3 Data

Warning View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Statistics Canada's Longitudinal Administrative Data base (LAD) consists of a random 20% sample of the T1 family file, a yearly cross-sectional file of all taxfilers. Individuals selected for the LAD are linked across years to create a longitudinal profile of each individual. The LAD contains demographic, income and other taxation information for the period from 1982 to 2005, which makes it possible to track individuals for a maximum of 23 years. As a result, it is possible to follow the evolution of the financial situation of individuals after retirement over a long period. Our focus is on six cohorts of Canadians who were aged from 54 to 56 years in 1983, 1986, 1989, 1992, 1995 and 1998 and who earned at least $10,000 at this age (in 2005 constant dollars). We exclude individuals earning less than $10,000 at age 55 since many of them did not file a return at the time.1 This implies that our focus is on individuals who had a significant degree of attachment to the labour market when they were in their mid-50s.

Our six samples (one for each cohort) were constructed as follows. First, individuals who were still alive in 2005 were included if they filed an income return for every year of the period of analysis.2 For instance, individuals from the 1983 cohort were included in the sample if a return was filed every year from 1983 to 2005. Second, individuals who died before 2006 were also included if a return had been filed for all years until the year before they died. For instance, consider an individual who was aged 55 in 1983 and who died in 1995 at the age of 67. To be included in our first sample, a return must have been filed for each of the years 1983 to 1994, which was the last complete year of his/her life. As a result of this process, we obtain six samples with a number of observations ranging from approximately 70,000 in 1983 to 100,000 in 1998 (see Table 1 for more information). Women comprised one third of the sample in 1983, but this share rose to more than 40% in 1998, which is consistent with the higher rates of labour market participation seen among younger cohorts of women. In this paper, we use our first cohort of 1983 most of ten because it covers the longest time period (20 years). The other samples are used only to examine differences across cohorts.

Our measure of income is based on adult-equivalent-adjusted (AEA) family income (on a constant basis), which includes the income of the spouse and all other family members in the Census family unit. For the most part, we use family income after tax because this measure of income is the best approximation of the level of financial well-being experienced by individuals. Our family income values are then adjusted by dividing total family income by the square root of family size to take account of economies of scale that accrue to people who live together in families.3 Finally, income levels by age are calculated on a 'permanent' basis, in order to account for temporary fluctuations that might not be representative of the true financial situation of the family. For example, the permanent income of someone aged 54 was calculated by dividing the sum of income levels reported at age 53, 54 and 55 by three.4 We also tested several alternative definitions of income to assess the robustness of our conclusions. All income figures are expressed in 2005 dollars adjusted with the consumer price index.

The income replacement rate is the standard indicator of welfare loss associated with retirement. We compute replacement rates by age, using permanent income at the beginning of the period (age from 54 to 56) as a benchmark when earnings are typically at their peak.5 In addition to median replacement rates by cohort, we also compute replacement rates across key points in the income distribution, again using permanent income at the beginning of the period as a benchmark to classify individuals across income groups.

1 With the introduction of the Goods and Services Tax in 1986 and the Child Tax Credit in 1992, low-income individuals became more likely to file an income tax return in order to apply for various tax credits. Prior to 1992, low-income individuals had fewer incentives to file. We get similarly defined cohorts by excluding all individuals with less than $10,000 in earnings, which is close to the basic exemption amount that was used for most years in federal tax returns and above which most individuals should be expected to file (which corresponds to approximately 50% of all individuals aged 54 to 56 years old in every cohort). One alternative could have been to include individuals with positive earnings. If this had been the case: (1) coverage would have increased by a little, albeit unequally across cohorts (from 53.1% among those aged from 54 to 56 years old in 1983, to 58.6% in 1998); and (2) our results would have been essentially the same, although replacement rates among low-income individuals would have been slightly higher.

2 It was necessary to exclude these individuals for reasons of consistency. Naturally, fewer individuals were lost in more recent cohorts because individuals were followed over a shorter period of time. In 1983, about 68,800 individuals were included in the final sample (out of 78,900 individuals aged 54 to 56 with at least $10,000 in earnings), which means that about 10,100 were excluded because of reporting problems (12%). In 1998, only 7,800 were excluded, out of 108,400 individuals (about 7% of individuals with at least $10,000 in earnings).

3 Changes in the family composition over time are taken into account in our calculations.

4 Individuals with less than $1,000 in permanent adult-equivalent adjusted income were excluded from our sample, but these amounted to a very tiny portion of the final sample (less than 0.1%).

5. Earnings peak at age 55, but total family income peaks around 60 years of age (see Figure 5).