Publications

    Income Research Paper Series – Research Paper

    Low-income Dynamics and Determinants under Different Thresholds: New Findings for Canada in 2000 and Beyond

    Empirical Results

    Skip to text

    Sample statistics
    Patterns of low-income dynamics from year to year
    Patterns of low-income dynamics over time
    How are transitions to low income affected by gender, family type and other factors?
    What determinants influence transitory and persistent low income?
    What are the robust changes from Panel 3 to Panel 4?

    Text begins

    Sample statistics

    When this research was conducted, there were four complete panels in the SLID database. Each panel consists of roughly 17,000 households and about 34,000 adults surveyed for a period of six consecutive years. For studying low-income dynamics after 1999, we focus on the SLID Panel 3 for 1999 to 2004 and the SLID Panel 4 for 2002 to 2007. The targeted population consists of individuals aged 16 and older during the panel years.

    In Panel 3 (1999 to 2004), women's population share is 51.4%; men's is 48.6%.1 Immigrants account for about 18% of the population, and more than 9% of the population are members of visible minorities. At the beginning of 1999, around one-quarter of the population are estimated to be in the 35-to-44 age group, more than 10% are 65 and older, three-quarters have at least a high school diploma, about 15% are students, and 18% have activity limitations. In terms of family composition, families headed by lone parents represent 5% of the population. The proportion of unattached individuals remains stable at 16% over time. The proportion of attached individuals with children decreases from almost 42% in 1999 to about 34% in 2004.

    For Panel 4 (2002 to 2007), 51.1% of the population are women; 48.9% are men. Immigrants account for almost 20% of the population, and more than 12% of the population are members of visible minorities. At the beginning of 2002, more than one-fifth of the population fall into the 35-to-44 age group, more than 12% are 65 and older, around 80% have at least a high school diploma, about 15% are students, and 23% have activity limitations. In terms of family composition, families headed by lone parents represent 5% of the population. The unattached individuals and attached individuals without children steadily increase their shares from about 14% to almost 17%, and from about 24% to almost 28%, respectively.

    Patterns of low-income dynamics from year to year

    We analyse year to year low-income dynamics by examining the empirical low-income transition probabilities of the population for each panel over time under various low-income thresholds. Probably the simplest way to examine year-to-year low-income dynamics is to examine the proportions of the people in and out of low income as well as the empirical transition probabilities from one year to another. Table 1 shows results for the targeted population and for men and women.

    The results suggest that while the year-to-year transition probabilities vary somewhat across the three different low-income thresholds over Panels 3 and 4, they are broadly consistent across the thresholds over time. That is, approximately two-thirds of the low-income individuals stay in low income year after year, and about one-third of them move out of low income over time. About 97% of the non-low-income individuals stay out of low income, and only about 3% of them fall into low income. This observation is quite robust for the total population as well as for both female and male subpopulations.2

    Patterns of low-income dynamics over time

    The low-income rates for both men and women, under both LICO and LIM, are quite comparable. Compared to LICO and LIM, MBM captures more individuals in low income. That is, some people whose incomes are marginally above LICO or LIM are classified as low-income individuals under MBM.

    Transitory and persistent low income

    Many Canadians experienced transitory low income (low income for 1 to 3 years) during the study period. Only a very small percentage of the population experiences persistent low income (for 4 to 6 years), regardless which low-income threshold is used. In Panel 3 (1999 to 2004), 5.2% and 5.5% of the population are in persistent low income under LICO and LIM, respectively; about 6.4% are in persistent low income under MBM. In Panel 4 (2002 to 2007), 5.1% and 5.6% of the population are in persistent low income under LICO and LIM, respectively; about 5.8% are in persistent low income under MBM.

    We find that more women than men are in low income for various durations in Panels 3 and 4 (1999 to 2004 and 2002 to 2007) regardless of the low-income threshold used. We also find that compared to men, more women experience transitory low income and fewer women experience persistent low income.

    Although senior citizens are much better off than members of other age groups in Canada due to the relevant social policy, such as Old Age Security (OAS) and Guaranteed Income Supplement (GIS), we find a persistent gender difference in low income among those 65 and older—more women than men were in low income. Under LICO, the low-income incidence is about 6% for elderly men but about 16% for elderly women. Under MBM, the difference between the two groups is the smallest; with low-income incidence for elderly men at 8%; for elderly women, 11%. This gender difference may be related to the difference in Canadian public pension payments depending on how much and how long workers contribute to the plan over their working lives. Because female workers tend to contribute less over shorter periods of time than their male counterparts in their working years, female retirees tend to receive lower public pension payments.

    Life cycle transitions

    In Panels 3 and 4, we note the remarkable patterns of life cycle transitions. While most people—two-thirds to three-quarters of the population—are never in low income, young people, students, unattached individuals and lone parents are more likely to be in transitory low income, reflecting their life cycle transitions.

    First, young people aged 16 to 24 have the highest share, more than 25%, in transitory low income; seniors 65 and older have the lowest percentage, less than 10%. The transitory low-income incidence reduces remarkably as people move through the age cohorts from 25 to 54 (i.e., 25 to 34, 35 to 44 and 45 to 54). In addition, the evidence of the highest transitory low-income incidence for young people aged 16 to 24 is strong. The rate stays above 25% under all three low-income thresholds for both Panels 3 and 4.

    Second, among all age groups, those 55 to 64 have the second highest transitory low-income incidence after those aged 16 to 24. This age group also has the highest persistent low-income incidence. People in this age group can be vulnerable in the labour market as well as in health and marriage situations. But, over time, the transitory low-income incidence for this group drops to about 20% in Panel 4 from about 25% in Panel 3.

    Third, seniors have the lowest transitory low-income incidence and the lowest persistent low-income incidence. The transitory low-income incidence for this age group is lower than 10% and the persistent low-income incidence is lower than 5%. Furthermore, compared to women, men have an even lower low-income incidence in this age group.

    One of the most important life cycle changes is the change of family composition. It can be complex: any change will involve more than one individual (e.g., spouse and children). As family incomes are shared among family members, a change of family composition often has a direct impact on the equivalent individual income of the family members and therefore whether the family members are in low income according to the established low-income thresholds.

    We find that, in Panels 3 and 4, unattached individuals and those living in families headed by lone parents have higher transitory and persistent low-income incidence regardless of the low-income threshold used. Unattached individuals include those singles who are in the earlier stage of their life cycle and those who end up living alone at various stages of the life cycle. These individuals either have less income or need more to get by, or both. Hence, it is not surprising that they experience both high transitory and persistent low income (above 15%).

    Similarly, lone parents are also likely to be in low income, as single breadwinners with one or multiple dependent children tend to spread their earnings across more family members. They have a much higher transitory low-income incidence (25% or higher) and a persistent low-income incidence (15% or higher).

    High-risk groups

    The high-risk groups we have identified are individuals with less than high school education, individuals with activity limitations, members of visible minorities, and recent immigrants.3 Our findings in this paper are consistent with the literature.4
    First, we find that, in Panel 3, more than 8% of the people with less than high school education are in persistent low income under all three low-income thresholds. In Panel 4, more than 8% of those with less than high school education are in persistent low income under both LICO and LIM, and just under 8% under MBM. As the theory of human capital predicts, when workers have less education, they get lower rewards for their human capital and are more likely to fall into low income.5
    Second, in Panel 3, we find more than 16% of those with activity limitations are in persistent low income under all low-income thresholds. This is a high percentage considering that less than 4% of those without activity limitations are in persistent low income under all low-income thresholds. In Panel 4, about 14% of those with activity limitations are in persistent low income under all low-income thresholds. Again, this percentage is much higher considering that only about 3% for those without activity limitations are in persistent low income.

    Third, Panel 3 shows that more than 10% of the members of visible minorities are in persistent low income under all three low-income thresholds; yet less than 6% of the people who do not belong to these groups are in persistent low income. In Panel 4, less than 9% of the members of visible minorities are in persistent low income, while only about 5% of those who do not belong to these groups are in persistent low income under LICO and LIM. Under MBM, the percentage of people who are members of visible minorities and in persistent low income is 11%, compared with about 5% among those who are not members of visible minorities. These findings are consistent with earlier findings in the literature.6
    Fourth, we find that, in Panel 3, more than 11% of recent immigrants (those who moved to Canada after 1986) are in persistent low income under all three low-income thresholds, compared with less than 6% of Canadian-born individuals. In Panel 4, more than 9% of recent immigrants (those who moved to Canada after 1989) are in persistent low income under all three low-income thresholds, while about 5% of native-born Canadians are in persistent low income. These results echo with the recent empirical findings on the Canadian immigrant population.7

    How are transitions to low income affected by gender, family type and other factors?

    It is possible to classify low-income transition over two adjacent years, and therefore, its probability, into four categories—getting out of low income, getting into low income, staying in low income, and staying out of low income. Because of the small sample sizes for certain groups of individuals,8 we looked only at the joint probability of entering or staying in low income. The complement cases are the transitions of getting out of or staying away from low income, which are exactly opposite to the transitions of entering or staying in low income.

    Because the patterns across genders and/or family compositions are clearly different, we will study the year-to-year transition of entering or staying in low income with respect to each subpopulation group. The goal of this analysis is to further identify the factors that determine low income and their marginal effects by gender and family composition.9 In addition to the general observation on the role of gender, we find that family composition dynamics, number of children, age and student status are important life cycle factors; important risk factors are low education, activity limitation, being a member of a visible minority and recent immigration status.

    First, let us discuss the role of gender and family composition in both Panels 3 and 4 under LICO, LIM and MBM, based on our estimates from the logit models of entering or staying in low income. The result shows that, under various low-income thresholds, women, lone mothers and, in particular, unattached women, have higher probabilities of entering or staying in low income than other people. We find, other things being equal, that unattached women and lone mothers are more likely to be in low income than their male counterparts. In addition, their probabilities of entering or staying in low income under LICO are higher than those under LIM and MBM.

    Second, let us examine the role of family-composition dynamics. Family-composition dynamics refers to the following year-to-year changes in family types: from unattached to other types of family composition (such as attached with no child, attached with children, lone parent, and other); from attached with no children to other types of family composition; from attached with children to other types of family composition; and from lone parent to other types of family composition.

    In general, chances of being in low income are low when individuals move from unattached to attached with or without children under all three low-income thresholds in both panels. In particular, the impact of this change in family composition is more pronounced for unattached women than for unattached men. In other words, getting attached appears to have more impact for single women than for single men to get out of or stay away from low income.

    For families headed by lone parents, changing to unattached families would mean a greater chance of being in low income. On the other hand, lone mothers, not lone fathers, saw less chance of being in low income when they become attached. This may reflect the fact that, when getting into a relationship—either married or common-law—lone fathers may not be as well-off as lone mothers in terms of pooled family income from the income perspective. In other words, lone mothers' new life partners tend to help their families  more to get out of or stay away from low income than lone fathers' new life partners do for their families.

    Third, the role of the number of children: attached individuals with more children are more likely to be in low income. The attached individuals with no children clearly have more resources for fewer people within their families: but they may decide not to have children because of insufficient resources to share among more family members. Those with more children must spread their income among more family members, but they can also take advantage of economies of scale. While a causal relationship is difficult to conclude, the robust empirical evidence shows that the attached individuals with more children are more likely to be in low income.

    Fourth, the role of age: Our empirical evidence shows that the existence of the OAS and GIS in Canada renders senior citizens (aged 65 and older) less likely to be in low income across different types of family composition. But the unattached young people aged 16 to 24 and the vulnerable age group from 55 to 64, particularly, are more likely to be in low income. While the former often have low income to begin with and are still accumulating human capital, the latter may end up in low income for various reasons such as job separation, marriage breakdown, activity limitation or changes in family composition (e.g., death of spouse, living with grown children).

    Finally, we also observe that members of high-risk groups generally have greater probability of falling into low income and multiple risks—for example, lone parents with less education and unattached individuals with activity limitations have particularly high probabilities of transitions to low income.

    What determinants influence transitory and persistent low income?

    The above analysis looks at the factors that affect the marginal probability of transition into low income year to year. Now we will use the logit models to examine the factors that lead individuals to be in low income for (a) at least one year, (b) at least four years, and (c) all six years. By doing so, we can disentangle the determinants for transitory low income from those for persistent low income. The results are contained in Tables 2-A to 2-F).

    We find that the dominating factors for being in low income for at least one year include family composition (unattached individuals and lone parents), activity limitation, less education, student status and recent immigration. But the dominating factors for being in low income for all six years are family composition (unattached and lone parents), activity limitation, and less education. This suggests that student status and recent immigration are more likely to be the key determinants for transitory low income, but family composition (unattached and lone parents), activity limitation and less education are more likely the key determinants for both transitory and persistent low income. The above observations are quite robust under all three low-income thresholds and across the two panels.

    Perhaps it is useful to assume that people can smooth their incomes over time. The relevant question then is, would low-income incidence over the six-year period differ significantly from that on the annual basis? To find out, we aggregated individual annual incomes and annual low-income threshold over the six-year panel and compared the aggregate individual incomes with the aggregated low-income threshold as well as the upper and lower bounds (125% and 75%) of the aggregate low-income thresholds. The result suggests that regardless of which low-income thresholds are adopted, the high-risk groups with the most pronounced low-income incidence are lone parents, unattached individuals, people with activity limitations and recent immigrants, in particular recent immigrants who are also members of visible minorities.

    What are the robust changes from Panel 3 to Panel 4?

    When we compare the estimates in Panel 3 across the three thresholds with those in Panel 4, we can identify some robust changes over time.

    First, the low-income transition for women improves under all three low-income thresholds over the two panels: we found both a lower immobility measure in low income and a higher immobility measure in non-low income. The low-income incidences for women, young people, people with less education and lone parents who are in persistent low income fall from Panel 3 to Panel 4. We also see the falling low-income incidence for those aged 45 to 54. However, more young people experienced transitory low income under MBM from Panel 3 to Panel 4.

    We paid particular attention to the SST index of low-income duration, which is a comprehensive and decomposable measure of low-income duration. The results are shown in Table 3. The first component, πTt=1(δ ≥ 1/T), the proportion of the population that had any low-income durations over a six-year period, increased under LIM but declined slightly under LICO and MBM from Panel 3 to Panel 4; the second component,  the normalized average duration in low income, dropped under LIM but rose slightly under the two other lines. The third component, 1+ G(δT), an inequality measure of low-income durations, increased under LICO and MBM but declined slightly under LIM.

    Putting these components together, we see that situations of low-income duration worsened slightly from Panel 3 to Panel 4. Under LICO, the index increased from 0.2045 to 0.2065 over the two panel periods. Under MBM, the index increased from 0.2276 to 0.2300. Under LIM, it increased from 0.2062 to 0.2089. The increase under LIM is slightly stronger than those under the other two thresholds.


    Notes

    1. To save space, we only present several key tables in this article. The rest of the empirical results (tables and figures) appear in the full version of the report.
    2. The female subpopulation has a slightly higher probability of staying in low income and a slightly lower probability of staying out of low income relative to their male counterpart under various low-income thresholds and over Panels 3 and 4.
    3. The concept of recent immigrants is dynamic in our study. That is, in the setting of Panel 3, recent immigrants refer to those who landed in Canada after 1986; in the setting of Panel 4, recent immigrants refer to those who landed in Canada after 1989. This addresses the fact that Panel 4 started three years later than Panel 3.
    4. Relevant research includes Morissette and Zhang (2001), HRSDC research paper (2009) and Valletta (2005).
    5. The theory of human capital can be traced back to Becker (1964) and Schultz (1971).
    6. See, for example, Statistics Canada (2001) and Samuel (2006).
    7. See, for example, Picot and Hou (2003, 2007).
    8. When some specific transitions (getting into low income) are analysed in a regression framework with a large number of covariates, the number of cases for these kinds of transitions with reference to some covariates is too small to be published under Statistics Canada's release guidelines.
    9. The marginal effect (also known as partial effect or net effect) in logit regression models here refers to the marginal contribution of each covariate to the probability of entering/staying in low income, with everything else held constant. When the covariates in the models are all binary variables (0/1), we can interpret the marginal effect of each covariate of interest in the following way: the marginal contribution of an individual covariate to the probability of entering/staying in low income is triggered by 'switching on' this covariate (from 0 to 1), while keeping all other covariates 'switched off' at 0. This is done by performing the logit command and its auxiliary command mfx using Stata (Ver. 11.0).
    Date modified: