IN THIS ISSUE:
Introduction | Methodological Issues when using Longitudinal Data | Disability and Employment | Disability, Education and Labour Force Participation
This is the second Disability Information Sheet published by the Canadian Council on Social Development. In response to many information requests received by the CCSD, our first Information Sheet described the major Canadian data sources on persons with disabilities and examined some of the potential research themes that could be explored using those data sources.
This second Information Sheet also covers several frequently asked questions, but it is more technical in nature than Sheet No. 1. In this issue, we examine methodological issues surrounding a disability status variable when using longitudinal data. In the second part of this issue, we examine the relationship between this longitudinal disability status variable and employment from 1993 to 1998. In the final part, we present two tables featuring education and employment for persons with and without disabilities - an area about which we have received countless requests.
People used to assume that disability was a permanent state. However, evidence from longitudinal databases indicates that there is actually a fair amount of change in disability status over time for some individuals. This was first documented using data from the 1989-90 Labour Market Activity Survey (LMAS). This longitudinal survey was one of the first data sources to identify persons with disabilities and survey the same individuals at different points in time. Examining the working-age population - that is, those aged 15 to 64 - we found the following:
Over the two-year period covered by LMAS, disability affected about 18 per cent of the working-age population. However, the disability rate for an individual year - either 1989 or 1990 - was just under 13 per cent.
With six years of new longitudinal data available from the Survey of Labour and Income Dynamics (SLID), a similar "turnover phenomenon" is evident. This raises new issues in the research on disability - issues which need to be resolved, or at least acknowledged, in order to avoid potential confusion about research results based on longitudinal databases.
In the first part of this CCSD Disability Information Sheet, we examine two such issues:
A screening question for disability used on some databases (including the first panel of SLID) offers individuals three choices of response: "Yes," "No," or "Don't know." When working with these data, it is common to limit the analysis to only those persons about whom you have known information - that is, those who answered either "yes" or "no." All others are treated as "missing cases" and eliminated from the analysis. However, examining six years of SLID data suggests that the "don't know" category should be reconsidered, particularly when dealing with the longitudinal file.
Only a very small fraction of individuals in the longitudinal file answered "don't know" to the disability screening question in every year of the six-year period, but about 18 per cent answered "don't know" in at least one year. However, to exclude this 18 per cent of the longitudinal sample poses a number of problems. First, it reduces the sample size and therefore threatens the reliability of estimates derived from the sample. Second, it may introduce a bias in that the excluded population is probably one in which individuals have experienced some degree of activity limitation, but are uncertain about the severity or longevity of their situation. This could mean that persons in the early stages of a disability or those with cyclical disabilities might be under-represented in the analysis.
Looking at the longitudinal panel from SLID for 1993 to 1998, the responses for each category for each year were as follows:
| 1993 | 1994 | 1995 | 1996 | 1997 | 1998 | |
|---|---|---|---|---|---|---|
| Per cent "Yes" (disability) | 10.1 | 10.0 | 10.0 | 13.7 | 15.1 | 16.3 |
| Per cent "No" (no disability) | 80.6 | 85.3 | 85.1 | 82.3 | 81.8 | 82.9 |
| Per cent "Don't know" | 9.3 | 3.5 | 4.0 | 3.2 | 2.6 | 0.8 |
| Per cent "missing" | 0.0 | 1.2 | 0.9 | 0.8 | 0.5 | 0.1 |
| (Note: Based on weighted data for all adults aged 16 and older in 1993; uses the 1998 longitudinal weight) | ||||||
The "missing" category includes records that are truly missing, perhaps due to an individual's withdrawal from the survey, difficulties in locating the individual, or even as a result of the death of the respondent.
It is important to remember that these percentages apply to a cohort of individuals who aged six years over the survey period. Given the link between age and the onset of disability, the incidence of disability would be expected to increase somewhat during this period simply due to the aging of the cohort itself. And as shown above, the percentage of individuals who reported having a disability rose in the final three years of the survey period.
Over the same six-year period, the percentage of people who answered "Don't know" decreased - from 9.3 per cent to 0.8 per cent. Perhaps respondents gained a better understanding of the question from one year to the next. Or perhaps those who were in the early stages of a health problem or those who had an accident required some time for a diagnosis about which they could be certain. It is also evident in the table above that as the percentage who answered "Don't know" decreased, the percentage who responded "No" increased. This suggests that many individuals changed their response from a "Don't know" to a "No" in a subsequent year. In fact, an analysis of all six years of the SLID data suggests that many of the "Don't knows" do convert to "No" at some point, but for many individuals, the situation is not quite that simple.
In some cases, people who responded "Don't know" one year may also have reported a "Yes" in all the other years. In other cases, the response for all other years may have been a "No." In such instances, it is reasonable to assume that the "Yes" or "No" response could be assigned for the entire six-year period.
Many individuals, however, have a more cyclical pattern of answers, with the "Don't know" responses likely representing a transition between having a disability and not having a disability, and vice versa. These individuals could be included in an analysis by assigning a disability status based on the overall pattern of their "Yes" or "No" responses.
Thus, by assuming an individual's disability status for certain years based on the pattern of their responses throughout the six-year survey period, it is possible to return many of the "Don't know" responses back into the analysis.
Another challenge when using longitudinal SLID data for persons with disabilities is how to deal with changes in disability status. Depending on the research question you are examining, a variety of options exist.
One option is to include only those individuals who had a disability in all six years. This yields a very select population of persons with disabilities for analysis and tends to exclude the following groups: individuals who have cyclical disabilities, those in the early stages of degenerative conditions, those who have had accidents or illnesses where the long-term impact is uncertain, and individuals who might have had difficulty receiving a proper diagnosis. This option also yields a much smaller sample size: 3.5 per cent of the population on the six-year longitudinal file, when treating responses of "Don't know" as a "No" response; or, 4.2 per cent when assuming disability status for "Don't knows" based on the pattern of responses in the remaining years.
Another option is to categorize disability status on the basis of the number of years the individual reported a disability during the six-year survey interval. About 25 per cent of the first longitudinal panel - that is, those aged 16 and older in 1993, with "Don't know" responses treated as "No" responses - reported a disability in at least one year during the survey period. To examine this option, we created a four-category variable in which all "Don't know" responses were treated as a "No." The following percentages were the result:
| No disability in any year: | 74.6 per cent |
| Disability in 1 or 2 years: | 12.1 per cent |
| Disability in 3 or 4 years: | 6.9 per cent |
| Disability in 5 or 6 years: | 6.5 per cent |
While this information can be useful when examining questions about work activity during the six-year period, the variable doesn't capture cycles of disability, transitions, or the direction and timing of changes in disability status. Yet these can be important dimensions of a longitudinal variable for disability status, and particularly so when analyzing work patterns.
A third option for a potential variable attempts to characterize changes in disability status in a more dynamic way. When creating this variable, responses of "Don't know" are treated as missing cases only if the individual had more than three "Don't know" responses over the six-year survey period. For other individuals who responded "Don't know" for three or fewer years, disability status is assigned based on their pattern of "Yes" or "No" answers in the other survey years. For example, if an individual reported having no disability for five years and responded "Don't know" in one year, that individual is assumed to have had no disability over the entire interval.
Looking at the SLID data using this third option, we divide individuals according to changes in their disability status. Included in these changes are transitions from one state to another, which are sustained as well as cyclical patterns. The following percentages result:
For this variable, we treat individuals who have changed from one state to the other without returning to the original state as "sustained transitions." However, we must be aware that this applies only to this six-year survey period - we might simply be picking up persons with cyclical disabilities, where the cycles are longer and therefore undetected over the six-year interval. The larger percentage of individuals entering into, rather than exiting from, a state of disability might also be due, in part, to the aging of this cohort.
The CCSD continues to try to perfect a longitudinal variable for disability status and we hope to report on new developments in future Information Sheets. In the next section of this Information Sheet, we present a collapsed version of this variable in an examination of employment patterns.
In this section, we describe an initial effort we made to examine the relationship between disability and employment over time. We used a collapsed version of the longitudinal disability status variable discussed above. Due to sample size restrictions in this particular analysis, we have reduced the longitudinal disability status variable to one which includes the following categories:
From previous research (such as In Unison 2000: Persons with Disabilities in Canada), we know that 1994 and 1995 were difficult years for persons with disabilities in the labour market. One interpretation of these findings is that the recession of the early 1990s might have had a very negative impact on persons with disabilities, even after an economic "recovery" was in progress. We asked the following questions:
Since we are interested in employment issues, we limit our analysis to the following population:
In order to begin to answer the questions posed above, we created an employment variable with the following categories:
Accessibility Note: The webmaster acknowledges that Table 1 still has some potential accessibility problems, particularly for people using screen readers. If you'd prefer to skip over it, you can go straight to the Table 1 summary.
| Table 1: Longitudinal Disability Status by Full-year Employment Patterns from 1993 to 1998, Working-age Persons during 1993-1998 | |||||||
| Longitudinal Disability Status (1993 to 1998) | |||||||
|---|---|---|---|---|---|---|---|
| Persons not disabled in all six years | Persons disabled in five or six of the years | Transitions (sustained change in disability status) | Cycles of Disability Status (not sustained changes in disability status) | Total | |||
| Longitudinal Labour Force Participation | Exit (has disability then does not have disability) | Entry (does not have disability then has disability | 3 or 4 years of disability status | 1 or 2 years of disability status | |||
| Early1 Full-year Employment Loss | 3.5% | 21.2% | -- | 17.6% | 13.6% | 8.0% | 4.9% |
| Middle2 Full-year employment loss | 4.6% | 10.0% | -- | 18.3% | 9.2% | 6.8% | 5.5% |
| Late3 Full-year employment loss | 3.2% | -- | 3.4% | 10.4% | -- | 5.0% | 3.6% |
| Full-year employment all six years | 62% | 39.6% | 63.6% | 43.1% | 53.1% | 60.7% | 59.8% |
| Fluctuating full-year employment | 26.6% | 26.2% | 21.1% | 10.7% | 19.7% | 22.2% | 26.2% |
| Total | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% |
| Number | 7,135,500 | 136,400 | 126,000 | 369,500 | 160,100 | 478,400 | 8,511,000 |
| 1"Early" refers to respondents who lost full-year employment after either 1993 or 1994 | |||||||
| 2 "Middle" refers to respondents who lost full-year employment after either 1995 or 1996. | |||||||
| 3 "Late" refers to respondents who lost full-year employment after 1997. | |||||||
| -- Sample size too small to provide reliable estimate | |||||||
| Source: Prepared by the Canadian Council on Social Development using Statistics Canada's Survey of Labour and Income Dynamics (masterfile). | |||||||
Overall, those without a disability in any of the six years surveyed had the most success at remaining employed all year long. Among this group, 11.3 per cent (3.5 per cent + 4.6 per cent + 3.2 per cent) lost their full-year work without regaining it; 62 per cent continued to be employed full-year all six years; and 26.6 per cent experienced fluctuations.
Among those with a disability, people who exited or left a state of disability and those who had shorter cycles of disability (only one or two years in total with a disability) experienced the most success at remaining employed all year long: 63.6 per cent and 60.7 per cent respectively continued to have a full-year job over the entire interval.
Those with a disability for five or six years during this period had the least success in hanging on to their full-year jobs from 1993: only 39.6 per cent were employed full-time during all six years. For this group, loss of full-year employment was most likely to be experienced early - in 1994 or 1995 - with 21.2 per cent having early loss of full-year employment without regaining it.
The impact of entering a state of disability is also evident, as those who became disabled during this period were the most likely of all to experience a loss of full-year work without regaining it at some point during the survey interval (early, middle, and late loss added together). This group was the least likely to experience a fluctuating pattern of full-year employment, that is, their loss of full-year work was permanent during this period.
Finally, those with longer or more frequent cycles of disability had less success at full-year employment than did those with shorter or less frequent cycles. For example, 13.6 per cent of those with longer or more frequent cycles of disability experienced a loss of full-year work in 1994 or 1995 (which was never regained), compared with 8 per cent of those with shorter or less frequent cycles. As well, only 53.1 per cent of those with longer or more frequent cycles were employed full-year for all six years, compared with 60.7 per cent of those with shorter or less frequent cycles.
We emphasize that much work still needs to be done in perfecting both the longitudinal disability status variable and the employment variable used here. The analysis above represents an initial effort to examine these important issues.
And now for something somewhat different:
Due to numerous requests received by the CCSD for information about education and labour force participation related to persons with disabilities, we present below two tables. Please note that due to space restrictions, we have presented data for 1998 only. People interested in reviewing these data for 1993 through 1998 should visit the CCSD's disability research information subsite at www.ccsd.ca/drip.
Persons with disabilities continue to have lower levels of educational attainment than their non-disabled counterparts (see Table 2 below). For example, working-age persons with disabilities are about twice as likely as those without disabilities to have not completed high school (36 per cent compared with 18.3 per cent). As well, persons without disabilities were almost one and a half times as likely as those with disabilities to have graduated from a post-secondary institution (college or university graduate): 51.4 per cent compared with 36.4 per cent.
Accessibility Note: The webmaster acknowledges potential accessibility problems in Table 2. You can skip over Table 2 if you prefer.
| Table 2: Highest Level of Education Attained, 1998 Women and Men with and without Disabilities Aged 16 to 64, Not Enrolled as Full-time Students | ||||
|---|---|---|---|---|
| Disability Status | Highest Level of Education Attained | Male | Female | Total |
| Person with a disability | Less than high school graduate | 37.9% | 34.2% | 36.0% |
| High School Graduate | 25.6% | 29.4% | 27.5% | |
| Post-secondary graduate | 36.5% | 36.4% | 36.4% | |
| Total | 100.0% | 100.0% | 100.0% | |
| Number | 827,500 | 857,600 | 1,685,100 | |
| Person without a disability | Less than high school graduate | 19.3% | 17.2% | 18.3% |
| High school graduate | 29.8% | 30.9% | 30.3% | |
| Post-secondary graduate | 50.9% | 51.9% | 51.4% | |
| Total | 100.0% | 100.0% | 100.0% | |
| Number | 6,533,800 | 6,608,700 | 13,142,500 | |
| Source: Prepared by the Canadian Council on Social Development using Statistics Canada's Survey of Labour and Income Dynamics (masterfile). | ||||
These lower levels of education among persons with disabilities have an impact on their opportunities in the labour market, as shown in Table 3. In 1998, both men and women with disabilities were much more likely to be employed all year if they were post-secondary graduates (51.8 per cent for men and 41.1 per cent for women) than if they had not completed high school (24.5 per cent for men and 14.3 per cent for women). Clearly, increasing educational attainment among persons with disabilities improves their chances of success in the labour market.
A similar relationship between education and employment is found among persons without disabilities (see Table 3). However, while it is clear that higher education greatly benefits persons with disabilities in the labour market, persons without disabilities experience even greater benefits. For example, 51.8 per cent of men with disabilities who were post-secondary graduates were employed all year in 1998; among their counterparts without disabilities, this figure was 82 per cent. Similarly, 41.1 per cent of women with disabilities who were post-secondary graduates were employed all year, compared with 73.9 per cent of their non-disabled counterparts. So while education is important for persons with disabilities in the labour market, they also encounter other barriers in the labour market itself.
Accessibility Note: The webmaster acknowledges potential accessibility problems in Table 3. You can skip over Table 3 if you prefer.
| Table 3: Yearly Labour Force Status by Highest Level of Education Attained, 1998 Women and Men with and without Disabilities Aged 16 to 64, Not Enrolled as Full-time Students | ||||||
|---|---|---|---|---|---|---|
| Highest Level of Education Attained | ||||||
| Disability status | Gender | Labour Force Status | Less than high school graduate | High school graduate | Post-secondary graduate | Total |
| Persons with a disability | Male | Employed all year | 24.5% | 43.1% | 51.8% | 39.2% |
| Unemployed part or all year | 4.4% | 7.9% | 5.5% | 5.7% | ||
| Some labour force / some NILF | 13.1% | 11.8% | 12.4% | 12.5% | ||
| NILF all year | 58.0% | 37.2% | 30.3% | 42.5% | ||
| Total | 100% | 100% | 100% | 100% | ||
| Number | 313,600 | 212,200 | 301,700 | 827,500 | ||
| Female | Employed all year | 14.3% | 28.0% | 41.1% | 28.1% | |
| Unemployed part or all year | 5.2% | 5.5% | 5.7% | 5.5% | ||
| Some labour force / some NILF | 7.3% | 16.9% | 13.3% | 12.3% | ||
| NILF all year | 73.3% | 49.5% | 40.0% | 54.2% | ||
| Total | 100.0% | 100.0% | 100.0% | 100.0% | ||
| Number | 293,400 | 251,700 | 312,500 | 857,600 | ||
| Persons without a disability | Male | Employed all year | 64.5% | 77.8% | 82.0% | 77.4% |
| Unemployed part or all year | 5.1% | 2.5% | 1.6% | 2.5% | ||
| Some labour force / some NILF | 21.7% | 16.3% | 13.3% | 15.8% | ||
| NILF all year | 8.7% | 3.4% | 3.1% | 4.3% | ||
| Total | 100.0% | 100.0% | 100.0% | 100.0% | ||
| Number | 1,262,600 | 1,946,200 | 3,325,100 | 6,533,800 | ||
| Female | Employed all year | 40.1% | 63.2% | 73.9% | 64.8% | |
| Unemployed part or all year | 5.6% | 3.8% | 2.3% | 3.3% | ||
| Some labour force / some NILF | 15.9% | 15.6% | 13.5% | 14.5% | ||
| NILF all year | 38.4% | 17.4% | 10.3% | 17.3% | ||
| Total | 100.0% | 100.0% | 100.0% | 100.0% | ||
| Number | 1,137,100 | 2,039,800 | 3,431,800 | 6,608,700 | ||
| Note: NILF indicates that the respondent was not in the labour force | ||||||
| Source: Prepared by the Canadian Council on Social Development using Statistics Canada's Survey of Labour and Income Dynamics (masterfile). | ||||||
People interested in reviewing these data for additional years should see the Supplemental Tables documents on the Research Page.
DRIP Home |DRIP Home |Research |Contact |Links |Français |CCSD Home