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.
- Data source, sample, and key concepts
- Summary measures of the working lives of Canadian baby boomers
- Job stability through the 1990s and 2000s
- Conclusion and implications
While decisions about when and how to retire are made towards the end of one’s working life, these decisions are shaped and constrained by the circumstances, choices, and events encountered over a life time. Accumulated wealth at age 55 or 60, for example, reflects employment choices and opportunities, savings decisions, public institutions, and macroeconomic conditions over previous decades. In the language of the life course perspective, “…to understand behaviour at any one life stage requires knowledge of prior transitions and trajectories” (Han and Moen 1999, p. 197). In this respect, information on the employment experiences and transitions of individuals over their working lives is relevant to understanding their retirement choices and options later on.
With the leading edge of the baby boom generation now in their mid-sixties, there is considerable interest in how and when these individuals will retire. To help place this issue in a broader context, the objective of this paper is to provide information on the employment histories of these individuals over a substantial portion of their working lives. Drawing on the strengths of the administrative data being used, the analysis is organized around the concepts of job tenure and employment mobility, and examines the employment history which older baby boomers are transitioning into retirement. The main questions are the following:
- What proportion of baby boomers are approaching retirement after having spent most of their careers or working lives in long-term employment with a single firm or organization? Conversely, to what extent have the working lives of older baby boomers been characterized by job mobility?
- How do the financial characteristics of individuals with low and high levels of job mobility compare?
- Is long-term employment becoming less prevalent?
This focus on employment duration is warranted for several reasons. First, long-term employment (or career jobs) is often the conceptual and empirical starting point for discussions of retirement transitions. Traditional (male) retirement, for example, is often understood as a single, clearly demarcated exit from long-term employment, while bridge jobs are typically defined as paid positions taken “…after an older worker has officially retired from a long-term career job, but before…complete withdrawal from the workforce...” (Dendinger et al. 2005, p. 22—emphasis added). Second, employment duration has implications for the expectations and certainty regarding retirement. For example, after controlling for a wide range of factors, including pension coverage and income, Ostrovsky and Schellenberg (2008) found a positive correlation between job tenure and plans for early retirement, certainty regarding retirement plans, and confidence in the adequacy of retirement savings. Similarly, Han and Moen (1999) found a positive correlation between orderly careers and both the expected and actual timing of retirement (p. 224). Third, employment duration and mobility are relevant to current discussions regarding Canada’s retirement income system. The portability of retirement savings vehicles and the extent to which they meet the needs of a mobile workforce is a salient theme in this discussion. How long individuals work for a single firm, or conversely, how often they change employers, is relevant in this context.
The longest duration of employment held by individuals is the central organizing concept in this analysis. Workers are categorized according to the longest job in which they are observed during a 28-year period with summary measures provided on their annual earnings, cumulative earnings, job turnover, layoffs, and pensionable service. The goal is to better understand the circumstances in which older Canadians are making retirement decisions.
The remainder of the paper is organized into several sections. Following a discussion of the data, sample selection, and key concepts in Section 2, a descriptive profile of employment histories is provided in Section 3. In Section 4, multivariate results examining whether employment duration has changed across successive cohorts of older workers is presented. Conclusions follow in Section 5.
2 Data source, sample, and key concepts
2.1 Data source
The Longitudinal Worker File (LWF) is an administrative database designed to provide information on employment dynamics in Canada. It is based on a 10% random sample of Canadian workers, and contains information on their demographic and income characteristics, on the jobs they hold, and on the firms in which they work. The LWF is longitudinal, providing information from 1983 to 2010. 1 The LWF’s longitudinal structure, coupled with information allowing workers’ movements from employer to employer to be identified, makes the LWF particularly well-suited to studying employment transitions over the life course. The long reference period, large sample size, and high-quality earnings data are further strengths.
Despite its strengths, a limitation of the LWF is that it does not contain information on some personal characteristics generally included in retirement studies, such as health status, occupation, and educational attainment. Consequently, the extent to which variations in employment outcomes are attributable to these characteristics cannot be specified. Similarly, information is not available on the number of hours worked each week or the number of weeks worked each year. Hence, earnings for each job held are observed on an annual basis only, and employment transitions must be defined on an annual, rather than monthly, basis. Furthermore, the LWF is structured using jobs and individuals as the units of analysis; information on the employment or earnings of couples or households is not available. Hence, individuals who are identified as having sporadic labour force attachment or low annual earnings may (or may not) be married to a higher-earnings spouse. Finally, although the LWF can be used to estimate individuals’ cumulative earnings and years of pensionable service, it does not contain direct measures of wealth, retirement savings, or the adequacy of retirement preparations. The issues are beyond the scope of this paper.
2.2 Sample and key concepts
The sample consists of Canadians who were aged 33 to 38 in 1983. These individuals comprise the first wave of the baby boom generation, born between 1945 and 1950 and reaching age 60 to 65 in 2010. 2 They were in their mid-thirties at the beginning of the reference period and by that point had likely completed their postsecondary education and made their way through entry-level positions at the start of their careers. Over the next 28 years they were in the prime of their working lives, and by the end of the period were close to retirement or had already retired.
Given the objective of the analysis, the sample is limited to individuals who had at least some employment at the start of the reference period; specifically, those who had T4 slips issued by their employers or reported non-zero net self-employment income on their T1 tax form in any one of 1983, 1984, 1985, or 1986. Some 11% of men and 20% of women were excluded on this basis. From the remaining sample, individuals who died between 1983 and 2010 (as indicated on their T1 tax files) were excluded, representing some 8% of men and 5% of women. Comparisons with Census-based population estimates show that the resulting sample comprises 86% of the total population in that age range (Table 1). 3 It is also important to point out that the sample excludes immigrants in the same birth cohort who had employment income in Canada only after 1986. It is also impossible to identify individuals who may have worked in Canada in the early 1980s but subsequently left the country to work abroad for any number of years. These individuals remain in the sample.
The observed duration of employment spells is the central organizing concept for the analysis. An employment spell is the number of consecutive years an individual is observed receiving T4 earnings from a specific firm or organization or income from self-employment. 4 The variable has an observed range of 1 to 28. The calculation of duration in this study allows for one-year breaks in employer-specific earnings, provided that receipt of earnings from the same employer resumes the following year. In other words, an individual observed with, say, three consecutive years of earnings from a specific employer, followed by a year with no earnings from that employer, and then a subsequent year with earnings from that same employer, is defined as having continuous employment with that firm. This allows for the possibility that individuals take time off to have children, raise families, return to school, attend to health issues, and so on. 5 Nonetheless, when two or more consecutive years go by with no earnings from the employer and then earnings subsequently resume, a second employment spell is said to start. 6 In the discussion below, the terms ‘employment spell’ and ‘job’ are used interchangeably.
For each individual in the sample, the lengths of all employment spells are calculated, and the longest one is flagged. If two employment spells have the same length, the one held most recently is flagged. If multiple spells with the longest duration exist concurrently and end in the same year, the one with the higher cumulative earnings is taken as the longest job.
The distributions of men and women across the longest employment spell observed are shown in Chart 1. Somewhat larger shares of men than women are observed in employment spells lasting 19 or more years. 7 Still, the distributions are not all that different between men and women, reflecting the extent to which women in this generation spent a large portion of their adult lives in the workforce. 8
The shapes of the distributions do not point to any obvious thresholds around which job duration categories could be defined. However, a 12-year threshold is purposefully applied in order to facilitate comparisons with the research literature. Many studies of retirement take as their starting point the ‘career jobs’, or ‘long-term jobs’, held by older workers, defined as jobs lasting ten years or more (e.g.: Cahill et al. 2006; Weckerle and Shultz 1999). 9 Given the annual nature of LWF data, an employment spell observed over ten consecutive years may have lasted little more than eight years if, for example, it commenced in December of year 1 and ended in January of year 10. To better approximate a ten-year threshold, employment spells lasting less than, or more than, 12 years are identified. The sample is further subdivided into groupings of approximately six years, as shown in Table 2. The intent is not to stake out and defend a specific definition of long-term employment, but rather to use employment duration as a transparent framework within which to present other employment measures. The use of alternative duration categories would yield estimates on the summary measures that would be predictably different.
Distributions across the categories are shown in Table 2. 10 At the low end, 12% of the cohort had a longest employment spell of one to five years, while at the high end 17% had a longest spell of 25 to 28 years. Using the Health and Retirement Study in the United States, Cahill et al. (2006) reported that 64.6% of Americans aged 51 to 61 in 1992 with any work experience past age 49 worked in long-term jobs (defined as those lasting 10 years or more with at least 1600 hours worked per year, i.e., full-time jobs). Applying the 12-year threshold to the LWF yields an incidence of long-term employment of 65% in Canada—67% among men and 62% among women. 11
Another benchmark against which to compare the above distribution is job tenure among employed respondents in the Labour Force Survey (LFS). Among employed (LFS) respondents aged 55 to 64 in 2010, 58% of men and 56% of women were working for an employer they had been with for 11 years or longer (Table 3).
3 Summary measures of the working lives of Canadian baby boomers
Using the duration of longest employment as an organizing framework, summary information on the employment experiences of older baby boomers is presented in Tables 4 and 5, for men and women, respectively.
Beginning with individuals whose longest job lasted five years or less, summary measures indicate that labour force attachment was typically quite limited. On average, individuals in this group had earnings of $1,000 or more in about 7 to 9 years over the period, and average annual earnings of the median worker during these years were modest, at about $16,300 among men and $8,600 among women. Together, limited employment and modest annual earnings yield modest cumulative earnings over the 28-year period, with medians of around $111,200 and $57,000 among men and women, respectively. By definition, all jobs were fairly short. The longest observed employment spell averaged 3.3 years, while the second-longest averaged about 2.5 years (conditional on a 2nd job being observed). In fact, of all jobs held by individuals in this group, over half (57%) were no more than one year in (observed) duration. Limited attachment to the workforce is further evidenced by the fact that individuals in this group typically exited their longest job at age 45, and were last observed with earnings at about 49 years of age. The Longitudinal Worker File (LWF) contains little information on why they left the workforce. Overall, these individuals had fairly marginal attachment to the workforce through the latter half of their working lives.
The second group, consisting of individuals whose longest job was observed for 6 to 11 years, differs in important ways from the previous one. Although the average duration of the longest job was less than 9 years, individuals in this group had fairly sustained involvement in the workforce. On average, they had earnings of $1,000 or more in almost 20 of the 28 years, and were last observed with earnings at about 58 years of age. Yet, in spite of this workforce attachment, individuals in this group typically worked for almost nine different employers over the period. In this respect, they experienced considerable employment mobility. Some of their jobs ended involuntarily, with layoffs more prevalent among men than among women. On average, men in this group experienced 2.8 permanent layoffs and 2.6 temporary layoffs over the period, 12 and received Employment Insurance income in 5.7 years. During the years they worked, average annual earnings of the median worker in this group were about $36,200 among men and $19,500 among women. 13 Further up the earnings distribution (80th percentile), men appear to have fared reasonably well, with annual earnings of about $60,300 and total cumulative earnings of $1.3 million. Overall, individuals in this group may be considered ‘mobile workers’, with sustained workforce participation and relatively high job mobility over the reference period.
The next three groups include individuals whose longest observed employment spell was 12 years or longer. By definition, individuals in these jobs had sustained attachment to the workforce, with earnings of $1,000 or more typically received in at least 24 of the 28 years. There are variations in the observed duration of the longest job within the three duration groups, from about 15 years at the low-end to 27 years at the high-end. However, the issue of left-hand censoring is more influential among these long-term-job holders than among the groups discussed above, as 36.8% to 85.4% of the longest employment spells observed among men were already in progress when first observed at the start of the reference period (in 1983). Similarly, right-hand censoring is also a consideration as 26.5% to 69.9% of the longest employment spells observed among men were still in progress at the end of the reference period (in 2010). Consequently, the observed duration of these long-term jobs underestimates their actual duration, but by how much cannot be determined.
Considering employment mobility, long-term-job holders typically worked for about three to six different employers over the period, including jobs held before or after their long-term job as well as jobs held concurrently with it, as in the case of holding multiple jobs, or ‘moonlighting’. Layoffs were typically not a recurring event during the working lives of long-term-job holders, with men experiencing 0.3 to 1.3 (permanent) layoffs over the 26 years and women experiencing 0.2 to 0.6 layoffs. Nonetheless, about 1% to 7% of individuals saw their long-term job end as a result of a layoff.
Among men in long-term jobs, average annual earnings of the median worker ranged from about $49,000 to $56,000, and median cumulative earnings ranged from $1.1 million to $1.4 million. Towards the upper end of the distribution (80th percentile), average annual earnings of men in long-term jobs ranged from $72,000 to almost $80,000. Annual and cumulative earnings among women are below those of men, but above those of women with more mobile or marginal workforce attachment. Job quality is also evident in terms of inclusion in a registered pension plan (RPP). While mobile workers typically accumulated 3 to 4 years of pensionable service over the 20 years for which data are available, longer-term-job holders typically accumulated 7 to 13 years. Among those whose longest job is observed for 19 or more years, accumulated pensionable service averaged 10 to 13 years (again, out of the 20 years for which data are available).
Among men, about one-fifth to one-quarter of those in long-term jobs were employed in the public sector (i.e., federal, provincial, or municipal government) or the near-public sector (i.e., education, health care, and social services). Conversely, about three-quarters of them were employed in the private sector, broadly defined as all other sectors. The public and near-public sectors were a more important site of long-term employment for women in the baby boom generation. Indeed, across the three duration categories of 12 years or more, 44% to 54% of women worked in these sectors.
Overall, the majority of the baby boom generation spent their working lives in what is termed ‘long-term employment’ in this study. When this term is defined using a restrictive definition—19 or more years observed with the same employer—about one-third of the sample is included in this group. These individuals appear to be approaching retirement on a ‘stable employment platform’ characterized by sustained employment with a single firm, relatively few layoffs or job changes, relatively high life-time earnings, and, in many cases, sustained RPP coverage. When defined using a less restrictive definition—12 years or more observed with the same employer—about two-thirds of the sample had long-term employment (67% of men and 62% of women). This less restrictive definition is accompanied by greater variability across summary measures. Note again that individuals in this study are first observed during their 30s, and therefore when they are past the age of highest job mobility. Morissette et al. (2013) show that hiring rates are around 51% for men and women aged 15 to 24, but drop to around 26% among individuals aged 25 to 34, and further to around 17% among those aged 35 to 44. 14
4 Job stability through the 1990s and 2000s
Given the considerable differences in the summary measures across these employment duration groups, one question that arises is whether the incidence of long-term employment has changed over time. Are younger baby boomers any more, or any less, likely than older baby boomers to enter their fifties in long-term employment? To answer this question, a discrete-time hazard model with a nonparametric baseline hazard function is estimated using a standard logistic regression framework. The model includes a series of dummies for duration of a job in years (the baseline hazard function), a series of dummies for the calendar year in which a job started, interaction terms between the duration and year dummies, a series of dummies for worker’s age at the start of the job and a gender dummy.
A slightly different definition of an employment spell is used in this than in earlier sections, as a job is treated as having ended in year t if it is observed in year t but not in year t+1. Any subsequent spells that a worker has with the same employer are excluded from analysis. Information from year 2010 is used to determine job status in 2009. All job spells observed in both 2009 and 2010 are censored at 2009. The sample for this analysis consists of all new job spells observed between 1985 and 2009. 15 By focusing on new spells, the left-censoring problem associated with employment spells in progress is avoided. Right-censoring is taken into account in one portion of the likelihood function of the model. Self-employment spells are excluded from analysis. The sample is also expanded to include workers aged 25 to 50 at the start of the spell.
Using the estimated coefficients, survival functions are computed for jobs starting in different years. The full set of survival rates is presented in Tables 7 and 8. Chart 2 presents selected survival curves for starting points which allow jobs to be (potentially) observed until at least year 12, the threshold for long-term jobs adopted in this study. The results suggest that between 5.3% and 6.5% of jobs that started between 1985 and 1997 lasted at least 12 years. This estimate jumps to 7.6% in 1998, representing jobs that reached their 12th year in 2009. The key message, however, is that the likelihood of a newly created employer–employee match resulting in a long-term employment spell did not decline between the mid-1980s and the late 1990s.
The most recent start year from which a job can be observed in the LWF until its 12th year is 1998; but what about jobs that started more recently than that? One might argue that, if no substantial changes over time are observed in the survival rate of jobs at lower durations, there is no reason to suspect a substantial change at year 12, all else equal. To get a sense of whether the probability of a job surviving to year 12 is likely to have changed among more recent job starts, the paper examines trends in the survival rates of jobs observed for fewer years. Chart 3 shows the percentage of spells that survive to at least n years, by year in which the jobs started. The results suggest that the share of jobs that survived to year 8 is no smaller among jobs that started in the beginning of the 2000s than among those that started in the 1990s or the second half of the 1980s. Similarly, there is no decline in the share of jobs that made it to at least year 5. These results are consistent with Brochu (2012) and Morissette et al. (2012). The evidence in Chart 3 also suggests that the higher rate of survival to year 12 among jobs that started in 1998 is a single-year phenomenon, as survival rates at shorter durations (e.g., 2 years and 5 years) are also higher that year but return to levels consistent with historical trends in subsequent years.
5 Conclusion and implications
Using observed duration of individuals’ longest job as an organizing framework, this paper has provided an overview of the working lives of Canadians at the leading edge of the baby boom generation.
About two-thirds of the baby boomers entered their fifties in a long-term job, defined as employment that had lasted 12 years or more. In fact, most had worked for the same firm or organization for far longer—often twenty years or more—and even this is an underestimation since most started their longest job prior to the start of our reference period. The prevalence and duration of long-term jobs among this cohort suggest that the working lives of most baby boomers were quite stable. The fact that at least 50% of long-term-job holders did not experience a single permanent or temporary layoff over the 28-year period is further evidence of this stability. Although hiring and separation rates calculated for the entire workforce show a high degree of worker reallocation, this is most prevalent among workers under 35 years of age (Morissette et al. 2013). By the time individuals are age 35 or older, most have sorted themselves into jobs that they are likely to hold for many years.
This has a number of implications. For researchers investigating the employment dynamics of older workers, the termination of a long-term job does provide a clean conceptual and empirical starting point. Simply put, what do older workers do after leaving the job they have held for much of their working lives? While useful, such a starting point excludes the approximately one-third of the workforce with more mobile or marginal employment histories. Furthermore, there is no reason to believe that retirement decisions and preparedness among workers entering their fifties in long-term employment can be generalized to all workers. The retirement transitions of those more mobile workers may be of particular interest to policy makers.
About one-quarter of the sample consists of what are termed ‘mobile workers’ in this study. In spite of their sustained attachment to the workforce, these individuals had more employment changes than the long-term job holders above. This mobility is evident throughout their working lives, with multiple employment changes evident among both men and women through their thirties, forties, and fifties. Compared to long-term-job holders, mobile workers have lower annual earnings, lower cumulative earnings, and fewer years of pensionable service. Nonetheless, mobile workers at the upper end of the earnings distribution appear to have fared reasonably well in terms of annual and cumulative earnings.
As well, just over one-tenth of the sample consists of what are termed ‘marginally attached workers’ in this study. These individuals had limited participation in the workforce, short-term employment, and low annual and cumulative earnings.
Finally, evidence from the Longitudinal Worker File does not indicate that there has been any decline over time in the incidence of long-term employment or, conversely, an increase in the extent to which individuals’ working lives are becoming more mobile.