Income Research Paper Series
Low Income in Canada - A Multi-line and Multi-index Perspective
Methodology and data
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.
Low income and poverty: Some conceptual issues
This report employs three low-income lines, the after-tax low-income cut-off (LICOs) and the after-tax low-income measure (LIM) developed by Statistics Canada as well as the market basket measure (MBM) developed by Human Resources and Skills Development Canada (HRSDC). The report also uses three indexes (the incidence, gap ratio and severity) to examine low income in Canada. It is not a report on poverty. As a statement by former chief statistician Ivan Fellegi makes it clear, "poverty is intrinsically a question of social consensus" and the determination of poverty criterion ultimately "involves value judgments" and, as such, it is not the role of Statistics Canada to define and measure poverty.1 Instead, Statistics Canada seeks to assist policy development and public discourse on poverty by providing low-income thresholds and low-income statistics, based on scientific principles and international best practices.
Scientific principles and international best practices indicate that low income and poverty are not the same. In the scientific community, poverty has been conceived and measured differently2. But low income and poverty are seldom treated as identical concepts. Peter Townsend (Rio Group, 2006) approaches poverty from a social exclusion perspective. He wrote that the "determination of a poverty line cannot be based on an arbitrary selection of a low level of income" and it is not "enough to describe poverty as a condition applying to those whose disposable income is low relative to that of others." The 1998 Nobel laureate of economics, Professor Amartya Sen views poverty from the capability approach. He says that poverty "must be seen as the deprivation of basic capabilities rather than merely as lowness of income." 3
One of the fundamental axioms on poverty is the focus axiom. It states that a poverty index should be independent of the non-poor population. The axiom implies that poverty measures should be about the 'poor' only. Many low-income conceptions and measurements, Statistics Canada's low-income lines included, violate the focus axiom. As shall be seen later, the thresholds of LICO are determined by the average spending on food, clothing and shelter of all Canadian households, not a subset of these households; the thresholds of LIM are based on the median of the income distribution of the whole population, while the baskets of the MBM represents a standard of consumption that is close to median standards of expenditure for food, clothing and footwear and shelter and somewhat below the median standard for other categories of expenditure. Hence, low income statistics under LICO, LIM and MBM are not independent of the incomes or expenditures of the non low income population. Therefore, if one accepts the focus axiom as a scientific principle in poverty measurement, one would have to distinguish low income from poverty.
But nobody would deny that poverty and low income are closely related, although answers to the question of how they are related differ. Townsend and Sen appear to suggest that low income is one aspect of poverty, so one cannot rely on low income alone to fully understand poverty. This is also consistent with authors who view poverty as a multi-faceted, multi-dimensional phenomenon in which income is a key dimension and the relative lack of income is a key facet of poverty.4 In contrast, the 'welfarist' approach tends to measure poverty based on level of 'utility' in the context of optimal consumer choices. Here income enters into the decision-making process through a budget constraint. A person is in poverty if he or she fails to attain some normative level of utility. The key element under the welfarist approach to defining poverty is the lack of command over commodities (Duclos and Araar, 2006), and low income is one of the determinants of poverty. If a person's income is low, he or she would be unable to acquire a sufficient amount of commodities to attain the non-poor level of utility.
Since we do not have a definition of poverty in Canada, we would not be able to explicitly and precisely link our low-income measures to poverty. But some links are conceivable. For example, LICO identifies a family as in low income if it spends a substantially higher proportion of its income on life's essentials, and is thus likely to live in "straitened circumstance". If one treats this straitened circumstance as a state of relative deprivation, then a person who falls in low income would be somebody who is at risk to live in relative deprivation. LIM identifies a household (and its members) as in low income if its income is below half of the median adjusted income. This is an internationally accepted standard across developed countries. A person who falls into low income under LIM would be someone who is likely to have difficulties to fully participate in the society. Thus the LIM conception seems to have a link with the social exclusion definition of poverty.5
In practice, both Statistics Canada and HRSDC explicitly state that their low-income thresholds are not designed to measure poverty or to make any explicit judgments about what in Canada is a minimally acceptable level of income. Rather, low-income statistics should be used as a tool to broadly characterize groups of Canadians who are facing a high risk of poverty. More importantly, the statistics have been employed to track the changes in these groups over time, and since their inceptions, low-income thresholds have been used to identify low-income problems associated with, for example, the elderly, single parents, and children. In all these cases, government programs have been developed to help low-income Canadians and have had impacts on them.6
This report is based on the understanding that poverty and low income are closely related and yet are not the same. This view is consistent with experts' consensus and international best practices. At the minimum, the low-income measures are useful tools to help us to understand poverty in Canada in lieu of knowledge on what constitutes poverty and how it should be measured.
Low-income lines or thresholds in Canada
As in poverty studies, there are two issues to resolve for low-income analyses: the identification issue and the aggregation issue. Identification tells us who is in low income and who is not, while aggregation tells us how many people are in low income, how far away they are from a given threshold and so on. Let's start with the low-income lines.
This report makes use of three low-income thresholds: LICO and LIM of Statistics Canada and MBM of HRSDC.7LICO is the oldest and the most widely recognized line. It is uniquely Canadian, not in use elsewhere in the world. LICO has been strongly criticized because of several methodological shortcomings (Wolfson and Evans, 1989) but it remains popular in practice. The current LICO thresholds are based on the 1992 consumption pattern of Canadian households. However, they are updated annually using the Consumer Price Index (CPI) to keep their real values.
More recently, HRSDC developed the MBM and began releasing results in 2000. The MBM is updated annually for price changes in the basket, and the basket is periodically rebased. The current MBM thresholds are based on a basket designed to meet the modest needs of Canadians, given the 2008 standard of living.8 The main strengths of the MBM are that it provides a basket approach to measuring low income and it is sensitive to regional variations in the costs of living.9 The measure is easy to communicate in the sense of comparing incomes with the cost of a basket of goods and services. But the details of goods and services in the basket are complicated. Numerous assumptions and experts' judgments need to be applied in deciding the items and their amounts. It is also very costly to rebase the basket and to calculate the thresholds.
The LIM line was introduced in the early 1990s based on the 1989 Wolfson and Evans report.10 Its threshold is calculated as half of the median of the contemporary income distribution. In this sense, LIM is rebased every year to reflect annual changes in the standard of living. The methodology has gained currency during the last few decades among academic researchers and government organizations in low-income measurement and international comparisons. It is straightforward to calculate, easy to communicate, and all of its underlying assumptions are highly transparent.
However, an oft heard criticism to the LIM is that it is a pure inequality measure. Our view is that, as other low-income measures, it is closely related to inequality, but it does not measure inequality. First, by definition, an income inequality measure gauges the disparities of income of a society among all its members, while the low-income rate under LIM tells the proportion of people whose incomes are below half of the median adjusted income. The gap ratio tells us how far from the threshold a low-income person's income is, while the severity index tells how unequally income is distributed among the low-income people (but not among all people in the population, which is always the base for inequality measures).
Second, low income under LIM behaves differently from inequality measures. In measuring inequality, the most important principle an index must satisfy is the Pigou-Dalton transfer principle. In the strong form of the principle, it states that when an amount of income is transferred from a rich person to a poor person while maintaining their income ranks, the inequality measure should decrease. In the weak form of the principle, the inequality index should not increase. The incidence of low income under LIM can easily violate the strong form. For example, when the amount transferred is not large enough to lift the recipient out of low income. On the other hand, if the transfer happens to reach a person in the middle of the distribution such that the income rank is not altered but the median is made higher than before, low income rate would increase under LIM because the threshold becomes higher.
Third, operationally, inequality measurement is a one-step exercise in which incomes from the top of the distribution play the same role as incomes from the bottom. But low income measurement is a two-step practice. Under LIM, top incomes only affect the thresholds, while bottom incomes affect both the low-income thresholds and low-income indexes. Furthermore, since the median is not sensitive to the extremes at the high and low ends, low income indexes under LIM are not as sensitive to top incomes as inequality indexes.11 Consider a simple case of two societies. The income of the richest person in one society is ten times as high as that of the richest person from the other, but all other aspects of the two distributions are the same. It is clear that inequality measures of the two societies will be different, but their low income measures under the current LIM configuration would be identical.
A related criticism is that low income will always be with us under LIM. Wolfson and Evans (1989) answered this question with a simple example. Figure A1 may help further to counter the criticism. Suppose the top graph of the figure describes the income distribution before certain transfers are made, while the bottom graph represents a situation after the transfers. The transfers may be designed such that the median is preserved, and hence the LIM thresholds (indicated by the vertical lines with an arrow) remain unchanged. It can be seen that low-income incidence, gap ratio and severity become zero in the bottom graph. Therefore, low income need not to be always with us under LIM.12
There are several similarities and differences between the three lines that will affect their interpretation. A major similarity is that all three lines set a standard that is based on the concepts of social exclusion and the ability to participate in Canadian society. Other thresholds set a lower standard and are probably more appropriate for concepts of survival or serious hardship.13 All of the three lines imply that individuals living in households and families below the threshold are in some sense deprived relative to others, but each line tells a different story about what it means to be deprived.
Figure A1. Income distributions before and after a transfer
The LICO implies that a person is deprived if he or she lives in a family that has to spend significantly more of its income on the necessities of life than the average family, and therefore has little discretionary income with which to participate fully in society. The LIM considers one to be deprived if his or her household income is less than half of the income of the median household. With much less income than the mainstream makes, a person would have difficulty participating fully in the society. The MBM considers as deprived individuals whose families lack the disposable income to purchase the goods and services of the 'market basket' that represents a modest yet decent standard of living (HRSDC, 2009).
The second similarity between the lines is that, regardless of how the thresholds were determined, they are all compared with resources defined in terms of income. Poverty and deprivation can be studied by examining what people actually consume or by examining their income—which determines their capacity to consume. Income information is more broadly available than consumption data in Canada, and so for practical reasons the development of indicators of deprivation has tended to use income.
While all three lines use income available for discretionary consumption there are some differences. The LICO and LIM make use of Statistics Canada's standard income concepts such as market income, total income and after-tax income. MBM starts with the same concept of after-tax income but goes further to deduct some other non-discretionary spending to end up with its own disposable income concept.14
All three lines take economies of scale into account through the use of equivalence scales, but they account for these economies differently. Economies of scale in consumption and resource-sharing within the family or household are important factors in determining the low-income thresholds. Larger families do not need the same level of per capita income to be as well off as smaller families because of economies of scale. However, the units of resource-sharing of the three lines are different, and each line assumes different equivalence scales. The unit of sharing of LICO and MBM is the economic family, while the unit of sharing under LIM, consistent with international practice, is the household. Practically, there is little difference in the definition, as 95% of households contain only one economic family and only 2.5% of individuals live in secondary economic families (Murphy et al., 2010).
Both the LIM and the MBM adjust income using the square root of the number of members living together; the square root of the economic family size in the case of the MBM and the square root of household size in the case of the LIM, while under LICO, the equivalence scales were estimated and implicitly built into the thresholds. However, the assumption of sharing of resources within the economic family or household under the three lines is identical: each member shares equally, implying that if the family or household falls in low income, each member is also in low income.
The three lines can be also compared in terms of how each threshold is updated. There are two fundamentally different ways—rebasing and indexing.15 Rebasing refers to the process of making judgments as to the relative level of income required to participate fully in society at a given point in time, while indexing refers to a simple adjustment of the dollar amount of the thresholds to account for inflation. Every low-income line, as such, represents a standard based on relative judgments that have been set at a given point in time. When this standard is not rebased to reflect current living conditions (i.e., making new relative judgments), and the thresholds are merely indexed to the CPI, it allows a comparison of Canadians' current situation to the distribution of well-being in an earlier time. Of course, at some point the relevance of this comparison becomes questionable as time passes. For example, it makes little sense to determine the low-income status using thresholds based on standards of living reached centuries ago, even though the thresholds are indexed to account for inflation.
The current LICO thresholds are based on the 1992 consumption patterns in Canada. Each year the LICOs are adjusted to account for increases in the cost of living using the CPI, but since 1992 Statistics Canada has not rebased the cut-offs themselves.16 As such, the real value of the LICO thresholds represents the well-being of Canadians living in the early 1990s, so they are considered 'fixed' thresholds.17 The first MBM basket was established in 2000. The basket was recently revised relative to the 2008 standard of living. HRSDC plans to rebase the MBM basket periodically by examining the contents of the basket and making appropriate changes. Each year the basket is indexed through a process of re-costing the basket at current prices, but the basket itself does not change. But different from both LICO and MBM, as mentioned before, the LIM thresholds are rebased each year to reflect changes in the income distribution and as such always based on a current standard.
There are similarities and differences between the three lines in terms of geographic reference. Every low-income line represents a living standard that has been set for a given geographic place. While all three lines can be considered national lines, the low-income lines vary in their treatment of geographic differences across Canada. The LIM is truly a national line, as no attempt is made to adjust the levels of its thresholds for different regions. Each Canadian is compared with the same national threshold for a given household size. It is the most analytically transparent line, as it is clear that the same standard is applied equally to all Canadians. While this transparency is desirable, the national LIM does not take into account price differentials across Canada.
The LICO on the other hand is a mix of national and sub-national calculations. The overall average proportion of spending on necessities is calculated at the national level while the actual thresholds are calculated for five different areas to reflect different consumption patterns between these areas. Thus the LICO for large cities would apply to any city over 500,000, regardless of geographic location. That is, the same line is used for Montreal, Toronto and Vancouver. The MBM varies geographically, as well, but the determination is more explicitly designed to take into account the price differentials across Canada. The MBM thresholds vary with the cost of the goods and services in the basket, not only between community sizes, but between communities of the same size in different provinces. Currently, the national basket of MBM is priced for 49 different geographic areas. While taking price variability into account, there can still be issues involved. For example all rural areas in a province have the same MBM threshold while some of them are decidedly closer to city centers than others.
Why multiple lines and indexes?
Requests persist for a single index under a single low income line so that simple, straightforward answers to questions such as, who are poor, how many are poor and how poor they are, can be obtained. On the surface, low income or poverty appears to be a simple binary event: a person is either poor or not poor; a person is either in low income or not. But the problem is that there is no universally accepted standard for determining poverty or low income and low-income lines or thresholds will always contain a certain degree of arbitrariness.18 Given these, one is practically "forced to use more than one criteria" to obtain robust results.19 Consequently, following international best practices, we employ multiple low-income lines (LICO, LIM and MBM) and indexes (the incidence, the gap ratio and the severity) in this study.
The arbitrariness is embedded in all three low-income lines in Canada. The LIM methodology chooses 50% of the median-adjusted income to determine the low-income thresholds. LICO determines its thresholds as the income of households that spent 20% of their income above the national average on food, clothing and shelter. There is no 'correct' answer as to why 50% was used; why not 55% or 45% in the case of LIM? Why 20% and not 19% or 21% in LICO? The arbitrary nature of the low-income lines implies that they are essentially tools to answer hypothetical questions, such as what would be the low-income rate if half of the median income were chosen as the criterion? Without a 'correct' definition of low income, several low-income lines would be preferred over a single line to better understand poverty through low-income statistics. At the very least, more robust results are possible when different low income lines are employed.
With any low-income line, several low-income indexes can be obtained. The most popular one is the low-income rate, also known as low-income incidence or headcount. It tells us what proportion of the population lives under a given threshold. This index is simple and easy to understand. But it is not the only useful index, nor does it provide complete information about low income. On the contrary, it has been criticized because of its 'all or nothing' character.20 In the extreme case, it means somebody whose income was one dollar below the line was identified as being in low income and somebody else whose income was one dollar higher was not. Such a scenario makes the headcount index undesirable. Best practice calls for the low-income incidence to be accompanied by other measures.
On the other hand, the low-income gap indexes tell us how far the income of a low-income person is from the relevant threshold. One of them is the Sen index. It measures the difference between incomes of the low income population and the thresholds they face. A more popular index calculates the average gap ratio among the whole population. Low-income severity indices are also available. One of them is the average of the squares of the gap ratio. This severity index indicates how 'low income' is distributed among the low-income population. An example may illustrate why higher-order indexes are useful. Consider a society that consists of ten people. The low-income line is $15,000. Three people have income below the line: $14,500, $10,000, and $5,000, respectively. The incidence is then 30%. Suppose the society transfers $1,000 to the three people. If the total is transferred equally among them, their well-being would all be improved. While the headcount would stay the same as before, the gap ratios would be able to capture the improvement.
To see the usefulness of the severity index, consider another society that has the same population size and low-income lines as above but the three people have incomes of $14,000, $10,000 and $5,500, respectively. If we rely on the low-income rate and gap ratio indices only, we may conclude that the two societies have equal low income. But because the income of the person from the very bottom of the second society is better-off than his/her counterpart in the first society, one can say that low income in the second society is less severe.21
Although multiple lines and multiple indexes help us obtain robust results, they are certainly accompanied by challenges. For example, multiple lines result in different incidences of low income. The tempting question is then which estimate is 'correct'? The simple answer is that they are all correct in their own contexts, because different lines answer different hypothetical questions. They measure low income from different angles and their interpretations are different. Another challenging question is, because of the arbitrariness inherent in the low-income lines, what is the usefulness of low-income statistics? The answer is that they can be used to study trends in low income as we did in the current report. A low-income line may lead one to overstate or understate the level of low income, but those biases do not invalidate low income comparisons over time if the line is rebased regularly and the underlying standard of living does not change dramatically. That is, if a given line overstates or understates low income in one year, it tends to do the same in the next year, producing a valid trend in low income over time, even though the estimated levels in low income are biased.
Figure A2. Low income rates under previous (2000) and current (2008) MBM thresholds
The above argument can be illustrated by the estimated low-income rates under the 2000 and the 2008 MBM lines for the period from 2000 to 2009 (2007 under MBM 2000). For each year, the top curve in Figure A2 shows what the low-income rates would have been if the old MBM basket (MBM 2000) were to be used, while the bottom curve indicates the estimated low-income rates under the current 2008 MBM basket (MBM 2008). Apparently, the estimated levels of the low-income rates (and hence the number of people counted as in low income) under the two MBM lines were quite different, we do not know which level is correct and we do not know the magnitudes of the biases contained in the two groups of estimates. But, no matter which MBM thresholds are employed, one would observe the same downward trend in low-income rate from 2000 to 2007.
The arbitrariness of the low-income thresholds has consequences for the interpretation of low-income statistics. Many observers are interested in the extent of low income in Canada. There is no definitive answer—only hypothetical ones—to the question of how many poor people there are in Canada. With three different lines, this report provides three estimates of the number of people with low incomes. In 2009, the highest estimate of the number of Canadians with low incomes was 4.4 million, according to LIM, while LICO and MBM counted 3.2 and 3.5 million people, respectively. In the year 2000, LICO and LIM produced the highest estimates of low income, at 3.8 million people, while the MBM counted 3.6 million. Still earlier, in 1996 it was LICO that had the highest estimate, at 4.4 million, a full 725,000 higher than the LIM.
These estimates of the number of people in low income are difficult to interpret and the relative ranking of the measures means little on its own. The estimates include certain individuals who one would likely not consider poor even though they have low income. Conversely, people may have income above the threshold but, for example, because of high spending on health care due to sickness, may have insufficient income to meet their basic needs. The issue is not that the LIM is finding more low-income Canadians than the LICO in 2009, nor is the issue that the LICO is finding more low-income Canadians than the LIM in 1996. Each line has its own metrics and should be looked at over time relative to itself. Where these trends over time differ between the lines, we can examine multiple lines and indexes to gain a clearer picture of the underlying phenomena.
Statistics Canada has several data sources that can be used for analysis of low income. Given time and resource constraints, the current report is based only on data from the Survey of Consumer Finances (SCF) and Survey of Labour and Income Dynamics (SLID).22
The SCF was an annual supplement to the April Labour Force Survey (LFS) until 1997. The samples for SLID are also selected from the LFS frame. The targets of the population of the two surveys are the same: all individuals in Canada, with the exclusions of residents of the Yukon, the Northwest Territories and Nunavut, residents of institutions and persons living on Indian reserves.23 The SCF is conducted to provide data on cross-sectional income for the Canadian population. The SLID serves the same purpose, in addition to providing longitudinal income data.24 We have employed SCF for the period from 1976 to 1995 and SLID for the period from 1996 to 200925. The only exception was that we used SLID from 1993 for studying the dynamics of low income. While we examined the evolution of low income for the 34-year period from 1976 to 2009, the focus was on more recent years, particularly from 2000 to 2009, in which some key variables for identifying vulnerable groups of individuals were available.
For cross-sectional, Canada-level analysis, we took all individuals from the surveys into account; no individual was excluded. The sample sizes varied from year to year. But, with the exception of 1976, 1978, 1980 and 1983, the sample sizes were large, ranging from 64,167 persons in 2008 to 115,966 in 1990. Even during the years in which the samples were relatively small, they were still reasonably large at the country level, varying from 38,315 in 1976 to 42,322 in 1980.
For analysis of low-income dynamics, we relied on balanced panels with various lengths, from two to six years. At the national level, the numbers of observations were 29,152 for the 1993-to-1998 panel, 29,527 for the 1996-to-2001 panel, 25,602 for the 1999-to-2004 panel, and 23,735 for the 2002-to-2007 panel. For provinces and cities, the overall sample sizes were also reasonable. For example, in 2008, although the total sample size was relatively small, the number of observations in the provinces ranged from 1,785 in Prince Edward Island to 18,509 in Ontario. As well, for the seven largest cities, the number of observations varied from 1,412 in Calgary to 3,415 in Toronto. However, when individuals within provinces or cities were classified by groups, the sample sizes were not large, particularly in smaller provinces and cities. As a result, the report excluded groups whenever they had less than 100 observations.
Finally, it should be noted that data for the low-income thresholds were somewhat different. The LIM thresholds were constructed using income and demographic data from SCF and SLID. Hence, low-income statistics under LIM as well as the LIM thresholds are estimated from the same data source. The LICO thresholds were constructed with data from the 1992 Family Expenditure Survey (FAMEX), with the thresholds for a particular year being adjusted by the CPI.26 The MBM thresholds, on the other hand, were constructed using price data collected by Statistics Canada separately, while the corresponding disposable income, against which the MBM thresholds were compared for low income identification, was based on data from SLID.27
1. Fellegi (1997).
2. There are different conceptions of poverty, there are the "welfarist" and capability approaches, and there are the subsistence, basic needs, and social exclusion conceptions of poverty. For a reference, see Duclos and Araar (2006).
3. Sen (1979).
4. For example, Asselin (2009) lists ten dimensions with income being the first in the list.
5. For relationship between the MBM thresholds and poverty, see Michaud, Cotton and Bishop (2004).
6. It is, however, important to recognize that government transfers, while significant and important to the families that receive them, have in general a much smaller impact on the distribution of incomes than do labour markets and financial markets. For this reason, caution is warranted in interpreting low-income statistics as an evaluation of government programs: much more careful analysis is required for such program evaluations.
7. For details and technical descriptions of these lines, see Statistics Canada (2009), Murphy et al. (2010) and HRSDC (2009).
8. The shelter component of the 2008 MBM basket is currently under review by Human Resources and Skills Development Canada
9. Strong price variations are able to affect the MBM incidence estimates. For example, if median house prices were to rise sharply, this would tend to increase the measured incidence of low income regardless of the actual housing costs paid by low-income individuals.
10. The LIM methodology was updated in 2010 to better reflect international developments (Murphy et al, 2010).
11. However, if one uses the mean income as basis for the threshold, as some countries do, then income from the top plays an equally important role in affecting the thresholds as those from the bottom do.
12. The example also helps to show that low income and inequality measure different things. After the transfers, low income would be erased, but inequality still exists.
13. Chris Sarlo, for example, has defined a basic need poverty line as the cost of a list of basic needs required for long-term physical well-being (Sarlo, 2001).
14. Market basket measure family disposable income is the sum remaining after deducting from total family income the following: total income taxes paid; the personal portion of payroll taxes; other mandatory payroll deductions such as contributions to employer-sponsored pension plans, supplementary health plans and union dues; child support and alimony payments made to another family; out-of-pocket spending on child care; and non-insured but medically-prescribed health-related expenses such as dental and vision care, prescription drugs and aids for people with disabilities. Generally, the average disposable income for the market basket measure is about 10% lower than average after- tax income used by LICO and LIM.
15. These are also sometimes referred to as relative updating and absolute updating. We chose not to use that terminology as it is sometimes confused with the absolute or relative nature of the actual basket or threshold.
16. The low-income cut-offs have been previously rebased in 1969, 1978, 1986 and 1992. Statistics Canada is not planning an update to the low-income cut-offs chiefly because the changes to the Survey of Household Spending make it impossible to replicate the methodology.
17. This, in a way, is similar to Orshansky's line in the United States, which has not been rebased for more than 50 years.
18. Atkinson (1983).
19. Sen (1979).
20. Saunders (2004).
21. For a complete list of low income indexes, see, for example, Duclos and Araar (2006).
22.We plan to explore data on low income from the Longitudinal Administrative Databank to assess the dynamics of low income among groups of individuals in more detailed regions over a longer time span and data from the censuses to examine low income in small areas.
23. These exclusions amount to less than 3% of the total population.
24. For detailed information on the two surveys, see Cotton (2000) and references therein.
25. Since 2007 standard publications on low income from SLID have used a combined and re-weighted SCF/SLID sample from 1993-1997. For this reason estimates provided in this report will not exactly match published estimates for those years.
26. Zhang (2010), among others, illustrates how the LICO thresholds were estimated for 1992 and how they were adjusted in other years.
27. See Tables A1 through A3 for the thresholds under the three lines in 2009.