Housing Statistics in Canada
Estimation of Energy Poverty Rates Using the 2021 Census of Population

Highlights

• The 2021 Census showed that 822,000 households in Canada (5.6%) were energy poor, and this was more prevalent in the Atlantic provinces (ranging from 10.7% to 13.7%). Households were energy poor if 10% or more of household after-tax income was spent on dwelling energy bills (electricity and fuel).
• The rate of energy-poor households was different across the territories: Yukon (11.1%), the Northwest Territories (9.5%) and Nunavut (3.7%).
• One-person households (10.4%) and households in dwellings that required major repairs (9.7%) were almost two times more likely to be energy poor compared to all households (5.6%).

This work was done in partnership with Natural Resources Canada.

In 2022, Natural Resources Canada commissioned Statistics Canada to prepare a brief text and statistical tables describing the energy poverty of households using data from the 2021 Census of Population. These findings have been edited collaboratively to better present the results and should be useful to the wider audience interested in this area.

Introduction

Energy poverty affects the living conditions of Canadians. Unaffordable energy costs may lead to utility disconnections, households experiencing hardship due to reduced energy use (such as setting a low temperature in winter) or trade-offs between energy and other necessities, which can affect a household negatively.

The 2019 Survey of Household Spending (SHS) showed that 860,000 households across the provinces were considered energy-poor, meaning the household spent 10% or more of its after-tax income on dwelling energy bills. The results also showed that energy poverty varied regionally and by household and dwelling characteristics.

Energy poverty statistics can also be calculated from the 2021 Census of Population.^Note While the SHS is conducted more frequently, the advantage of the census is that data are collected on reserves and in the territories, and it has a large sample that permits detailed examination by various characteristics.

This paper uses data from the 2021 Census to respond to two questions on measuring energy poverty:

1. What were the energy poverty rates based on the census?
2. The current calculation of energy poverty includes only dwelling energy costs that were billed directly, which is the information available in the SHS. How would indirect payments for dwelling energy—that is, payments included in rent or other payments—affect energy poverty rates if they could be included in the calculation of energy poverty?

Part 1: Energy poverty rates based on the census

The 2021 Census showed that 822,000 households in Canada (5.6%) were energy poor. Chart 1 shows the rates by province and territory. The relative energy poverty rates across the provinces were like those observed in the SHS data, with Quebec, Ontario and British Columbia having the lowest rates of energy-poor households, followed by the Prairie provinces and then the Atlantic provinces.

The rate of energy-poor households was different across the territories. Yukon (11.1%) and the Northwest Territories (9.5%) had rates near the upper end of the provincial range. However, the rate in Nunavut (3.7%) was the lowest across all provinces and territories. While the colder climate and remoteness of Nunavut might be associated with higher energy consumption and a higher cost of delivery, the low rate of energy-poor households reflected the high proportion of subsidized rental housing. About 70% of households in Nunavut live in subsidized rental housing.

Data table for Chart 1



Data table for chart 1
Table summary
This table displays the results of Data table for chart 1 All households, calculated using percentage of households spending 10 percent or more of household after-tax income on energy units of measure (appearing as column headers).

	All households
	percentage of households spending 10 percent or more of household after-tax income on energy
Can.	5.6
N.L.	13.7
P.E.I.	11.4
N.S.	12.5
N.B.	10.7
Que.	4.2
Ont.	4.8
Man.	5.6
Sask.	6.9
Alta.	7.3
B.C.	4.9
Y.T.	11.1
N.W.T.	9.5
Nvt.	3.7
Source: Statistics Canada, Census of Population, 2021.

In addition to the rates of energy-poor households in the territories, the census provides the rates on reserves. There were 119,000 households on reserves (0.8%), and the proportion of on-reserve households that were energy poor (12.9%) was more than double the rate for all households (5.6%).

Chart 2 shows rates of energy-poor households by various characteristics. The energy poverty rate was higher than the national average (5.6%) for households maintained by a person 55 years and older (6.9%). It was also higher for one-parent (6.4%) and one-person households (10.4%). Households that owned their dwelling also had a higher energy poverty rate (6.1%). Households living in dwellings that are single-detached house (6.8%), built in 1960 or before (7.7%), or needing major repairs (9.7%) had higher energy poverty rates compared to the national average.

Data table for Chart 2



Data table for chart 2
Table summary
This table displays the results of Data table for chart 2. The information is grouped by Characteristics (appearing as row headers), Percentage of households spending 10 percent or more of household after-tax income on energy (appearing as column headers).

Characteristics	Percentage of households spending 10 percent or more of household after-tax income on energy
Panel A – Select housing characteristics
Age of primary maintainer
15 to 29 years	4.8
30 to 39 years	3.7
40 to 54 years	4.5
55 to 64 years	6.5
65 years and older	7.2
Household type
Couple with children	2.4
Couple without children	3.8
One parent	6.4
One-person household	10.4
Other	3.4
Structural type
Single-detached house	6.8
Double, row, terrace, duplex	5.0
Apartment or other	3.8
Panel B – Select housing characteristics
Period of construction
1960 or before	7.7
1961 to 1980	5.9
1981 to 2000	4.9
2001 to 2021	4.2
Condition of dwelling
Only regular maintenance and minor repairs needed	5.3
Major repairs needed	9.7
Housing tenure
Owner	6.1
Renter	4.3
All households	5.6
Source: Statistics Canada, Census of Population, 2021.

Part 2: Indirect payments for dwelling energy in relation to energy poverty rates

The energy poverty rates shown in Part 1, as well as those calculated using the SHS, are based on electricity and fuel payments that were made directly by the household. In other words, they do not account for payments that were included elsewhere, such as in rent or condominium fees.

In the 2021 Census, 15.8% of households reported that electricity or fuel costs were included in other payments. Among the 84.2% of households that reported direct electricity or fuel costs, 36.0 percentage points corresponded to households that reported $0 for either electricity or fuel. The remaining 48.2 percentage points reported a non-zero payment for both electricity and fuel (see Chart 3).

When electricity and fuel were considered separately, households reported indirect electricity payments in a similar proportion to fuel (10.7% for electricity and 12.5% for fuel). However, households that reported a zero-dollar cost did so mainly for fuel (38.5%) and not electricity (4.7%). This could be related to households that heat using electricity and not fuel. Overall, 84.6% of households reported a non-zero electricity payment and 49.0% reported a non-zero fuel payment.

Data table for Chart 3



Data table for chart 3
Table summary
This table displays the results of Data table for chart 3 Percent (appearing as column headers).

	Percent
Electricity
Included	10.7
Zero dollars	4.7
Non-zero	84.6
Fuel
Included	12.5
Zero dollars	38.5
Non-zero	49.0
Electricity or fuel
At least one included	15.8
None included (at least one zero dollars)	36.0
None included (both non-zero)	48.2
Source: Statistics Canada, Census of Population, 2021.

The proportion of households reporting indirect electricity or fuel payments varied by characteristics—in particular, by housing tenure, rent subsidy status, province or territory, and structural type (see Table 1). Over one-quarter of tenant households reported indirect electricity (27.3%) and fuel (28.5%) payments, while among owner households, 2.5% reported indirect electricity payments and 4.5% reported indirect fuel payments. The prevalence of the utility being included in other payments also varied across the provinces and territories, especially for fuel costs among renter households, which ranged from 13.0% in Newfoundland and Labrador to 58.9% in Nunavut. Among owner households, those in apartments were more likely to have electricity (11.5%) and fuel (22.6%) included in other payments—possibly in condominium fees. Lastly, those in subsidized rental housing were more likely to have indirect utility payments than those paying market rent (41.2% versus 25.5% for electricity; 33.5% versus 27.9% for fuel).



Households that had indirect electricity or fuel payments were less likely to be energy poor because the measure does not include indirect payments. Because of the definition of energy poverty, households with both electricity and fuel payments included in other payments cannot be energy poor. For households with one of the two utilities included in other payments, 3.2% were energy poor. For households with neither of the two utilities included in other payments, 6.3% were energy poor.

Because indirect payments for utilities varied by characteristics, differences in energy poverty rates across these characteristics may not reflect the true differences in energy affordability. To include indirect energy payments in the energy poverty measure, one approach could be to estimate and include the portion of the indirect payments associated with electricity and fuel.

To illustrate the potential impact of including indirect energy payments, an estimation method was developed to create the energy poverty rates in Table 2. The method is based on an analysis presented in Appendix A.

For the energy poverty estimates in Table 2, when electricity or fuel was paid indirectly, a percentage of the overall shelter cost was included in the energy poverty measure. The percentage was derived separately for electricity and fuel. It was based on the average percentage paid by households that paid both utilities directly and share the same category for the following characteristics: tenure, mortgage status, subsidized rent status, province or territory, structural type, and household type.^Note Future research using other estimation methods, such as donor imputation, are also being explored to enhance the estimates of energy poverty when indirect energy payments are included.

Table 2 shows that, when indirect energy payments were included, the energy poverty rate nearly doubled for renter households (from 4.3% to 8.3%), while the rate among owner households was about the same (from 6.1% to 6.5%). This was consistent with the fact that tenants were more likely to have indirect payments.

Because a small proportion of owner households had indirect energy payments, the rates of energy poverty for this group did not change much when indirect payments were included. The one group of owner households that saw a large change was those in apartment buildings—the one group that had a higher incidence of indirect payments. With indirect energy costs included in the energy poverty measure, owner households in apartments had a rate of 7.6% (from 5.1%), which exceeded that of owner households in single-detached houses (6.6%).

With indirect energy costs included, the average annual energy costs for owner households in apartment buildings increased from $1,460 to $2,060, but they remained below the average ($3,160) for owner households in single-detached houses. Thus, the higher prevalence of energy poverty among owner households in apartments, compared with single-detached houses, was also related to the lower income of these households.

Among tenant households, there were large changes in the energy poverty rate when indirect payments were included. Moreover, the relative trends across characteristics changed. Tenant households with a primary maintainer aged 15 to 29 and one-person households saw larger increases in the energy poverty rate than other tenant households. The differences across the provinces and territories also widened, with Prince Edward Island’s tenant energy poverty rate almost tripling, from 6.3% to 18.4%, while Quebec’s rate increased from 3.8% to 6.4% (less than double). Lastly, the energy poverty rate among subsidized renter households was higher than that of non-subsidized renter households (10.9% versus 8.0%), while without the indirect energy payments the rates were similar (4.6% versus 4.3%).

In conclusion, energy poverty rates increased when indirect payments for energy were included in the measure (from 5.5% to 7.1%). The increase was larger among tenant households (4.3% to 8.3%), while the effect on the rate for owner households was smaller (6.1% to 6.5%). This was related to the prevalence of indirect energy payments among the different housing tenures. For tenant households, the relative trends of energy poverty across the characteristics changed when indirect energy payments were included. The tenant households that saw the largest increases in energy poverty rates were households with a maintainer aged younger than 30 years, one-person households, households in Prince Edward Island and households with subsidized rent.



Appendix A: Shelter cost trends when electricity or fuel payments were included in other payments

This appendix describes the background analysis that led to the estimation approach used in the main document and draws the following conclusions:

Conclusion 1: Shelter cost was generally lower when the energy cost was included in other payments. As such, estimating the energy cost as a percentage of the shelter cost is one way to account for this difference.

Conclusion 2: At the census subdivision (CSD) level, there were almost always some respondents who reported energy costs separately. As such, there are respondents upon which to base an estimation. However, in some areas, as additional characteristics are used to group subpopulations, it may be necessary to use information from outside the CSD to have enough respondents to produce an estimation of the energy costs.

Conclusion 3: For some households, the direct electricity payments might include fuel costs, or the direct fuel costs might include electricity payments. It is important to adjust the energy cost estimation to avoid double counting fuel or electricity costs.

A.1 Shelter cost amounts when energy costs were included in other payments

While the main shelter cost components were collected via the census, it is not known which of those components included the electricity or fuel costs when indirect payments for those utilities were reported. Accordingly, the overall shelter cost is the primary choice from which to estimate the indirect payments for electricity and fuel.

The distribution of shelter costs for those that paid energy costs directly versus indirectly was examined through a series of histograms. This review was done separately for four groups for which shelter costs are fundamentally different: renters with a subsidy, renters without a subsidy, owners with a mortgage and owners without a mortgage. Moreover, these distributions were examined by the different characteristics used in the main document (age of primary maintainer, household type, structural type, period of construction, condition of dwelling, and province or territory), as well as a few additional variables (household size, number of bedrooms and number of rooms).

The examination of the histograms shows that the average shelter cost was lower when the electricity was paid for indirectly versus directly. This trend generally holds for both renters with and without a subsidy, as well as when the distributions were examined by the other characteristics.

A regression model confirmed that, after controlling for the other variables, renter households without a subsidy that had indirect electricity payments had $113 less per month in shelter costs, compared with those that paid directly. Among renter households with a subsidy in the regression model, the monthly shelter cost was $130 less when electricity was paid for indirectly.

A closer examination found that it was not just the combined shelter cost that was lower, but the rent charged was also lower when electricity was included. This was unexpected because, all else being equal, rents should be higher if electricity is included. Given the magnitude of the difference—which is close to the entire cost of the utility itself for a household—this suggests that there were unobserved characteristics related to indirect electricity payments that caused rent to be lower in these dwellings. For example, less desirable units may be advertised with utilities included to attract renters. Another example is that these units might more likely be conversions or accessory suites and not purpose-built rentals, because the former are less likely to have dedicated utility meters.

The relation of fuel payment to the overall shelter cost was like that of electricity payment. That is, the average monthly shelter cost was less when the payment for fuel was indirect versus direct. However, the difference was larger than that observed with electricity payments. The regression showed that renter households without a subsidy had $204 less per month in average shelter costs when fuel was paid indirectly versus directly. Among renter households with a subsidy, the average monthly shelter cost was $252 less.

Among owner households, the relation between shelter cost and indirect payments for energy was less straightforward. The histograms showed that average shelter cost was not consistently lower across the various characteristics when payments were indirect. This was because, among owner households, indirect payments for energy were not common and occurred mainly in apartments (likely organized as a condominium). As a result, differences in shelter cost between those that paid for energy indirectly versus directly probably reflected the type of households that were likely to live in condominium apartments and where these apartments were located.

A regression model for owner households that included an interaction term for the geography and structural type showed that the average shelter cost was more consistently lower when energy payments were indirect versus direct. Moreover, a review of histograms for each province and territory and structural type confirmed that the average shelter cost was lower when energy payments were indirect. One exception occurred in some regions where the average shelter cost within apartments was higher for those that pay energy costs indirectly. This exception occurred more frequently among owners without a mortgage. It could reflect further subprovincial geographical concentration of owner-occupied apartments with indirect costs.

In conclusion, households that pay energy costs indirectly generally had lower overall shelter costs. There were unobserved factors leading to lower shelter costs when energy costs were indirect. The analysis does not point to a specific method for estimating indirect energy costs. Instead, it highlights differences between the group of households that pay directly and that of households that pay indirectly, implying that it could be difficult to perfectly estimate energy costs when they are paid indirectly. Assumptions may be required, and a certain level of inaccuracy may need to be accepted.

Estimating an energy cost as a percentage of the overall shelter cost could be one way to account for the differences in shelter costs. It assumes that any discounts on shelter costs are proportionately allocated to rent and energy.

A.2 Energy cost payment methods at the census subdivision level

In some areas, the billing of electricity and fuel costs may be such that it was included in other payments for all households. This would make it difficult to find local respondents who could be used to estimate the proportion of the shelter cost that was spent on energy. These situations may require using information from outside the local area and may lead to less accurate estimates.

Each census subdivision (CSD) was examined to see whether electricity or fuel costs were always paid indirectly. Generally, the frequency at which this occurred was small. Of all the CSDs where an estimate would be required, most of them had at least 10 respondents who reported direct energy payments. Among those that did not have at least 10 respondents, most also did not have many respondents reporting indirect payments (see Table A.1).

Overall, there were seven CSDs with a meaningful number of respondents with indirect payments for electricity and not enough households that paid directly. These seven CSD were all in northern Quebec, in the range of CSD2499075 to CSD2499140. For fuel payments, there were 23 CSDs in the same situation. These included the same CSDs in northern Quebec, where electricity was also mostly paid indirectly, as well as several other CSDs (mostly reserves) in other provinces.



In conclusion, at the CSD level, there were almost always respondents who reported energy costs separately. Therefore, there were respondents upon whom to base an estimation. For the few CSDs where energy costs were mostly paid indirectly, a different imputation method may need to be developed. Also, in some areas, as additional characteristics are used to group subpopulations, it may be necessary to use information from outside the CSD to have enough respondents to produce an estimate of energy costs.

A.3 Fuel costs included in electricity payments, and electricity payments included in fuel costs

For some households, it is also possible that electricity and fuel costs were billed together. In these instances, respondents might have reported the combined value as one of the two utilities and indicated that the other utility was included in other payments.

This creates two estimation challenges. First, if these households are used to estimate the proportion of overall shelter costs spent on electricity or fuel that is then applied to other households with indirect payments, it may result in an overestimation of energy costs. Second, for these households, when the indirect payment is estimated and then added to the payment that covers both electricity and fuel, it will result in a double counting of the utility that was paid indirectly.

Among households assessed for energy poverty, 815,800 reported a non-zero direct payment for either electricity or fuel and indicated that the other utility was included in other payments. Among these households, 90,400 reported a non-zero value that was 90% or more of what both utilities cost for similar households in the area. These households could be those where the direct payment includes the other utility.

To mitigate the risk of overestimating the proportion of overall shelter costs that were spent on a utility, one method is to derive the estimate only from households that directly paid both electricity and fuel. This approach ensures that the electricity payments do not include fuel costs and vice versa.

To mitigate the risk of double counting, the 90,400 households mentioned earlier can use the utility that was reported as the overall energy cost and avoid estimating the indirect payment. In other words, when households paid directly for one utility and indirectly for the other, and the amount reported for the direct-payment utility is high enough to represent both utilities, then the amount reported can be considered as the full energy cost that represents both electricity and fuel costs.

In conclusion, a small proportion of households may have reported a direct payment for electricity or fuel that included the other utility. In these instances, mitigation strategies can be applied to avoid double counting and overestimating energy costs.