Income Research Paper Series
Low Income in Canada - A Multi-line and Multi-index Perspective
The dynamics of low income in Canada
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.
In the previous chapters, we looked at the trends in low income in Canada. We now turn to a new set of questions on the dynamics of low income. Are the same people in low income year after year or are different people going in and out of low income? For those who fall into low income, how long does it take them to get out? This chapter attempts to answer these questions by looking at the proportion of people who experienced multiple spells of low income, the transition rates in and out of low income, the re-entry rate and the duration and persistence measures of low income under all three low-income lines.
To address the questions above, we followed the same people over time using the Survey of Labour and Income Dynamics (SLID). The data enabled us to examine the annual dynamics of low income in Canada as well as the longer term dynamics for several six-year periods: 1993 to 1998, 1996 to 2001, 1999 to 2004 and 2002 to 2007 and to correlate the dynamics of low income with key individual and family characteristics.1
Rates of transition to and from low income
Low-income dynamics describe the flow of people in and out of low income and the length of their stay in low income. The time span underlying the low-income transition matrix can be long or short. We first look at the pattern of year-to-year transitions under the low-income cut-offs (LICO), low-income measure (LIM) and market basket measure (MBM).
Table 5.1 may help illustrating several concepts of low income transitions. The rate at which people enter into low income between any two adjacent years, say 2008 and 2009, is the proportion of people who fell into low income in 2009 as a share of people who were not in low income in 2008. This is represented by PNL->L in Table 5.1. The exit rate measures the proportion of people who exit low income in, say, 2009, as a share of people who were in low income in 2008. The exit rate is represented by PL->NL in the table. The other two elements of the table, PL->L and PNL->NL, measure immobility of and resistance to low income, respectively. The former represents the proportion of population who stayed in low income in both years, while the later represents the proportion of the population who stay out of low income in both years.2
Table 5.1 An illustration of the low-income transition matrix, 2008 to 2009
Table 5.2 contains our estimates of the four elements of the transition matrices under different lines, from the 1993-to-1994 period to the 2007-to-2009 period. It shows that many people who fell into low income in one year were able to exit it in the next year. Under LICO, the one-year exit (or escape) rate, PL->NL, varied from 28% to 40% in the 15 two-year periods from 1993 to 2009. Under LIM, the exit rate changed from 29% to 38% in the same period. While under MBM, in the eight two-year periods from 2000 to 2009, it ranged from 35% to 42%.
However, the underlying trends in the low-income exit rates under different lines were mixed. Over time, the rates increased slightly under LICO and decreased slightly under LIM, while under MBM, no clear trend could be detected, although in more recent years, i.e., from the 2006-to-2007 period to the 2008-to-2009 period, the exit rates declined under all of the three low-income lines, most notably under LICO and MBM. Notice that the immobility measure, PL->L, changed in the opposite direction as the exit rate. Again, no clear trend can be detected, except those from the 2006-to-2007 period to the 2008-to-2009 period in which immobility rose under both LICO and MBM.
Table 5.2 The one year transition matrices of low income, 1993 to 2009
The entry rates, PNL->L, under LICO, LIM and MBM appeared to follow a declining trend under all three low-income lines from the 1993-to-1994 period to the 2006-to-2007 period. In the 1993-to-1994 period, entry rates under LICO and LIM were slightly above 5%. In the period from 2000 to 2001, entry rates under all three lines varied from 3% to 4%. By the 2006-to-2007 period, they dropped to their respective historic lows. They all started to edge up thereafter. The entry rate even reached a historically high of 4.3% under MBM during the 2008-to-2009 period. These changes can also be observed from Figure 5.1 (top panel).
The declining entry rate implies that the resistance to low-income, PNL->NL, under different lines should have increased over time since PNL->NL= 1 - PNL->L. In the 1993-to-1994 period, about 95% of Canadians who were not in low income in 1993 stayed out of low income in the next year, under both LICO and LIM. In the 2000-to-2001 period, this measure varied from 96% to 97% under the three lines. By the 2006-to-2007 period, the measures reached their historic highs, especially under LICO and MBM. However, the latest recession pulled the resistance rate down slightly. Figure 5.1 (bottom panel) plots the overall evolution of Canadians' ability to resist low income.
Multiple spells of low income
The previous section drew a picture of how Canadians enter, stay in and exit low income. However, low-income dynamics may also be examined by looking at the incidence of multiple low-income spells, both for the overall population and for various groups of people. In both cases, if the proportion of people experiencing low income continuously in a period is high, then low-income persistence is strong. Otherwise, if the proportion is low and many people experience low income for at most one or two years, low income would be characterized as transitory. Between the extreme cases, if a person experienced low income for a substantial number but not all years within a given period, the person is usually identified as experiencing chronic or recurrent low income.
The results based on data from SLID are shown in Table 5.3. An immediate observation is that low income in Canada is largely transitory. More people experience low income for at least one year in a six-year period than did people in any particular year. For example, from 2002 to 2007, from 20% to 24% of Canadians experienced low income in at least one year under LICO, LIM and MBM, while in any particular year, low income rates under the three lines varied from 9% to 13% during that period.3 At the same time, fewer people experienced low income for more than four years or for all six years in the panel. For example, under MBM, only 1.4% of Canadians had low income in every year from 2002 to 2007. The corresponding proportions were 2.1% under LICO and 3.5% under LIM, suggesting that persistent or chronic low income was unlikely to affect a large proportion of the Canadian population.
As well, low-income dynamics between high-risk groups varied a great deal. The groups most affected by persistent low income were non-elderly unattached people and lone parents. In the period from 1993 to 1998, 51% to 61% of non-elderly unattached people experienced low income for at least one year under LIM and LICO. Nearly one in four of them was in low income for all six years under LIM, and under LICO one in three was in low income for six years. The situation improved over time. By the 2002-to-2007 period, 13% to 17% of unattached non-elderly people were in low income for all six years under MBM, LICO and LIM. Even so, low-income persistency among them was still the strongest relative to other vulnerable groups.
As documented by Richards (2010), low-income incidence among lone parents declined substantially during the past decade. After examining their low-income dynamics, we found that, similar to people from unattached non-elderly group, lone parents experienced a high transitory low-income rate. The proportion of lone parents in low income for at least one year from 1993 to 1998 was 59% under LICO and 62% under LIM. By the 2002-to-2007 period, it varied from 48% under LICO to 57% under LIM to 53% under MBM. These indicate that transitory low income declined over time for this group. But the transitory low income of lone parents was the highest among all vulnerable groups in the latest period. Furthermore, lone parents also experienced a persistently high level of low income, second only to unattached non-elderly people. Similar to the latter, the persistence of low income declined among lone parents over the 1993–to-2007 period, although the decline among them was not as strong as that among unattached non-elderly people.
Table 5.3 Multiple spells of low income across groups of people
The proportion of people with low income for at least one year was also high among recent immigrants, people with activity limitations and off-reserve Aboriginal people. For the period from 2002 to 2007, about 40% of recent immigrants and about 30% of people from the other two groups experienced low income for at least one year under different lines. Despite the relatively high transitory low income among recent immigrants, their level of chronic low income was not as high as that of lone parents and unattached non-elderly people. People with activity limitations also had relatively high rates of persistent low income.4 From 2002 to 2007, 8.4%, 10.9% and 5.4% of people with activity limitations had low income over all six years under LICO, LIM and MBM, respectively. These rates were lower than those for unattached non-elderly people but higher than those for virtually all other vulnerable groups.5
Table 5.5 Multiple spells of low income across communities
Rates of transitory low-income for children were only slightly higher than the population average, while the persistence of low income among children was similar to the population average, particularly in more recent years. Among seniors, rates of transitory and persistent low income were relatively low under LICO. But under LIM, transitory and persistent measures increased over time from 1993 to 2007 for seniors, indicating that their incomes did not grow as much as that of the general Canadian population.
Table 5.4 contains results related to multiple low-income spells across provinces. The persistence of low income declined substantially in Alberta from 1993 to 2007. In the 1990s, the proportion of Albertans experiencing low income for at least four years was close to the national average. By the 2002-to-2007 period, it fell to a level far below the national one. Quebec also experienced substantial improvements in this period. During the 1993-to-1998 period, the persistence of low income in Quebec was almost the greatest among all provinces, but during the most recent period, 2002-2007, it dropped to a level close to the national average. In contrast, the persistence of low income in British Columbia increased over time. In the early 1990s, the proportion of British Columbia residents who experienced low income for four years was below the national averages under both LICO and LIM. By the 2002-to-2007 period, the proportion of people in low income for at least four years exceeded the national average under both lines.
In contrast, the persistence of low income in Ontario was mixed, with different low income lines giving different results. Under LICO, the persistence improved between the 1993-to-1998 and the 2002-2007 periods, while under LIM, the situation appeared to deteriorate between the two periods. However, when we compared the persistence measures of the province with those of the country, we found that low income persistence in Ontario was consistently below the national level.
Not surprisingly, the changes across provinces were mirrored in the changes across the largest cities in these provinces (Table 5.5). For example, the proportion of people in low income for at least four years dropped substantially in Montréal from 1993 to 2007 under LICO and LIM, and there was also a decline in Edmonton under the LICO.6 In contrast, the proportion of people in Vancouver experiencing low income for at least four years either increased (under LIM) or stayed at a relatively high level over time (under LICO), while in Toronto there was evidence to suggest that low income persistence had increased over time.
An alternative measure of chronic and persistent low income
While low-income persistence and recurrence can be measured by the number of years in low income within an observation window, the measurements have some limits. For example, they do not measure the depth of low income. Consider two people. One is in low income all years but in each year his or her income is only slightly below the threshold. The other person is in low income for only two years, but in each year, this person's income is far below the threshold. The measures used in the previous section would identify the first person as in persistent low income while the second would be treated as in transitory low income. Yet the second person might have to borrow against future income to cope with the severe low income of a particular year and, as a result, may have less income to consume in future years. Even if his or her income would not be considered 'low' in those years, he or she would be more likely to live in straitened circumstances.
To deal with the problem, we use a version of the long-term low-income incidence measure developed by Rodgers and Rodgers (1993). We follow Velleta (2005) to compare average after-tax income (or disposable income under MBM) of an economic family (or of household under LIM) over a period of time with the corresponding average low-income threshold. If the average income is below the average threshold, the family is identified as in persistent or chronic low income.7 The long-term low-income status of a family or household is then assigned to each member. The results are presented in Table 5.6.
Table 5.6 Incidence of long-term low income, using an alternative measure of persistence
Across the at-risk groups, the results under the alternative measure of persistence are consistent with findings from the previous section. The incidences of long-term low income for unattached non-elderly people and lone parents declined from 1993 to 2007, but by the most recent period, these two groups still had the highest long-term incidences among all at-risk groups. The persistence of low income among people with activity limitations was also high: although it was not as high as among unattached non-elderly people and lone parents. However, persistence of low income of people with activity limitations was still more than twice as high as the national average. Furthermore, from 1999 to 2007, the persistence of low income under LICO and LIM changed little for people in this group, while considerable improvements occurred among other vulnerable groups in this period.
The results show that, first, the improvements in the persistence of low income in Alberta and Quebec and the deterioration of it in British Columbia, which were identified in the previous section, were also observed under the alternative measure. Under LICO, the incidence of long-term low income in Alberta dropped from about 8% in the first period (1993 to 1998) to 2% in the last period (2002 to 2007), while under LIM, it dropped from about 7% to 3%. In Quebec, the incidence dropped from 14% to 6% under LICO and from 13% to 10% under LIM in the same period. In contrast, the incidence of long-term low income in the province of British Columbia either stayed relatively high or increased between the two periods. Under LIM, the incidence increased from less than 6% to almost 10%, while under LICO, it stayed at around 8%. However, the relatively stable long-term incidence under LICO does not mean that British Columbia was not worse off over time because nationally, the incidence declined and the declines in all other provinces were considerably larger than in British Columbia.
As for persistence of low income in Ontario, the alternative measure confirmed our findings under the multiple spell approach: the trends varied under different low income lines. Under LICO, some improvement occurred between the first and the last periods of observation (1993 to 1998 vs. 2002 to 2007), while results under LIM seemed to suggest a different story. Nevertheless, the results again suggested that low income persistence in Ontario was below that at the Canada level when the Rodgers and Rodgers (1993) approach was employed.
The alternative measure provided more useful information for different communities than did the multiple spell approach. It is now clearer that persistent low income dropped, sometimes substantially, in Edmonton, Calgary and Montréal from 1993 to 2007. For Edmonton, in the 1993-to-1998 period, incidence of long-term low income was above or very close to the national average of 8%-to-9%. By the 2002-to-2007 period, it dropped to the 2%-to-3% level. This low level was also seen in Calgary, although the incidence of long-term low income there was already below the national average as far back as the 1993-to-1998 period. The incidence of long-term low income also dropped significantly in Montréal. Under LICO, it dropped from 21% in the 1993-to-1998 period to 8% in the 2002-to-2007period, and under LIM, from 17% to 9%. In the 1993-to-1998 period, Montréal had the highest long-term incidences of low income across the country, at more than twice the national average. By the most recent period, its long-term incidences were much closer to the national averages.8
The situation in Vancouver was different. Under LICO, the incidence of long-term low income stayed the same over time. In the 1993-to-1998 period, it was at about 12%. By the 2002-to-2007 period, it was still at about 11%, and the incidence under LIM increased from about 5% to 11%. In the most recent period of observation, Vancouver had a higher incidence of long-term low income compared to most of the large cities under all three low-income lines. The only exception was Winnipeg, where the incidence of long-term low income under LICO and LIM were similar to Vancouver's. However, under the MBM, the incidence in Winnipeg was lower than in Vancouver, and in the period from 1993 to 1998, the incidences under both LICO and LIM were considerably higher in Winnipeg than they were in Vancouver, suggesting that the decline in the persistence of low income in Winnipeg was probably much stronger than in Vancouver.
Finally, in Toronto, although long-term low income incidence had not been much higher than that of other cities in recent years, a deteriorating trend seemed to appear. In the period from 1993 to 1998, the incidences under both LICO and LIM were the lowest in Toronto among the major cities. These incidences varied over time, but by the period of 2002-to-2007, they were higher than those of Edmonton and Calgary.
The duration of low income
The majority of those who fell into low income in one year were not able to escape it in the next year. The question is how many years, on average, would a person in low income stay there? Given that each panel of SLID observes a person for at most six years, estimates of the duration of low income is subject to the caveat of left and right censoring of a low-income spell. Left censoring occurs if, in the first year of observation, the person in question was already in low income and it is unknown whether he or she was in low income in the years before the observation. Right censoring happens when a person is in low income in the last year of observation and we do not know if he or she would continue to be in low income in years after the observation. As a result, the average length of low-income spells will generally underestimate the true duration when censoring is ignored.9
Nevertheless, the estimated duration provides a lower boundary for each group, and comparisons of spell length across groups in the same period and over the years are still valid.10Table 5.7 presents the estimated average duration of all low-income spells (completed or censored) in each panel of SLID. Overall, the average low-income spells last more than two years. Under LICO, the average duration declined slightly, from 2.7 years in 1993-to-1998 to 2.4 years in 2002-to-2007; under LIM, it stayed at about 2.5 years.
The results across groups of people and communities are more interesting. Table 5.7 suggests that lone parents and unattached non-elderly people were the two groups that had longer durations of low income. In the six-year period from 1993 to 1998, their average lengths were, respectively, 4.5 and 4.3 years under LICO and 4.3 and 3.8 years under LIM. By the latest six-year period, from 2002 to 2007, these declined to 3.6 and 2.8 years under LICO, and 3.9 and 3.6 years under LIM.
Across provinces, residents of Quebec appeared to have a longer duration of low income in the 1990s and the early 2000s, but their situation improved over time. These findings are consistent with those described in the previous sections, which showed that Quebecers had a strong persistence of low income in the 1990s and that the persistence declined thereafter. In British Columbia, the duration of low income increased slightly over time, but the level was no higher than that in several other provinces. In Ontario, the average duration increased slightly over time but it was generally below the national average. However, in the Atlantic provinces, except Prince Edward Island, low-income spells typically lasted longer than in the rest of the country, particularly when LIM was used. For example, under LIM, the average duration of low-income spells in Newfoundland and Labrador, Nova Scotia, and New Brunswick were 3.0, 3.0 and 3.2 years in the 2002 to 2007 period, the highest in the country.
Between communities, again, we found that people in Montréal and Vancouver had longer spells of low income. The difference between them was that the duration declined over time for those in Montréal, while the duration of low income for those in Vancouver increased. The duration of low income in Toronto was generally below the national average in the 1993 to 1998 period, but over time, it approached or surpassed the national average by the 2002 to 2007 period. Surprisingly, residents from the city of Winnipeg had a long duration of low income under LICO and sometimes relatively high durations even under LIM, and although the situation improved over time, progress seemed slow. The duration of low income was also relatively long in Ottawa–Gatineau before 2000, but has declined since then.
Looked at from the year-to-year perspective, there is a long-term declining trend in the low-income entry rate under all three thresholds and an increasing trend in Canadians' resistance to low-income, though the recent recession lead to an increase in the entry rate and a decrease in the resistance.
The six-year panel data allows us to examine low-income transitions over a longer time span. The data suggest that a significant proportion of low income is of a transitory nature in Canada. One-third of those who fell into low income escaped in the next year. Still, over 20% of Canadians experienced low income at some time in a six-year period, although very few of them experienced low income for all six years. Nevertheless, several groups of people experienced low income more persistently than others. The most noticeable of these were lone parents and unattached non-elderly people and, to a lesser extent, people with activity limitations. Even though their persistence in low income dropped, they were the groups most likely to be in low income, and once in low income, they would stay the longest.
The persistence of low income dropped considerably in Alberta and Quebec, as well as in the major cities in these two provinces in recent years. In Alberta, in general, and in Edmonton and Calgary, specifically, persistent low income became rare. But the situation deteriorated in British Columbia, although the duration of low income did not appear to be particularly long, while in Ontario and in Toronto in particular, although low income persistence was generally at or below the national average, there was evidence to suggest that the situation deteriorated over time. Relatively long spells of low income were found in Newfoundland and Labrador, Nova Scotia and New Brunswick as well as in the city of Winnipeg.
1. An alternative source that would allow longer periods than six years is the Longitudinal Administrative Databank (LAD), which is constructed from a 20% random sample of Canadians' income tax records. Its sample size is huge, and it covers a much longer period of time than SLID. However, the LAD has few individual characteristics, and as the census family is the basic unit of observation for LAD, none of the three low-income lines—low-income cut-off (LICO), low-income measure (LIM) or market basket measure (MBM)—that we have employed is applicable in LAD. However, it is not impossible to construct a LIM line in the database to study low-income dynamics. For an example, see Finnie and Sweetman (2003). We leave this for future research.
2. Note that PL->L + PL->NL = 1 and PNL->L + PNL->NL = 1.
3. See Table 2.1 for low-income rates in a particular year in this period.
4. The definition of people with activity limitations changed in 1999.
5. Because of small sample sizes, the proportions of low income for all six years for recent immigrants (under all three lines) and for off-reserve Aboriginal people (under LICO) were not presented. But they were not as high as for other groups.
6. The proportions for Edmonton and Calgary also dropped, but unfortunately, the samples for these cities were small and it was difficult to evaluate the changes precisely.
7. This simplified version of incidence long-term low income essentially assumes that the discount rate for future income is 0.
8. However, in the incomplete panel of Survey of Labour and Income Dynamics, from 2005 to 2009, the incidence of long-term low income increased again in Montréal and, to a lesser extent, in Toronto and Calgary, while they decreased in Vancouver.
9. Further, it may be that the longer the spell has lasted, the lower the probability it will end in the next period.
10. Comparisons for the same group over time may be problematic when the economic circumstances are different between periods, for example, if a recession occurs in one period while expansion occurs in the other. In the former, a low-income spell can last longer while in the latter, it can be shorter.