In table 1 (attached) we will look more closely at changes in the income distribution. A summary measure like G cannot indicate where in the distribution any changes occur. One way of shedding light on this is to divide persons into ten equal income classes (deciles) sorted by the size of income. Decile 1 will then be the ten percent with the lowest household income, whereas decile 10 will be the ten percent with the highest income. The table shows the percentage of total income available to each decile. If the distribution was entirely equal, each decile group would have available 10 percent of the total income.

Table 1 (attached) confirms that income dispersion has taken place in the period, and that changes have particularly occurred at the top of the distribution. The decile with the lowest equivalent income (decile 1) has recorded a reduction in its share of total income from 4.1 percent in 1986 to 3.7 percent in 1997. We also find that all other decil groups, with the exception of decile 10, have also gradually recorded a reduction in their share of total income in the period. The decile with the highest equivalent income (decile 10) has, on the other hand, increased its share of total income by a considerable margin at the expense of all others, i.e. from 18.6 percent in 1986 to 22.1 percent in 1997.

In table 2 (attached) we look more closely at income changes in the period 1986-1997. The table shows average equivalent income for the decile groups, measured in constant prices. The table shows that all income groups had an increase in the level of income, but that decile 10 had an increase substantially higher (34 percent) than other income groups, while average income of decile 1 increased somewhat less than other groups. As a consequence the ratio between average income of decile 10 and decile 1 increased in the period, from 4.5 in 1986 to 6 in 1997.

In addition to look at changes in average income for the decile groups we also presents changes in decile thresholds (table 3, attached). Decile threshold P10 will be the observation which separates decile groups 1 and 2, while decile threshold P90 will be the cut-off between decile groups 9 and 10. P50 will then be the median, i.e. the middle observation in the distribution.

Table 3 (attached) gives a more subtle picture than what was presented in table 2. It is still the case that persons at the top of the distribution have increased their income more than persons at the bottom, but the differences are now considerably smaller. We see, for example, that P90 only rose by 10.4 percent between 1986 and 1997, while the corresponding increase for P10 was 6.4 percent. This may be interpreted to mean that the sharp growth in average income for decile group 10 was primarily due to the small number of persons in this group that has recorded a sharp increase in income. While this ratio has risen substantially with regard to decile averages (see table 2), the changes are considerably smaller with regard to the ratio of decile thresholds. The ratio between P90 and P10 was only marginally higher in 1997 (2.9) compared to 1986 (2.8).

When, on the one hand, the ratio between decile cut-offs only changed marginally in the period, while, on the other hand, the ratio between decile averages increased substantially, this suggests that there has not been a sharp rise in the number of "rich" and "poor" households. It is more likely that the "rich" became richer between 1986 and 1997.

In the next section we will examine whether there have been changes in the structure of household income, which may help to explain the observed changes in inequality.

According to the income account (see Technical Box), total household income can be divided into four different types of income: wage and salary income, net income from self-employment (after depreciation), property income (i.e. interest income and the return on securities) and transfers (pensions and social security benefits).

Table 4 (attached) shows how these income components are distributed among the various decile groups for the years 1986, 1991 and 1997. We see, for example, that the decile of persons with the highest equivalent income (decile 10) received 19 percent of all wage and salary income in 1986, whereas the decile with the lowest income only received 1 percent of all wage and salary income. The distribution of income from self-employment was considerably more skewed than for wages. Here, the decile with the highest income received as much as 39 percent of all income from self-employment. As expected, the situation was the reverse for transfers. Persons with the lowest equivalent income were also those who received the bulk of the transfers.

The distribution of wage income has been relatively stable in the period from 1986 to 1997. Decile group 10, however, has increased its share of wage income slightly, although this change occurred mainly between 1986 and 1991. The distribution of self-employment income has also been fairly stable in the period even though, here as well, we register an even greater concentration at the top of the distribution. The most dramatic shifts, however, have occurred in property income. The table shows that while decile group 10 accounted for 30 percent of all property income in 1986, it accounted for as much as 75 percent in 1997, and most of this change took place in the 1990s.

It is thus particularly property income, which has become more unequally distributed in the 1990s, and it appears that it is primarily changes in share dividends and realised capital gains that have resulted in a more unequal distribution of income. Share dividends have traditionally been a small source of income for Norwegian households. According to Tax Return Statistics, only about NKr 2 billion was paid as dividends to households for each of the years 1986 to 1991. This changed substantially in 1993 when close to 9 billion (1997-NKr) was paid as dividends. Dividend payments increased further in successive years.

Net realised capital gains increased even more than dividends between 1986 and 1997, i. e. from less than one billion Nkr in 1986 to more than 13 billion Nkr in 1997. Note, however, that some of this increase can be explained by changes in tax rules. Prior to 1992 profit from the sales of shares that had been owned in more than 3 years and that were registered at the Stock Exchange was tax-free income in Norway, and thus not included in the definition of income.

Share dividends and realised capital gains are, however, sources of income that are very unevenly distributed among households. In 1996, for example, the 10 percent with the highest equivalent income received 89 percent of all share dividends.

We have defined low-income on the basis of the relative distance to the median equivalent household income, for instance 50, 60 or 70 percent of the national median equalised income. This is an approach often used in international poverty studies. We do not, however, use the concept 'poverty' to describe household with such low income, because it is now generally recognised that annual cash income below a certain threshold is not necessarily identical with poverty.

Table 5 (attached) shows the number of individuals with income below certain cumulative thresholds of equivalent median household income. As can be seen there has only been a modest increase in the number of individuals with income below half of the median, i.e. from 6.4 percent in 1986 to 7.4 percent in 1997. However, the increase in the number of individuals with income slightly above this threshold, for instance 70 percent of the median is more substantial, i. e. from 19.4 percent in 1986 to 21.4 percent in 1997.

There are two age groups that stand out as being more likely to have low income than other groups, i.e. those between 18 and 25 year old and those that are 75 years or older (table 6). Within these two age groups there is a clear over-representation of women with low incomes.

The trend from 1986 to 1996 shows, furthermore, that while there has been a reduction in the number of elderly people with low income, there has been an increase in the number of young people with low income.

Technical Box

This analysis is based on data from the annual Income Distribution Survey for households (The IDS). The IDS are sample surveys, which collects most of its data from Income Tax Records. Household composition is, however, established after a household interview. Non-response households are included in the survey, but missing information on actual household composition is substituted with information on 'family composition', which can be derived from the Central Population Register. For many households there will be no difference between these two concepts.

The table below reports the number of households and individuals in the IDS from 1986 to 1997.

After-tax income is derived on the basis of the following account:

1. Earned income

• wage and salary income (including sickness benefits)
• net income from self-employment

2. +Property income

• interest received
• share dividends
• net realised capital gains
• other property income

3. +Transfers

• social security benefits
• occupational pension, annuities etc
• unemployment benefit.
• alimonies etc.
• family allowances
• housing allowances
• scholarships
• parent's tax deduction
• social assistance

4. =Total income

5. -Assessed taxes and negative transfers

6. =After-tax income (4-5)

No account has been taken of imputed rent for owner-occupied accommodation or of interest payments.

In the analysis, the unit of analysis is the person, but each person has been allotted the household's income. From an economic welfare point of view, it may be argued that it is better to use the individual rather than e.g. the household as the unit of analysis. If the household is used as the unit of analysis, for example, a household of six persons will "count" just as much as a single person, i.e. both the single person and the household of six would be counted as one unit of analysis, i.e. the household. When, on the other hand, a person is used as the unit, all persons will count the same - irrespective of how many persons there are in the household. This approach still involves problems because it assumes that the financial resources of household's are equally distributed among the household members.

When comparing the level of income and living standards for households of varying size, income is often adjusted with the help of equivalence scales or consumption units. An equivalence scale, for instance, provides an indication of how much income a household of four must have in order to achive the same standard of living as a single person with the same income.

The degree of economies of scale may otherwise be expressed through a parameter E. The parameter can vary from 0 to 1. The larger E is, the smaller are the economies of scale assumed for the household, and the higher the income a multi-person household must have in order to be able to have the same living standard as a single person. E=0 corresponds to an unadjusted household income, whereas E=1 corresponds to a household income divided by the number of persons in the household (per capita).

An equivalence scale with an elasticity of 0.5 is used in all the tables presented in this article. This scale assigns weights that are proportional to the square root of the number of persons in the household.

Tables

• Table 1 Distribution of equivalent household income after taxes, between persons. 1986-1997. Decile shares
• Table 2 Changes in equivalent household income after taxes, between persons. 1986-1997. Decile average in 1997-Nkr
• Table 3 Changes in equivalent household income after taxes, between persons. 1986-1997. Decile thresholds (percentiles). In 1997-Nkr
• Table 4 Distribution of household income. 1986, 1991 and 1997. Decile distribution of equivalent household income after taxes. Percent
• Table 5 Cumulative proportions below percentiles of median equvivalent household income after taxes. Persons. 1986-1997
• Table 6 Percent individuals with low income by sex and age. 1986 and 1996