Poverty in Canada (2nd Edition)

STATISTICS CANADA PROVIDES CANADIANS WITH the closest thing to an official poverty line. Virtually all poverty researchers, journalists and anti-poverty organizations use the Statistics Canada Low Income Cut-Offs (LICO) as if it were "the" poverty line. Indeed the National Council of Welfare, the federal government's advisory body on poverty and social policy and purveyor of widely distributed annual reports on poverty in Canada, uses the terms "poverty line" and "low income line" interchangeably. However, in spite of the almost universal acceptance of LICO as the poverty line, Statistics Canada is less than enthusiastic about its use as such. "Although Statistics Canada's low income cut-offs are commonly referred to as official poverty lines, they have no officially recognized status nor does Statistics Canada promote their use as poverty lines." For good reason!

The use of LICO as poverty lines, official or unofficial, is fraught with numerous hazards. However, before discussing the inadequacies of LICO, it is essential first to outline in detail its methodology.

LICO methodology

The construction of LICO is a multi-stage process. It begins with information on how much families in Canada actually spend on the three necessities—food, shelter and clothing. This information is obtained from the Family Expenditure Survey conducted every four years by Statistics Canada and published as Family Expenditure in Canada (FAMEX). Recent LICO lines continue to use the 1978 FAMEX data which reveals that the average family spent 38.5 percent of its gross (before tax) income on the three necessities. Statistics Canada then makes the judgement that any family that had to spend at least 20 percent more than this value (i.e., 58.5 percent) on the necessities would be in "straightened circumstances" or be "low income."

Finally, in order to translate the 58.5 percent line into a numerical LICO, the relationship between spending on the necessities (food, shelter and clothing) and gross income is required. Using regression analysis on raw FAMEX data, a sort of "necessities" consumption function is estimated. The regression, in log-linear form, includes family size and degree of urbanization "dummies." These are expected to be and were found to be statistically significant influences on necessary expenditures. Not surprising to anyone with even passing familiarity with consumption functions, the results indicate that spending on necessities varies directly but not proportionally with income. Simply put, for a given family size and degree of urbanization, family spending on necessities rises as income rises but not as fast. This means that the ratio of expenditure on necessities to income declines as income rises. This evidence confirms intuitive understanding of expenditure behavior. Now, once family size and degree of urbanization have been specified, we have a line relating expenditures on necessities and income and we simply have to find the place on the line where expenditures on necessities are .585 of income. The corresponding income level is the LICO for the base year (currently 1978). LICO are updated annually using the previous year's CPI.

Illustration

A numerical illustration is useful in providing a better understanding of the LICO methodology. There are two equations used in the determination of LICO. The first equation can be referred to as the low income line because it represents Statistics Canada's judgement as to the proportion of spending on necessities that constitutes "low income." As we have seen, the currently used proportionality is .585. The second equation is simply the "necessities consumption function" which is estimated using regression analysis. It gives us the relationship between expenditures on necessities (food, shelter and clothing) and income, family size, and size of community. To simplify the illustration, let us focus our attention on a family of four in an "average" sized community in Canada (population between 30,000 and 99,000). Given this information, our second equation reduces to a relationship between expenditure on necessities and gross family income. The slope coefficient of this equation, the marginal propensity to consume necessities, is incorporated in the income elasticity of necessities, denoted e. Close examination of both the 1978 and 1986 FAMEX surveys reveal that e lies somewhere between .6 and .8. While it is impossible to be precise using "average-of-range" data, it does appear that e has declined somewhat between 1978 and 1986. For convenience, this illustration will assume that e= .65. This value is within the estimated range and results in a slope coefficient of .25. Therefore, the LICO equations for 1978 would be:

(1)N = (.385 + .2)X

(2)N = 3599 + .25X

Where N = expenditure on Necessities (food, shelter, and clothing)

and X = Gross Family Income

The solution to this simple system of equations (X = $10,750) gives us LICO for a family of four in an average sized community in Canada in 1978. Using data on the distribution of income, it is relatively easy to determine the number of families (in this particular category) that fall below LICO and thereby derive the "incidence of low income." LICOs are determined in this manner for seven different family sizes and five different population categories each year.

LICO lines are inflation adjusted each year according to the rate of growth of the C.P.I. Formally, this involves increasing the intercept term by the inflation rate. In this manner, future poverty lines can be estimated using "expected" inflation rates. As long as there is no real economic growth, "updated" LICO lines keep pace with rising incomes. However, to the extent that living standards improve, i.e., real incomes increase, inflation-adjusted LICO falls behind rising incomes. In this case LICO loses some of its "relativity" unless it is rebased. Rebasing LICO involves a change in equation (1) to reflect the fact that, as living standards increase, the average family spends a smaller portion of its income on necessities. It may also involve changes in equation (2) because the "necessities consumption function" must be re-estimated. The end result of rebasing is higher LICO lines reflecting higher average living standards.

Figure 3-1 can be used to demonstrate the essential features of the LICO methodology. Line 0A corresponds to equation (1) and represents the N/X ratio that, in the judgment of Statistics Canada, qualifies a household as "low income." In 1978 that line had a slope of .585. The line BC (equation 2) is the "necessities consumption function" or the estimated relationship between expenditures on necessities and income in the base year. The intersection of OA and BC determines the low income cut-off. This occurs at X* and is referred to by most users as the "poverty line."

Click here to view Figure 3-1: How LICO lines are determined

Annual inflation adjustments can be shown as increases in the consumption function intercept, point B. Line BC shifts up parallel cutting OA further to the right and results in an increase in X*. If, over a period of time, the economy experienced 20 percent inflation but no real income growth, LICO would increase by that same 20 percent. If, however, we experienced a 20 percent improvement in average living standards and no inflation, then, in the absence of rebasing, LICO would not change. It would therefore fall behind in relative terms. What has happened in this case is that, because of real income growth, the typical family is not spending as great a proportion of their income on necessities as they used to. This means that "true" OA is now flatter. Only if LICO is rebased to reflect the new ratio will X* rise and recover ground lost to real growth.

This, of course, presents a problem for relativists. As Gunderson (1981) states "In essence, at least some of the fall in the incidence of poverty associated with the revised [Statistics Canada] poverty line occurs because the revised poverty line grows slower than does average income; hence proportionately fewer people fall below that poverty line."

Another (and far more obvious) problem with LICO is the arbitrary 20 percent "rule." Why is it that families spending 20 percent more than the average proportion on necessities are regarded as low income? Why not use 15 percent, or 25 percent or any other value? Wolfson and Evans (1990) explain that the originator of LICO, Podoluk (1967), concluded that a family which spent more than 70 percent of its income on food, clothing and shelter "were in 'straightened circumstances' relative to the rest of the population." This percentage was 20 percent higher than the Canadian average. While Statistics Canada has decided to maintain the 20 percent parameter for consistency, they do acknowledge that it is entirely arbitrary.

Had Statistics Canada chosen to define a "poverty line," the arbitrary 20 percent add-on would have never been acceptable to serious researchers. However, the concept they selected was "low income cut-off" and an arbitrary judgment (the 20 percent rule) is less likely to be challenged. This term is conveniently ambiguous and does not contain the heavy emotional overtones of the former concept. The problem is that the low income cut-offs are widely interpreted as "poverty lines." In spite of their caveat, Statistics Canada may well have intended this. They have certainly acquiesced in this use of LICO over the past decade. Indeed the only practical use of LICO is as a poverty line. What else could it be used for? The credibility of the LICO values will obviously depend on people's perception of poverty. To the extent that poverty means relatively less well off than average, LICO will have its adherents. On the other hand, if being poor implies the inability to afford the basic necessities of life then LICO will not be a believable poverty line.

It is impossible to rule out a political motivation in the selection of 20 percent. By using that value in the construction of LICO, Statistics Canada calculates that about 15 percent of the population are low income or poor.  If they had constructed LICO using a 25 percent or greater "add-on" there would have been only 10-12 percent defined as poor and this would clearly not have been acceptable to the social welfare lobby. On the other hand, the use of 15 percent (or less) "add-on" would have resulted in substantially more than 22 percent classified as poor, and that would have been very difficult to sell to the government and ultimately to the taxpayers who could expect to pay the cost of the alleviation of such distress. A value around 20 percent may well have been viewed as a politically acceptable compromise.

At first glance, there is a tendency to regard LICO as a composite of both absolute and relative definitions of poverty. The reference to expenditure on necessities gives the appearance, albeit superficially, that it is measuring subsistence. However, the close connection to average spending patterns and the proximity of LICO values to such obvious "relative" poverty lines as that constructed by the CCSD give LICO the appearance of having a strong "relative" component. The indirect nature of the construction of LICO explain, to some extent, the confusion and superficiality regarding its classification.

An absolute definition involves the development of a list of basic necessities and the costing of that list. LICO does not do this, not even obliquely. The dollar values for LICO calculated by Statistics Canada are derived from the proportion that an average family spends on food, clothing and shelter, not the amount that an average family needs to cover a (wider) list of necessities. Does this indirect connection to average spending patterns make LICO a relative measure of poverty? It would be fair to say that LICO is a relative measure but not a very good one.

As has already been mentioned, annually updated LICO lines, using the CPI, fall behind relatively to the extent that real living standards improve. There are problems, however, even if LICO is frequently rebased. LICO is defined as that level of income at which a household spends a + .2 of their income on necessities, where a is the ratio of necessities to income for the average household. As average real income rises, we move rightward along the necessities consumption function and a declines. The decrease in a pivots line OA downward leading to a higher "poverty line." This process gives LICO its relativity. However, as long as necessities are income inelastic, incomes rise faster than a declines. This means that rebased LICO falls as a proportion of income. For example, forecasts using equations (1) and (2) show that a doubling of average real income from current levels and rebasing LICO annually will reduce the LICO to average income ratio by 68 percent. Long before that happens, LICO will have been abandoned by relativists. LICO is clearly in trouble. The decline in a has in fact slowed over the years. It was .50 in 1959, .42 in 1969, .385 in 1978, and .362 in 1986. Given its existing methodology, LICO will not keep pace with rising living standards. In 1980, LICO as a proportion of average income for a family of four was .47. By 1988 that proportion had declined to .39. However, even if LICO had been rebased using the 1986 FAMEX data, the LICO to average income ratio would have still fallen to .425.

This analysis is predicated on the assumption that the necessities consumption function remains fairly stable. It could change however, if there is a fundamental variation in people's behavior regarding necessities. Most favorable for LICO would be a systematic increase in the marginal propensity to consume necessities. This change would work to offset the slowdown in the decline of a and result in a faster rising LICO. Visually, this amounts to a counter- clockwise pivoting of BC in figure 3-1. This combined with the downward pivoting of OA due to the decline in a produces large increases in X*. Such a rise in the necessities propensity could be the result of a significant change toward environmentally friendly production processes. Families could be forced to spend a much greater proportion of income on basic necessities because of the sharp rise in costs of non-polluting production. This possibility combined with the recent trend toward "cocooning" (families spending more time and money in their homes) may rescue LICO. On the other hand, technological improvements may permit a continuation of the long-term decline in the real cost of necessities. This would allow the necessities propensity to stabilize or even fall. It would be dangerous to bet against technology.

These and other difficulties have prompted Statistics Canada to undertake a review of LICO. In a recent discussion paper, Wolfson and Evans (1990) rather thoroughly outline the problems with LICO and suggest a number of possible options including scrapping LICO and starting over. Some of its disadvantages, such as the inability to account for such considerations as subsidized shelter and mortgage-free home ownership; in kind benefits and services; wealth; and underreporting of income—all of which tend to overstate the incidence of "low income"—are, of course, common to all poverty lines. A number of criticisms, including the arbitrary 20 percent parameter, are unique to LICO. Most compelling of these has to do with its interpretation, an issue the authors skirt around rather than deal with head on. The fact is that a "low income" line is essentially meaningless unless it is understood as a poverty line. But "Statistics Canada repeatedly insists that the LICOs are not poverty lines" [my emphasis]. Then what exactly are they? Is someone living at the LICO level better off or worse off than someone living at the poverty level? This ambiguity results, predictably, in varied interpretations. Some users regard LICO lines as bare survival levels while others see them as "comfort lines" consistent with a modest array of social amenities.

The authors correctly point out that the complexity of LICO methodology contributes to the difficulties with interpretation. I think it is fair to say that most users do not fully understand how LICO lines are constructed. Their use of LICO as poverty lines is based on their confidence in the reputation of Statistics Canada and not on their own critical evaluation of the worth of LICO. Recognizing this, the agency argues that a simpler, and easy-to-understand methodology would be desirable. Indeed it would.

While readily admitting that LICO methodology is flawed in a number of ways, the authors come to a rather extraordinary conclusion. "However, the current figures relating to the incidence of low income are generally well accepted. Thus, it is worthwhile considering alternative methods which would produce results similar to the existing LICOs in terms of dollar levels for the LIL's (low income lines) and the overall incidence of low income for a recent year, while avoiding to the extent possible the deficiencies of the current method." In other words, Statistics Canada is claiming that they have the cut-offs about right and the proportion of "low income" Canadians about right, now all they need is a better methodology capable of generating those values. It is hard to believe that this highly respected institution would even consider such an approach. It seems to me that the correct, scientific method would be to start with a clear and precisely defined concept, followed by a carefully explained and justifiable methodology for measurement and then finally, actual measurement, letting the chips fall where they may. By starting with the "right" answer and working backwards to find a way of getting there, Statistics Canada would be rejecting scientific methodology. Is their goal to deliver "acceptable" values or the truth?

The claim by Statistics Canada that it does not promote LICO as poverty lines is not convincing. Virtually every individual and group studying poverty in Canada uses LICO as the "official" poverty line. In spite of their disclaimer, the agency has consistently acquiesced in the use of LICO as a poverty line. It must be judged as such. In my view, LICO is neither a useful nor a credible tool for measuring poverty. It is based on an entirely arbitrary assumption that anyone spending 20 percent more than average on food, shelter and clothing is "poor." It does not relate in any way to the cost of living facing the poor. It uses, indirectly, a relative approach but one which is badly flawed. There is no question that it has already fallen behind compared to average living standards and, under the most likely assumptions for the future, LICO stands to lose ground even faster. I can't imagine why anyone who favours a relative approach to the measurement of poverty would prefer LICO to a simple, elegant and clearly relative measure such as that used by the CCSD. LICO's complex methodology gives it a mystique and credibility that is entirely undeserved. The fact is that the use of sophisticated statistical analysis is wasted on such an ambiguous and ultimately meaningless concept, "low income." It is hard to understand why Statistics Canada would cling to so flawed and inadequate a measure of something so important.

For almost a century, Statistics Canada (and its forerunner, the Dominion Bureau of Statistics) have been in the business of telling Canadians about themselves. They have developed a well earned reputation for technical skill and integrity in their selection, analysis and presentation of information about this country and its people. Regrettably, Statistics Canada has not yet developed a way to tell us about our poor. It seems unbelievable that at this point in our history we simply do not know how many among us are unable to acquire all the basic necessities of life. We have a mountain of information of far less interest and importance.

My advice to Statistics Canada would be to develop a genuine poverty line, one which is related to the cost of a list of basic necessities. Everyone recognizes that it would be impossible to obtain universal agreement on a list of necessities. Without question, what is considered essential is relative to a particular time and place. Therefore, my preference would be to limit the list to physical necessities, have a panel of experts make judgements or compromises where disagreements cannot be resolved through discussion and then keep the list fixed. The measurement of poverty requires a unambiguously constant standard to permit valid intertemporal comparisons. Canadians deserve to know how many people in this country have resources insufficient to acquire all the necessities of life. We should know many other things as well, but we should at least know this.

A "necessities-style" poverty line would be far less complex, far less arbitrary and because it would be based on a definition which directly describes a state of impoverishment (the absence of any physical necessity) far more credible than the existing approach. I also believe it would be much less expensive to calculate. Indeed, Statistics Canada and other government agencies are already collecting much of the needed data. It should be possible to determine poverty lines and poverty rates back at least as far as the 1970s. With this information we can find out whether (and the extent to which) poverty is decreasing in Canada.

Some will argue that, using this approach, poverty is now or will soon be virtually eliminated and that we need another measure of "well-being" related in some way to improvements in living standards. It may well be desirable for Statistics Canada to produce a companion measure to the poverty line which I might refer to as a "comfort level." This would involve calculating or estimating the cost of an expanded list of items (necessities as well as non necessities) on a regular basis. While the development of such a list would be far more difficult and involve a greater degree of subjective judgement, on balance it is likely to be worthwhile. It is both interesting and useful for Canadians to know how many of us are unable to attain a satisfactory level of comfort as judged by a panel of experts. However, such a measure cannot be a substitute for the poverty line. The determination of the number of people living "in poverty" should be the first priority.