs varying between 10.8% and 13.9 % at the regional level.