Income Research Paper Series
Preliminary poverty estimates in Canada: A new modelling approach

Introduction

Official poverty estimates in Canada are determined using the Market Basket Measure (MBM).^Note The MBM is a basket of costs based on five main components: food, clothing, shelter, transportation and other basic necessities.^Note According to the MBM, individuals or families are considered to live in poverty if their disposable income is insufficient to purchase a predetermined basket of goods and services required to achieve a modest, basic standard of living given their family size and area of residence.

Canada’s annual poverty estimates are produced using the Canadian Income Survey (CIS).^Note The CIS is a cross-sectional sample survey that collects information during the first six months of a calendar year (i.e., January to June). As is the case with many statistical programs, there is a trade-off between the timeliness and the accuracy of the published estimates. For example, income estimates, including poverty estimates, are not typically available until about 16 months following the reference year. More specifically, the 2022 CIS reference year ends in December 2022, while the income estimates were released in April 2024. The main reason for the delay in producing official estimates is the CIS’s dependence on tax files, and especially the T1 file. Because most of the income data from the CIS are derived by tax linkages to respondents, the tax data must be complete and cleaned before they can be used in the CIS production.

Typically, throughout the year following the reference year, the Canada Revenue Agency will provide Statistics Canada with different versions, or “cuts,” of the tax data. As the year progresses, the data become more complete. For example, late tax filers and self-employed individuals (who have a later filing deadline) are often missing from the initial versions of the files, but as new tax data become available later in the year, the files become more complete.

This paper proposes a method of producing preliminary poverty estimates up to seven months before the official release by using preliminary tax slips (approximately 9 months after the end of the reference year), while ensuring the estimates maintain reasonable revision limits. Following the release of this paper, Statistics Canada will continue to provide preliminary poverty estimates each fall following the reference year using the methodology described in this paper.^Note The official poverty estimates will continue to be released in the spring of the year after the reference year.

Method

To produce preliminary poverty estimates, the proposed methodology uses the sample of CIS respondents from the previous year and either their tax slips for the current year (where available) or historical imputation to derive an estimate of disposable income for the current year. The previous year’s sampling weights will be retained, with adjustments for non-response and calibration to the current reference year’s population totals for certain subgroups to address coverage issues (e.g., emigration or immigration, deaths or births)^Note , and to mitigate the risk of bias. Finally, the CIS respondents’ age will be adjusted to reflect the current year.

Official poverty estimates use T1 Income Tax and Benefit Return data and data from various tax slips.^Note The final T1 data for a given reference year are not available until near the end of the calendar year following the reference year. However, tax slips are issued by employers and banking institutions, and they are generally available in the early fall of the year following the reference year. Therefore, only tax slips are used for this approach. The following tax slips will be used:

• T4 Statement of Remuneration Paid
• T4A Statement of Pension, Retirement, Annuity, and Other Income
• T4E Statement of Employment Insurance and Other Benefits
• T4A(OAS) Statement of Old Age Security
• T4A(P) Statement of Canada Pension Plan Benefits
• T5007 Statement of Benefits
• T3 Statement of Trust Income Allocations and Designations
• T5 Statement of Investment Income
• T5013 Statement of Partnership Income.

When tax slips are unavailable for a respondent in the reference year, their disposable income is either imputed using the value from the previous year or derived from the tax slips of a donor record. The figure below illustrates the imputation strategy based on tax slip availability for the current year.

Description for Figure 1

Figure one shows a flow chart of the imputation strategy described in the paper. The top box has the text, “A person is selected from the previous year’s Canadian Income Survey (CIS)”. It is then split into two paths. The left path has the text, “The person responded to the previous year’s CIS”. Below that box are two additional boxes. The one on the left has the text, “At least one tax slip is available for the current year”, while the one on the right has the text “No tax slips are available for the current year”. Below the left box is one box with the text “Use the current year’s tax slip to derive disposable income”, while below the right box is one box with the text “Impute using disposable income from the previous year”. Starting back at the top, but following the right path, the box has the text “The person did not respond to the previous year’s CIS (income was imputed via a donor record)”. Under that box are two more boxes. The box on the left has the text “At least one tax slip is available for the current year” and has one connected box under it with the text “Use the current year’s tax slip to derive disposable income”. The box on the right has the text “No tax slips are available for the current year” and two additional boxes underneath it. The box on the left has the text “The current year’s tax slips are available for the same donor used in the previous year” with one box under it with the text “Impute using the current year’s tax slips to derive disposable income for the same donor”. The box on the right has the text “The current year’s tax slips are unavailable for the same donor” with one connected box having the text “Impute using disposable income for the previous year”.

Source: Authors' visualization.

Table 1 shows the number of CIS respondents for selected reference years, along with the availability of their tax slips and donor tax slips used for imputation.



The disposable income is derived for a specific year based on all available tax slips. An estimate (or proxy) of disposable income is derived by integrating tax slips from the previous and current years. In addition, adjustments by type of tenure were applied to the current year after adapting for inflation. Once an estimate of disposable income for the current year is available, the percentage of Canadians who fall below the MBM threshold represents the preliminary poverty rate.

Results

National official poverty rates and estimated preliminary poverty rates for selected family types and demographic groups from 2018 to 2023 are presented in Table 2.^Note The preliminary poverty rates were produced using the most recently available tax slips from late September of each year.  According to the modelling methodology previously described the preliminary poverty rate for Canada was estimated at 10.6% in 2023. This would suggest that the poverty rate will continue its recent upward trend in 2023, which began in 2021 (7.4% in 2021; 9.9% in 2022).  The modelled rates suggest that the increase should impact all the selected age groups and those in families. For the preliminary estimates of the number of people in poverty, please see the tables in Appendix A.2.



Table 3 presents the preliminary poverty rates by province. The majority of provinces (7-out-of-10) are predicted to have an increase in their 2023 poverty rate compared to the previous year, with Quebec (1.1 percentage points) and Alberta (1.0 percentage points) projected to have the largest increases.



Preliminary poverty rates by selected economic family types are presented in Table 4.^Note According to the model, persons in economic families experienced an increase in their 2023 poverty rate (1.0 percentage points) from 2022 (6.6%), with non-elderly families experiencing a relatively larger increase in 2023 (1.1 percentage points) compared to the previous year (7.1% in 2022).



Next steps

Statistics Canada’s intention is to continually improve the methodology presented in this paper. Resources will be put in place to enhance data quality indicators, expand the availability of preliminary estimates and determine whether timeliness can be improved. Efforts will be made to expand the preliminary estimates from poverty to sources of income and other income-based estimates that are currently available.^Note Finally, further improvements to timeliness will be evaluated to determine whether earlier versions of tax slip files can be enhanced and used instead.

Limitations

Several assumptions were required when applying this method, including the assumption that all sociodemographic characteristics of respondents in the previous year’s CIS sample—such as region of residence, type of tenure and family composition—had not changed for the current year. However, this assumption, which may not hold true, could lead to an incorrect estimate of disposable income. Additionally, some components of disposable income were not adjusted, as they were unavailable on tax slips. These components include net income from self-employment; income tax; capital gains tax adjustments; and various expenses excluded from disposable income, such as employment insurance contributions, Canada Pension Plan or Quebec Pension Plan contributions, registered pension plan contributions, child care expenses, alimony paid, and direct medical expenses. Future studies are planned to evaluate the effects of these limitations.

Conclusion

This paper marks the first time Statistics Canada has produced preliminary poverty estimates using a probabilistic sample and tax slips. By using a methodology similar to the one presented in this paper, preliminary poverty estimates will continue to be produced in a timelier manner (e.g., seven months earlier, or approximately 9 months after the end of the reference year), compared with the official poverty estimates. While this method has effectively produced poverty estimates that require relatively few revisions compared with the historically produced official estimates, work to improve this methodology will continue.

Appendix

   

References

Deville, J.-C., & S¨arndal, C.-E. (1992). Calibration Estimators in Survey Sampling. Journal of the American Statistical Association, 87 (418), 376–382.

Djidel, S., Gustajtis, B., Heisz, A., Lam, K., & McDermot, S. (2019, December 20). Defining Disposable Income in the Market Basket Measure. Income Research Paper Series. Catalogue no. 75F0002M.

Fuller, W. (2009). Sampling Statistics. New York: John Wiley & Sons, Inc.

Horvitz, D., & Thompson, D. (1952). A generalization of sampling without replacement from a finite universe. Journal of the American Statistical Association, 47, 663-685.

Isaki, C., & Fuller, W. (1982). Survey Design Under the Regression Superpopulation Model. Journal of the American Statistical Association, 77 (377), 89-96.

Nambeu, C.-O, Olson, E., Yin, L. (2024). Provisional Estimates of Poverty Rates in Canada. Internal Document. Advisory Committee on Statistical Methods. Meeting No.75.

Rao, J., & Wu, C. (1988). Resampling Inference with Complex Survey Data. Journal of American Statistical Association, 83, 231-241.

Särndal, C.-E., Swensson, B., &Wretman, J. (1992). Model Assisted Survey Sampling. New York: Springer-Verlag.

Statistics Canada. Poverty Reduction: Canada’s Poverty Reduction Strategy Dashboard. Government of Canada, 2021. Accessed 15 Oct. 2024.