Catalogue no. 97-563-GWE2006003
Information in the census on income and its sources, including earnings, allows for the compilation of income statistics for all people in Canada, their families and households.
Governments use income statistics to develop income support programs and social services, such as the Canada Child Tax Benefit, Employment Insurance, provincial income supplements, and welfare payments. They also use these statistics to ensure that programs supplementing family incomes do so efficiently and to identify specific geographic areas that need assistance.
Businesses, large and small, use these statistics to locate stores near consumers and to develop new products and services.
Private sector and public sector researchers as well as academics often make use of earnings data from the census to study labour markets and industrial patterns.
Income data were collected during the 2006 Census for the non-institutional population comprised of persons aged 15 years or over. All income received during the preceding calendar year (2005) was to be reported, even nontaxable income, except: capital gains and losses; withdrawals from registered retirement Savings Plan (RRSP) and other savings plans to cover living expenses or to pay off debts; inheritances received; lottery winnings and lump sum insurance settlements.
All income variables collected by the census or derived by aggregation of component sources are defined in the 2006 Census Dictionary.
The sum of income received during 2005 from all different income sources together produces the variable:
To assist in respondent recall when filling out the questionnaire and provide a richer dataset for analysis, personal income was classified into income source categories.
For further explanation of the interdependency of the components, please refer to the following figure: Components of income in 2005: A related derived variable is the:
New to the 2006 Census is the addition of a variable for income tax paid. This allows for the calculation of after-tax income, arguably a more useful measure of funds available to a household, family or individual for consumption, savings and investment.
Although total income and after-tax income are collected at the individual level, since income is often pooled or shared, these are also derived for Canadians in private households at various other levels of aggregation:
Definitions for Census family, Economic family and Household explain the various concepts and the figure Economic and census family membership and family status describes in a concise way the relationships and classifications of people at each aggregation level.
Measures of central tendency (for example, mean or median) for all income variables and the standard error of the mean due to sampling can be calculated by our tabulation system in a standard way. Notes on the methodology behind the derivation of these statistics at the population level and other levels of aggregation are available in Appendix Q Derived Statistics.
In standard Statistics Canada tables, adjustments were made to better reflect the context of income data, such as factoring for inflation for the purpose of comparability between years (Constant dollars income).
A new variable available only in special tabulations has also been produced that adjusts total income and after-tax income for family size and composition for the purpose of comparability between families as well as with persons not in families.
Total income, after-tax income, employment income and all other income variables collected or derived for the 2006 Census are for calendar year 2005. Two other labour-related variables also have that same reference period and can usually be employed in a relatively coherent framework with the earnings field or other income data: Weeks worked in 2005 and Full-time or part-time weeks worked in 2005. A data quality note signals differences in coherence compared to prior censuses.
However, some inconsistencies may become apparent when using income variables with other labour variables and some care must thus be exercised. For the following variables the reference period is the week prior to Census (May 7 to 13, 2006) and not calendar year 2005: Hours worked for pay or in self-employment and Labour force activity.
Further, some labour variables refer to the job held in the week prior to the census or to the most recent job held since January 1, 2005, so they may or may not correspond to the economic activity seen in the Earnings or employment income field or its sub-components: Class of worker; Industry (based on the 2002 North American Industry Classification System [NAICS]) and Occupation (based on the 2006 National Occupational Classification for Statistics [NOC-S 2006]).
Income data is used traditionally with Owner's major payments and Rent, gross to compute one other standard census variable, Owner's major payments or gross rent as a percentage of household income. Minor inconsistencies arise as these shelter cost variables, as well as their components (Condominium fees, Annual payment for electricity, Annual payment for oil, gas, coal, wood or other fuels, Annual payment for water and other municipal services, Annual property taxes, Monthly mortgage payment and Rent, monthly cash), are sometimes collected for the most recent month or for the last 12 months before Census Day, whereas the income data are always for the previous calendar year.
Income data for private households were obtained from questions 51 or 52 on the long census questionnaires (2B, 2C and 2D). The 2B questionnaires were used to enumerate a 20% sample of all private households in Canada, while the 2C questionnaires were used to enumerate Canadians living abroad and the 2D questionnaires were used to enumerate private households on Indian reserves and in remote areas.
Question 51, new for 2006, gave respondents the option of allowing Statistics Canada permission to obtain income information from 2005 income tax files instead of completing question 52. For persons that marked 'Yes,' Statistics Canada attempted to find them in a file received from the Canada Revenue Agency and, if found, combined the information from various tax items to produce responses to all parts of question 52.
Question 52 asked respondents to indicate the presence and amount of income from various sources as well as total income. For the 2006 Census, two new items were added: one for Child benefits and one for Income tax paid.
This new mixed method of collection with part of the data coming from tax files and part of the data coming directly from the respondent introduces some issues of historical comparability. However, it permits greater comparability with other Statistics Canada sources of income data. The sections Data quality and Historical Comparability provide more information on these issues.
For the non-institutional population comprised of persons aged 15 years or over, some classifications used in several tables were based on the following income characteristics:
Once total income and after-tax income are aggregated to economic family level, economic families and persons not in economic families were classified according to low-income status. Respondents reporting income below a certain threshold, defined by family size and family status, and the size of area of residence, are classified as low income.
These variables are important in identifying the prevalence of low-income and the social, economic and geographic characteristics of those identified as members of low income families:
The actual threshold amounts for calendar year 2005 are supplied in the Low income after-tax cut-offs (1992 base) for economic families and persons not in economic families, 2005 and Low income before tax cut-offs (1992 base) for economic families and persons not in economic families, 2005 reference tables.
Users should be cautious when comparing 2006 Census data with previous census years as some variables have changed slightly in 2006 and some may not have retained comparability over time under all circumstances. Please see the section below on data quality for more information.
Throughout the census-taking process, every effort was made to ensure high-quality results. Rigorous quality standards were set for data collection and processing and the public communications program assisted in minimising non-response.
Although considerable effort is made throughout the entire process to ensure high standards of data quality, the resulting data are subject to a certain degree of inaccuracy.
To assess the usefulness of census data for their purposes and to understand the risk involved in drawing conclusions or making decisions on the basis of these data, users should be aware of their inaccuracies and appreciate their origin and composition. This reference guide, specific to income and earnings will provide the basic information required but other methods have also been used to inform the users. Changes to the methodology and their impacts were summarily described in a short Info-Census document published May 1, 2008; data quality notes were attached to specific tables and products (accessible through a footnote linked to the table title in web-based products or via the File/Summary in Beyond 20/20 tables); other technical documents related to coverage, sampling and weighting, aboriginal people and education variables are published or planned and for the average of any dollar amount, we systematically produce the standard error of the average to inform users of the estimated impact of sampling variability.
The evaluation of income variables consisted of the following at the national level and where possible at the provincial and territorial levels:
Household non-response was slightly higher in 2006 compared to 2001. For persons, we believe this translates into 2.45%. This affects the coverage and possibly the distribution of variables. We have adjusted the long form database to cover the missed households and people (756,730) but if the characteristics of the missed are different from those that responded that we used to adjust, there might be some bias introduced in the data.
Net undercoverage was also measured by a variety of means and that exercise has shown both components (overcoverage and undercoverage) rising in size, though the amplitude of the net effect on population counts (2.67%) is lessened when compared to 2001 (2.99%). The Census has missed more people it should have counted than before but we have also more duplicates than before in the database.
Specific sub-parts of the income question had different response rates. Item-by-item response rates were all better compared to the 2001 Census except for the wages and salaries variable which was slightly worse.
|(a) Total wages and salaries||83.1||79.3|
|(b) Net farm income||74.3||80.3|
|(c) Self-employment income||73.1||80.2|
|(d) Child benefits1||.. 2||68.9|
|(e) OAS, GIS and Allowance||80.1||80.6|
|(f) CPP, QPP||79.6||80.6|
|(g) Employment Insurance||77.6||80.4|
|(h) Other income from government sources||78.7||80.6|
|(i) Dividends, interest and other investment income||78.6||80.3|
|(j) Retirement pensions, superannuation, and annuities||77.8||80.3|
|(k) Other Money income||77.0||80.2|
|Total Income in prior year from all sources||78.7||67.4|
|Income Tax Paid on prior year's income1||..2||76.6|
For respondents providing income data via the internet, an automated prompt was made to verify their response if it fell outside of usual parameters. This was to help identify situations where the respondent might have entered a non-annual amount or a very high amount by mistake. There was also a summary screen that gave the respondent the opportunity to review their answers and go back to change them.
The validation prompts on the internet were activated for 42,460 fields. Editing by the respondents reduced the count of unusual fields by 10,527 or approximately one quarter of the time.
For the income questions, certain errors, if left uncorrected, could lead to distortions which could have serious repercussions on the quality and credibility of census income data. For example, a respondent-provided amount of $90,000 in wages and salaries could be entered erroneously with an additional zero, changing the original amount to $900,000. Similarly, an amount of $9,000 in Employment Insurance benefits could be erroneously keyed in as $90,000. A few errors of this magnitude in the first source could quickly add millions of dollars to wage estimates, while similar errors in the second source could lead to obvious errors in the estimates.
To safeguard against such errors, an online editing system was established as in previous censuses that checked all respondent-provided amount entries against specified limits. If an amount reported on paper or via the internet was in excess of the specified limits, the relevant source was highlighted on an electronic display of the income question, along with various individual characteristics (sex, age, education, weeks worked, etc.), to assist in validation of the response. In many cases, it was also necessary to examine the electronic image of the questionnaire. Responses were then either accepted as reported or modified as required on the database. Some 64,750 amounts were examined which resulted in a fairly high rate of changed data. This rate is much higher than in prior censuses due to difficulties with the capture system.
The objective of edit and imputation was not to 'create' data but to ensure the reasonable accuracy and consistency of the data supplied by the respondents. With this objective in view during each of the many phases of edit and imputation, a record was kept of all changes made to the data. The table below gives the percentage change between the original and the final number of income recipients, the aggregate amount received from different income sources and the average amount received.
count of recipients
|(a) Total wages and salaries||21.3||12.7||17.8||10.0||-2.9||-2.4|
|(b) Net farm income||31.0||-0.2||25.6||-23.6||-4.1||-23.5|
|(c) Self-employment income||7.2||10.1||-7.0||11.4||-13.3||1.2|
|(d) Child benefits 1||..2||770.5||..2||984.0||..2||24.5|
|(e) OAS, GIS and Allowance||23.9||10.7||57.4||16.2||27.0||5.0|
|(f) CPP, QPP||46.0||10.9||45.6||10.5||-0.2||-0.3|
|(g) Employment Insurance||68.7||12.0||67.6||12.3||-0.6||0.2|
|(h) Other income from government sources||182.9||127.7||83.6||58.5||-35.1||-30.4|
|(i) Dividends, interest and other investment income||25.1||10.6||23.4||10.2||-1.3||-0.4|
|(j) Retirement pensions, superannuation, and annuities||27.9||9.8||25.8||9.2||-1.7||-0.6|
|(k) Other Money income||25.8||12.0||26.0||12.2||0.2||0.1|
|Total income in 2005 from all sources||17.0||38.8||16.9||29.7||-0.1||-6.6|
|Income tax paid on 2005 income 1||..2||13.5||..2||21.3||..2||6.9|
The table shows the impact of edit and imputation on both the number of recipients and aggregate income from each source. At the end of the process, the number of income recipients as well as the aggregate amount of income had increased by about 28%. This is largely due to the tax data not providing Child benefits, so a large proportion of persons sourced from tax data had the child benefits field empty and total income missing. In general, the proportion of income assigned for most sources is commensurate with the proportion of records imputed. Farm income is one source where it does not hold as mainly large amounts were transferred to other sources to make the responses more consistent with the labour variables.
In 2006, for the first time Canadians had the option of granting permission to retrieve income information directly from their tax records. This reduced respondent burden and improved the quality of the income data. Those who did not select this option were required to provide the income information on the paper form or via the Internet. The tax option was very popular: 82.4% of the population aged 15 and over agreed to the method.
Due to the precision of income amounts in tax records, there is less rounding in the data. As well, the large proportion of tax-sourced data has reduced the tendency of counts to cluster around 'round' dollar amounts such as $10,000, $25,000 or $30,000.
Counts are now more evenly spread out, which may cause detailed income distributions to be more variable when compared to data from a previous year on a bracket-by-bracket basis. As a result, comparison of detailed 2006 Census income distributions with data from previous censuses should be done with caution.
|Employment income amounts
Dollar amount ends in :
|01 to 99||34.2||87.2|
The second large difference with the new response modes introduced in the 2006 Census is the increased reporting of small amounts. The tax system has built-in reminders and verification mechanisms that together manage to encourage reporting. Therefore, records from tax filers have higher propensity to include small amounts.
This increase in small amounts of income in the 2006 Census increases the number of persons with earnings and modifies the distribution and the calculated statistics. As a result, averages and medians from earlier censuses appear larger when looking at populations for whom small amounts are a frequent occurrence. This change is seen most often in earnings (such as wages and salaries and net income from self-employment) but may also be present in other sources such as employment insurance benefits and investment income.
Full-year, full-time earners are far less subject to this effect as most had accurately reported their earnings in previous censuses, which minimizes the reporting differences over time and by response mode (paper and Internet versus tax records). Statistics Canada recommends comparing the earnings of full-year, full-time workers when possible. For other groups, these comparisons must be done with caution.
As with data product, the quality of the 2006 Census income information released was evaluated internally prior to publication. The income and earnings data were compared, as much as possible, with alternative data sources. The main sources of comparison were the estimates from the System of National Accounts and the Provincial Economic Accounts, the Survey of Labour and Income Dynamics, as well as the Longitudinal and Administrative Databank.
Census estimates of aggregate income in 2005 were compared to similar personal income estimates from the national accounts. After adjustments to the personal income estimates for differences in concepts and coverage, the census estimate of aggregate income in 2005 from comparable sources was 1.2% lower than the national accounts estimate. As in the past, census estimates for some income components and for some provinces compared more favourably than for others.
Census estimates of aggregate wages and salaries, the largest component of income, were slightly higher (1.0%) than the national accounts estimates. This was partially offset by the difference (-7.8%) between the census estimates of aggregate self-employment income from both farm and non-farm self-employment and the adjusted national accounts figures. Overall, estimates of aggregate employment income or earnings were nearly identical (0.3% difference).
Census estimates of Old Age Security pensions and the Guaranteed Income Supplement were slightly lower (-1.4%), as they were for Canada/Quebec Pension Plan benefits (-0.9%), than adjusted national accounts estimates. Employment Insurance benefits reported in the census were smaller by 6.1%. Census estimates of aggregate Child benefits were 2.0% higher than the adjusted national accounts estimates. Census estimates of other government transfer payments, which include such items as social welfare benefits, provincial income supplements to seniors, veterans' pensions and GST/HST/QST refunds, were significantly below (-39.2%) the estimates from the national accounts. Overall, census estimates of aggregate income from all government transfer payments were lower by 12.0%. The census estimate of aggregate investment income in 2005 was slightly lower (-2.7%) than the comparable national accounts estimate. This is a significant improvement when compared to previous census comparisons.
Census income statistics were also compared with similar statistics from the annual Survey of Labour and Income Dynamics (SLID). SLID estimates reflect adjustments made for population undercoverage, while census estimates do not include such an adjustment. This adjustment contributes to census estimates showing fewer income recipients (-2.1%) and earners (-1.4%) than SLID estimates. However, due to higher average amounts, census estimates of aggregate earnings are 2.8% higher than the SLID estimate, while the census estimate of aggregate total income of individuals is 2.3% higher. Most of the observed provincial differences were considered acceptable in the light of sampling errors in the survey. The all-person low income prevalence rates for Canada (excluding the Territories) were almost identical in both sources for the before-tax measure at 15.3% and only slightly higher (0.6 percentage points) in census than SLID for the after-tax rate.
Users should be cautioned that Statistics Canada income definitions do not always match concepts used by other organisations. In particular, while the use of tax data has forced the harmonisation of certain concepts, the definition of Total income in the census still does not correspond to that used by the Canada Revenue Agency.
One should also note that the populations covered differ from that in other surveys, the National Accounts and administrative results. For more information on the comparability of Census income data with income from other data sources, please refer to Comparing Income Statistics from Different Sources: Aggregate Income, 2005.
The 2006 Census underwent significant changes, including expanded income content, new collection methodology and processing techniques for the income questions, and slight conceptual changes to existing questions. As a result, there is an impact on historical comparability for some variables. This section outlines some of the impacts of which users should be aware.
The biggest impact is the mixed mode of collection: the presence of tax data and respondent-provided data has provoked significant differences in a few areas. These have for the most part been discussed in an earlier section.
The new layout of the questionnaire to facilitate automated capture has also led to many more capture errors and more significant ones but we think the interactive verification steps have caught the most significant of these.
The absence of a long guide delivered with the questionnaire may also have contributed to confusion of what amounts to include and where to include them. It is thought that this may not have had a tremendous impact because of the reliance on tax responses and its availability online for Internet respondents.
For the 2006 Census, some components of income sources were added, deleted or reallocated to align the census income concept with that of Canada Revenue Agency. For example, taxable allowances and benefits are now included as part of wages and salaries. Research grants and royalties, formerly components of other money income, are also now included as part of wages and salaries. Compensation payments from a provincial or territorial government agency for a criminal act or for motor vehicle accident victims, formerly included as part of Other income from government sources, are now excluded. For further information, see the detailed definitions for each source of income.
The census shows a larger count of very high income earners (employment income greater than $1,000,000) when compared to administrative data. While this has a very minimal impact on the medians, it increases slightly the average for the whole population. Income shares when distributed by quintiles or percentiles are also affected.
Considering a larger group of high income earners such as all those earning $100,000 or more or those earning $150,000 or more does not present the same issues and the counts at those levels should be considered reliable and comparable from census to census.
One component of income for which growth from census to census is different from other sources is the variable 'other government income.' The estimates from the 2006 Census are in line with alternate data sources. Comparisons with 2001 should be used with caution as this income component was higher in the census in 2001 than in the other sources. This was the case particularly for families without children and persons not in families.
Farm self-employment income is another component of income that shows abnormal change between the 2001 and 2006 censuses. The census data is now much closer to other sources of data but now shows far lower levels of income and more negative amounts. This is likely mainly due to the introduction of the mixed collection modes in 2006. It is possible that respondents may have always had a tendency to not take into account all deductions such as the capital cost allowance on the census form. The users of the tax option may have been more systematic in mentioning their valid deductions. This renders comparisons from census to census more difficult for this income component.
Changes to the collection methodology for income data and to the editing procedures create an apparent inconsistency for more records in 2006 compared to 2001. For example, a larger amount of full-year, full-time workers (or part-year workers) are without reported earnings (306,510 vs. 170,675). There are also more persons with earnings that do not report work activity during the previous calendar year (1,159,425 vs. 429,265). The table below shows the counts of persons with and without income by work activity and we can see the increase in inconsistency from census to census.
These impacts are also visible for wages and salaries and net income from self employment.
|Presence of employment income||Year||Work activity during reference year|
|Worked full year, full time||Did not work||All others||Total|
|Without employment income||2000||170,675||7,030,150||284,755||7,485,580|
|With employment income||2000||8,685,225||429,265||7,301,295||16,415,780|
In the 2006 Census, individuals who resided in institutions or residences with distinct, separate living quarters, and who were able to complete the census questionnaire, received their own census form to complete.
These individuals were excluded from measurements of income in prior censuses. This census their incomes have been set to zero. This results in a slight overestimation in the count of population 15 years and over, and primarily the age group 65 years and over, without income (or without earnings).
Counts and income statistics for families or persons not in families are not affected, as individuals in these types of collective dwellings have always, and continue to be excluded from those populations.
Confidentiality suppressions are always applied to all areas and statistics to preserve respondents' confidentiality. Data quality suppressions are applied to most products when the estimates are subject to high sampling variability but we recognize that some users may tolerate larger errors to be able to obtain data for extremely small population groups. Thus some basic tables are made available with lower suppression limits. Users are cautioned to always beware of statistics based on cells with small counts. The standard error of the average provides a good indicator of the quality of the average but it only addresses sampling variability and other types of errors do occur.