Highlights Highlights Main menu Editor's corner More news Contact us Survey information Back issues Statistics Canada home page In depth Français
 
Statistics Canada logo

75-001-XIE

system menu - text links at bottom of page
mast-head for "Perspectives on Labour and Income"
sub-heading "The online edition"

February 2002     Vol. 3, no. 2

Wealth inequality

René Morissette, Xuelin Zhang and Marie Drolet

The distribution of income has attracted considerable interest in most OECD countries including Canada. In this country, individual earnings inequality has risen since the beginning of the 1980s, at least among male workers (Morissette, Myles and Picot, 1994; Beach and Slotsve, 1996). In contrast, inequality in family disposable income did not increase between the mid-1970s and the mid-1990s (Wolfson and Murphy, 1998). Wealth inequality, however, has not received much attention.

Using the 1984 Assets and Debts Survey and the 1999 Survey of Financial Security, this article examines changes in wealth inequality between 1984 and 1999. Most of the analysis uses three different samples: all families, all families except those in the top 1% of the wealth distribution, and all families except those in the top 5% of the distribution (see Data sources and definitions).

Average and median wealth

Between 1984 and 1999, real (that is, adjusted for inflation) median wealth grew by roughly 10% (Chart A). Real average wealth rose between 28% and 37%, depending on the sample. Excluding the top 1% of families lowered the growth rate of average wealth from 37% to 31%, indicating that the choice of sample is important. The growth in median and average wealth occurred despite an increase in the percentage of families with zero or negative wealth (11% in 1984 versus 13% in 1999).

Because older families have had more time to accumulate savings, wealth increases with the age of the major income recipient, at least until age 65 (Table 1). Shift-share analysis reveals that between 30% and 39% of the growth in average wealth appears to be related to the aging of families. The rest is caused by growth in average wealth within age groups.

Did wealth inequality increase?

Although some segments of the population enjoyed increases in real wealth, others did not—with the result that between 1984 and 1999, wealth distribution became more unequal. note 4  Real median wealth fell in the bottom three deciles but rose at least 30% in the top three (Table 2). Only families in the upper two deciles of the wealth distribution increased their share of total net worth (Chart B). For the other eight deciles, the share of total net worth fell. These results imply that only families in the upper two deciles saw their average wealth increase faster than overall average wealth.

Wealth inequality did not rise uniformly. As measured by the Gini coefficient, it increased much more among non-elderly couples with children and among lone-parent families than among unattached individuals and non-elderly couples with no children (Table 3). Among non-elderly couples with children under 18, real average wealth fell roughly 15% in the second quintile but rose about 20% in the fourth quintile and even more in the fifth quintile (Table 4).

Changes in the wealth structure

The growth of wealth inequality occurred in conjunction with substantial changes in the wealth structure. Real median wealth and real average wealth evolved very differently for different families. First, both rose much more among families whose major income recipient was a university graduate (Table 5). Second, both fell among families whose major income recipient was aged 25 to 34 and increased among those whose major income recipient was aged 55 to 64. The rise was even greater among families whose major income recipient was 65 or over. Third, both increased among Canadian-born families and foreign-born ones living in Canada for 20 years or more, but fell among foreign-born families living in Canada for less than 10 years. Fourth, both increased faster among non-elderly couples with no children than among non-elderly couples with children under 18.

In many population sub-groups, real median wealth grew much more slowly than average wealth. For instance, among families whose major income recipient was aged 25 to 34, real median wealth fell 36% while real average wealth fell only 4%. Similarly, non-elderly couples aged 25 to 54 with children under 18 experienced almost no change in their real median wealth but enjoyed an increase of 30% in their real average wealth (Chart C). note 5 

Young couples with children under 18 with a major income earner aged 25 to 34 experienced drastic changes. Their real median and average wealth fell 30% and 20%, respectively. The percentage of these couples with zero or negative wealth rose from 9.5% in 1984 to 16.1% in 1999. The decline in median wealth reflects a 39% decrease in net equity on the principal residence, which more than offset a 12% increase in financial wealth. note 6 

Among families whose major income recipient was between 25 and 34, the decline in real median wealth was unlikely caused solely by a decrease in real median after-tax income. While the former dropped by 36%, the latter fell by only 7%. note 7  However, growth rates of average wealth and average after-tax income diverge to a much lesser extent (-4% and 1%, respectively). Inheritances and inter vivos transfers (for example, parental financing of education or of a house down payment) are unlikely to be factors since the parents in 1999 are unlikely to be poorer than those in 1984.

In contrast, the dramatic increase in real median wealth and average wealth (56% and 51%, respectively) of families whose major income recipient was 65 or older likely reflects a combination of factors: larger inheritances possibly received by the 1999 respondents; higher income from private pensions; and higher income from the Canada or Quebec Pension Plan, Guaranteed Income Supplement, or Old Age Security.

In summary, families whose major income recipient was a new entrant to the labour market—that is, a young individual or a recent immigrant—lost ground relative to older families. Furthermore, within a given age group, families whose major income recipient did not have a university degree lost ground relative to families headed by a university graduate. note 8 

Aging and wealth inequality

The substantial changes in family structure over the last two decades may have affected wealth inequality. Specifically, the growing proportion of unattached individuals and lone-parent families, which generally have lower-than-average wealth, may have contributed to the growth of wealth inequality. Accordingly, the 1999 data were re-weighted so that the relative importance of various types of families was equal to that observed in 1984. note 9  The inequality measures resulting from this re-weighting were then calculated.

The inequality measures used were the Gini coefficient, the coefficient of variation (CV), and the exponential measure. While the Gini coefficient is sensitive to changes in the middle of the wealth distribution, the coefficient of variation is sensitive to changes at the top, and the exponential measure is sensitive to changes at the bottom (Table 6).

Whether or not changes in family structure tended to increase wealth inequality cannot be said with certainty. When all families are considered, the effect is ambiguous. Applying the 1984 family structure to the 1999 data decreases the Gini coefficient and the exponential measure, but increases the CV (compared with their 1999 actual values). For the sample in which the top 1% of the wealth distribution is excluded, wealth inequality would have been lower in 1999 if the composition of families had remained the same as in 1984. For this sample, changes in family structure accounted for 14% to 22% of the growth in wealth inequality. note 10  For the sample in which the top 5% of the wealth distribution is excluded, changes in family structure accounted for 25% and 23% of the growth in the Gini coefficient and the CV, respectively. note 11 

The aging of the population may also have affected wealth inequality. However, its effect is unclear since it is associated with a decline in the relative importance of young families, who have lower-than-average wealth, and an increase in the relative importance of older families, which tend to have higher-than-average wealth. To assess the effect of aging, the 1999 data were re-weighted with the 1984 age structure, using six age groups. If the 1984 age structure had prevailed in 1999, wealth inequality would have been higher than it was in 1999. Hence, the aging of the population tended to reduce wealth inequality.

What would wealth inequality have been in 1999 if permanent income note 12  and other attributes of families had remained at their 1984 levels and families had kept their 1999 net worth? The other attributes to be considered are age of major income recipient (five age groups), education level of major income recipient (two levels), a lone-parent family indicator, family size, provincial controls, and a rural-urban indicator. note 13  For all three samples, the hypothetical inequality measures for 1999 are always higher than the actual inequality measures. This means that if the distribution of permanent income and other family attributes had remained at their 1984 level and families had kept the net worth observed in 1999, wealth inequality would have been higher than it was in 1999. At the very least, this suggests that permanent income and other socio-demographic characteristics as measured with cross-sectional data are not major factors behind the growth of wealth inequality.

Explaining wealth inequality

Several factors may have contributed to the growth of wealth inequality. First, young people have been staying in school longer before entering the labour market, thus decreasing the number of years over which they have had significant incomes. This and the greater debt load of students (Finnie, 2001) probably account for part of the decrease in their real median wealth. note 14  Second, the booming stock market of the 1990s likely contributed to the rapid revaluation of financial assets observed in Canada over the last decade (Yan, 2001). Since financial assets are held predominantly by families at the top of the wealth distribution, this revaluation is likely to have contributed to the growth of wealth inequality. Third, easier access to credit or changes in preferences may have induced some low-wealth families to accumulate debt to finance expenditures, thereby decreasing their net worth. Fourth, increases in contributions to RRSPs made by families in the middle of the wealth distribution could have widened the gap between them and poorer families if these greater contributions caused an increase in their savings rate. Fifth, differences between low-wealth and high-wealth families in the growth of inheritances and inter vivos transfers may also have played a role. These factors cannot be quantified with existing data sets.

Summary

Wealth inequality increased between 1984 and 1999. The growth was associated with substantial declines in real average and median wealth for some groups, such as young couples with children and recent immigrants.

Only the 10th decile (and for some samples, the 9th decile) increased their share of total net worth between 1984 and 1999. Wealth inequality increased more among non-elderly couples with children and lone-parent families than among unattached individuals and non-elderly couples with no children.

Real median wealth and real average wealth rose much more among families whose major income recipient was a university graduate than among other families; both fell among families whose major income recipient was aged 25 to 34 and increased among those whose major income recipient was 55 or over.

The aging of the population between 1984 and 1999 had two important effects: it tended to increase the average wealth of Canadians and to reduce wealth inequality.

Young couples with children experienced a 30% decline in their median wealth. This led to a substantial decrease in their net equity on principal residence. Furthermore, a growing proportion had zero or negative wealth and therefore could not rely on savings to provide liquidity in periods of economic stress.

 

Data sources and definitions

The 1984 Assets and Debts Survey (ADS) was a supplement to the May 1984 Survey of Consumer Finances. The 1999 Survey of Financial Security (SFS) was conducted from May to July 1999. For both surveys, the sample was based on the Labour Force Survey frame and represents all families and individuals in Canada except residents of the territories, members of households located on Indian reserves, full-time members of the Armed Forces, and residents of institutions. note 1  Data were obtained for all family members aged 15 and over.

Some differences between the two surveys are worth noting. First, in ADS, all information on components of assets (except housing) and debts were collected for each member of the family aged 15 years and over and then aggregated at the family level. In contrast, in the SFS, information was collected directly at the family level. Second, unlike ADS, the SFS contained a supplementary 'high-income' sample (consisting initially of about 2,000 households), which was included to improve the quality of wealth estimates. note 2  The final sample of ADS included 14,029 families, and the SFS sample 15,933. Families include unattached individuals.

Because records of the current value of assets and debts are not as readily available as records of income, the quality of wealth data is viewed as lower than the quality of income data. Also, the value of real assets (such as housing and vehicles) is judged to be of higher quality than that of financial assets.

To make the concept of wealth comparable between the two surveys, contents of the home, collectibles and valuables, annuities, and registered retirement income funds, which were not included in the 1984 survey, were excluded from the 1999 data.

The wealth of a family is defined as the difference between the value of its total assets and the amount of its total debts. Excluded are the value of work-related pension plans, and future entitlements to social security provided by the government in the form of Canada or Quebec Pension Plan benefits or Old Age Security. Also excluded are the family's human capital, measured in terms of the value of the discounted flow of future earnings for all family members.

One particularly difficult issue with wealth data is the measurement of the upper tail of the wealth distribution. Using a variety of data sources, Davies (1993) estimates that, using ADS, the share of total wealth held by the top 1% of families in 1984 could increase from 17% to between 22% and 27% after adjustments. Similarly, the share of total wealth held by the top 5% of families could increase from 38% to between 41% and 46%.

Since this article compares wealth at two points in time, a further complication is that the degree of truncation of the wealth distribution may change over time. More precisely, assume that the true wealth distribution was unchanged between 1984 and 1999. Extending the argument of Davies (1993,160) to the analysis of changes in the wealth distribution, if no Canadian family with wealth over $10 million consented to an interview in 1984, and if no Canadian family with wealth over $50 million consented to an interview in 1999, ADS and SFS would show an (incorrect) increase in wealth inequality—which could simply be due to the use of better interviewing techniques in the later survey than in the earlier one. note 3  For these reasons, most of the analysis described in this article uses three different samples: all families, all families except those in the top 1% of the wealth distribution, and all families except those in the top 5%.

The Gini coefficient and the exponential measure are two measures of inequality, which would equal one if one family owned the total wealth of society while all others had zero wealth. Both measures would equal zero in the case of perfect equality—that is, if all families had the same wealth. The coefficient of variation, defined as the ratio of standard deviation to the mean, would also equal zero in the case of perfect equality. It would increase—but not necessarily equal one—if one family owned the total wealth of society while all others had zero wealth.

Notes

  1. These include institutions such as penal institutions, mental hospitals, sanatoriums, orphanages and seniors residences.
  2. Having a high-income supplement in 1999 increased the precision of wealth statistics (for example, average, median, and inequality measures) compared to ADS, while still leaving them unbiased (like those of ADS).
  3. Weighting procedures cannot correct this problem since no family with wealth over $10 ($50) million would be observed in the sample.
  4. More precisely, if the bottom 0.5% of the wealth distribution is excluded, one can say unambiguously that wealth inequality rose between 1984 and 1999—that is, the 1999 Lorenz curve lies below the 1984 Lorenz curve at all points of the wealth distribution. See Morissette, Zhang and Drolet (2002) for a detailed analysis.
  5. Couples with children under 18 are defined as couples with at least one child of the major income earner under 18.
  6. Financial wealth is net worth minus net equity in housing and net business equity. Median financial wealth of young couples with children under 18 rose from $7,200 in 1984 to $8,000 in 1999. Their median net equity on principal residence fell from $26,000 in 1984 to $16,000 in 1999.
  7. This statement must be made with caution since changes in wealth depend, among other things, on changes in the set of annual after-tax incomes received in the past, and not only on changes in current after-tax income measured by cross-sectional data. In other words, while current after-tax income dropped by 7%, accumulated after-tax income could have dropped by more than 7%.
  8. Since there is evidence that financial assets were better reported in 1999 than in 1984 (Morissette, Zhang and Drolet, 2002), the growth rates of wealth observed for groups with growing wealth must be interpreted with caution. They likely represent an upper bound for the true growth rates of wealth of these groups.
  9. Families were defined according to 14 categories.
  10. If the 1984 family structure had prevailed, the coefficient of variation in 1999 would have been 1.498 rather than 1.517. Hence, 22%—that is, (1.517-1.498)/(1.517-1.429)—of the growth in the coefficient of variation can be accounted for by changes in family structure.
  11. The decrease in the exponential measure for this sample (in the actual data) occurs because the Lorenz curves for 1984 and 1999 cross below the bottom 0.5% of the wealth distribution.
  12. A family's permanent income is defined as the predicted income of this unit when the major income recipient is aged 45 and the spouse (if present) age is set equal to what it would be when the major income recipient is aged 45. See Morissette, Zhang and Drolet (2002) for further details.
  13. To implement this approach, the 1984 and 1999 data were first pooled. Second, a logit model was estimated in which the dependent variable equals 1 if a family unit with a given level of permanent income and other given attributes was observed in 1984, 0 if it was observed in 1999. Third, the 1999 data were re-weighted by the factor (Pi84/Pi99).(K99/K84), where Pi84 and Pi99 are the probability of family i being observed in 1984 and 1999, respectively, and K99 and K84 are the sum of weights for 1999 and 1984, respectively. Fourth, after the 1999 data were re-weighted, the inequality measures were calculated. The explanatory variables used in the logit model include permanent income and other attributes defined above. For further details, see DiNardo, Fortin and Lemieux (1996).
  14. Young individuals now get married later, thereby delaying the benefits from the economies of scale associated with cohabitation. However, this may be offset by some young individuals staying longer with their parents or cohabiting in other ways. Similarly, the downward shift in the age-earnings profile of young men (Beaudry and Green, 1997) may have tended to reduce real wealth of young men. However, its effect may have been partly offset by the growing number of dual-earner couples among young families.

References

  • Beach, C.M. and G.A. Slotsve. Are We Becoming Two Societies?: Income Polarization and the Myth of the Declining Middle Class in Canada. Social policy challenge series, vol. 12. Toronto: C.D. Howe Institute, 1996.
  • Beaudry, P. and D.A. Green. "Cohort patterns in Canadian earnings and the skill-biased technical change hypothesis." University of British Columbia Department of Economics Discussion Paper: 97/03. British Columbia: University of British Columbia, 1997.
  • Davies, J.B. "The distribution of wealth in Canada." Research in Economic Inequality, Vol. 4, E. Wolff (ed.). Greenwich, CT: JAI Press, 1993, 159-180.
  • DiNardo, J., N.M. Fortin and T. Lemieux. "Labour market institutions and the distribution of wages, 1973-1992: A semiparametric approach." Econometrica 64, no. 5 (September 1996): 1001-1044.
  • Finnie, R. "Student loans: the empirical record." The Canadian Journal of Higher Education, 2001 (forthcoming).
  • Morissette, R., J. Myles and G. Picot. "Earnings inequality and the distribution of working time in Canada." Canadian Business Economics 2, no. 3 (Spring 1994): 3-16.
  • Morissette, R., X. Zhang and M. Drolet. "The evolution of wealth inequality in Canada, 1984-1999." Analytical Studies Branch research paper, no. 187. Statistics Canada, Ottawa, 2002.
  • Wolfson, M.C. and B.B. Murphy. "New views on inequality trends in Canada and the United States." Monthly Labor Review 121, no. 4 (April 1998): 3-23.
  • Yan, X. "Understanding saving and wealth accumulation." Working paper. Income and Expenditure Accounts Division, Statistics Canada, 2001.

Author

The authors are with the Business and Labour Market Analysis Division. René Morissette can be reached at (613) 951-3608 or rene.morissette@statcan.gc.ca; Xuelin Zhang at (613) 951-4295 or xuelin.zhang@statcan.gc.ca; Marie Drolet at (613) 951-5691 or marie.drolet@statcan.gc.ca.

Statistics Canada FIP identifier Government of Canada wordmark
Highlights ]
Main menu | Editor's corner | More news | Contact us | Survey information | Back issues ]
Statistics Canada home page | In depth | Français ]

© Statistics Canada - Conditions of  use Published: 2002 02 22