CHANGES IN MEDIAN HOUSEHOLD INCOME:
1969 TO 1996
From 1969 to 1996, median household income rose a very modest 6.3 percent in constant dollars (from $33,072 to $35,172). At the same time, per capita income rose by a robust 51 percent in constant dollars (from $11,975 to $18,136). The two data series are not inconsistent, because the median simply identifies the income level of the household in the middle of the income distribution while per capita income is calculated by dividing aggregate income by the total number of individuals in the universe. Although the two series are not inconsistent, the difference between the two is sufficiently large that an attempt to identify the reasons behind the difference seems useful. The material below will focus on changes in income inequality and changes in the size and composition of households as factors that help explain the different growth rates in the two income series.
The fact that median household income showed only a small gain during the same time period that per capita income rose sharply suggests that a major change occurred in income inequality. The distribution of income changed dramatically over the period, but it changed in such a way as to have a small effect on the median.
As part of the attempt to identify reasons why the growth rates in median household income and per capita income look so different, this report presents data on the median income of households that have been categorized according to their size and composition. A time-series on the median income of households can be misleading if the characteristics of households change in a substantial way over time. The 1969 to 1996 stagnation in median household income may, in fact, be largely a reflection of changes in the size and composition of households rather than a reflection of a stagnating economy. The data in this report do show that the characteristics of households changed substantially from 1969 to 1996, and data users should be aware of these changes when attempting to interpret time-series data on the median income of households.
Most of the data presented in this report come from special tabulations of the March Current Population Surveys. The medians are based on data for individual households and differ slightly from the official Census Bureau data on the median incomes of households which are calculated after individual cases have been grouped into intervals. The 1969 to 1996 period was chosen because 1969 is the earliest income year for which the Census Bureau has produced a data file that can be easily accessed by analysts and 1996 is the most recent year for which data are available.
The data in Table 1 show 1969 to 1996 changes in a number of economic statistics including some from the National Income and Product Accounts as produced by the Bureau of Economic Analysis (BEA). The various series are presented to give some idea of the time-series that are available for the purpose of measuring the economic well-being of the consumer sector. Both the Census Bureau and the BEA provide data that can be used to measure per capita income, and the two sources provide estimates of the percent change in per capita income that are quite close when differences in the income definition are considered. A major difference in the income definitions used by the two agencies is that the BEA counts noncash benefits and employer contributions to social insurance programs as part of personal income while the Census Bureau does not (the Census Bureau counts only money income). The fact that the BEA series on per capita income rose 62 percent over the period while the Census Bureau series rose 51 percent is consistent with the fact that supplements to wages and salaries ( part of the BEA definition) were a considerable larger share of total compensation in 1996 than they were in 1969 (21.8 percent versus 11.6 percent). /1
Other data in Table 1 show why per capita income has shown a sizeable gain at the same time that median household income has shown little change. The incomes of households below the median were essentially stagnant over the 1969 to 1996 period while the incomes of households above the median surged. The average income of households at or below the median grew by only 3.9 percent, but the average income of households above the median grew by 30 percent.
Two factors that are likely to be associated with changes in household income are (1) changes in the employment status of household members, and (2) changes in the educational attainment levels of household members. Data on the employment status of members of households with two or more persons (presented in Table 1) help explain the differing rates of growth in average income. In 1996, the average number of year-round full-time workers in two-or-more-person households with incomes above the median was 1.4, up 14 percent from the 1969 level. The comparable figure for two-or-more-person households with incomes at or below the median was 0.5, unchanged from the 1969 level.
The impact of changes in the level of educational attainment is less clear. Data in Table 2 show a sharp increase in the educational attainment levels of individuals regardless of the income position of the household. From 1969 to 1996, in households with incomes above the median, the proportion of individuals 25 years old and over with a college degree grew from 16 percent to 33 percent. Over the same time period, in households with incomes below the median, the proportion of individuals 25 years old and over with a college degree grew from 5 percent to 11 percent. The fact that the average income of households at or below the median was essentially stagnant at the same time that the educational attainment levels of people in these households rose sharply presents something of a puzzle. The result leads to the conclusion that, for some very significant proportion of the population, increases over time in educational attainment did not translate into increases in income.
The reference unit for most of the discussion above has been the household. It was noted earlier that comparisons over time can be misleading if the characteristics of the reference unit change substantially. The data in Table 3 and in Figure 1 clearly show that there have been substantial changes. The changes mean that is is difficult to interpret time-series changes in median household income without an examination of data for specific types of households. For the analysis in this report, households have been categorized into 12 types. Data in Table 3 show the percent distribution of households by type in 1969 and 1996 and also show how certain critical characteristics have changed over the same time period for each type of household. The most important changes in the size and composition of households were the relative decline in the number of married-couple households with children (from 41 percent to 26 percent), the relative increase in the number of households with children but no spouse present (from 6 percent to 11 percent), and the relative increase in one-person households (from 17 percent to 25 percent).Table 3
The data in Table 3 also show that there have been important changes over time within household types. One change has been the increased labor force participation of women. Table 3 data for married couples show that, from 1969 to 1996, the proportion of wives working year-round full-time rose from 17 percent to 39 percent in households with children, from 42 percent to 60 percent in households with no children and a householder less than 40 years old, and from 31 percent to 46 percent in households with no children and a householder between the ages of 40 and 64 years. Material presented later in this report isolates the income effect of these increases in labor force participation by showing time series data on the median incomes of married-couple households both when the earnings of wives are included in the income definition and when the earnings are excluded.
Another major time series change in the characteristics of households concerns the increased educational attainment level of householders or husbands and wives. The data in Table 3 show that 25 percent of all householders had a college degree in 1996, up sharply from the 1969 figure of 13 percent. In 1996, in married-couple households with children, 29 percent of husbands and 26 percent of wives had a college degree compared with 1969 figures of 16 percent and 8 percent. In married-couple households with no children and with a householder under 40 years old, 37 percent of husbands and 39 percent of wives had a college degree in 1996 (the two percentages are not statistically different). The comparable 1969 figures were 22 percent and 16 percent. In fact, for 11 of the 12 household types, the proportion of householders (or husbands and wives) with a college degree was higher (generally much higher) in 1996 than in 1969 (the change for male householders with children and no spouse was not statistically significant).
The increases in year-round full-time employment and in educational attainment for most of the 12 types of households leads to the expectation that the median income of most types of households should show an increase over the 1969 to 1996 period.
Data in Table 4 and Figure 2 show that median incomes did increase for most types of households over the 1969 to 1996 period, and for some household types, the increase was substantial. Table 5a presents data on the percent changes in median household income for each type of household. Table 5b presents the same information but with the percent changes shown in terms of annual rates. Tables 4, 5a, and 5b present data for married-couple households that show median income data based on an income definition that excludes the earnings/income of wives as well as for an income definition that includes the earnings/income of wives. In the three tables, data are shown for the full reference period of 1969 to 1996, and for five subperiods. The subperiods, chosen to generally reflect business cycles, are 1969 to 1979 (generally characterized by rising incomes), 1979 to 1983 (generally characterized by falling incomes), 1983 to 1989 (generally characterized by rising incomes), 1989 to 1993 (generally characterized by falling incomes), and 1993 to 1996 (generally characterized by rising incomes).
The median income of married-couple households with children rose 25 percent from 1969 to 1996. The median income of these households grew substantially during the years 1969 to 1979 and 1983 to 1989, and showed a further increase during the years 1993 to 1996. The periods 1979 to 1983 and 1989 to 1993 were periods of declining incomes.
Although the median income of married-couples with children increased substantially between 1969 and 1996, the effect on this increase of excluding the earnings of wives from the income definition is striking. When the earnings of wives are excluded, the median income of these households showed a 2 percent increase over the period, rather than a 25 percent increase.
Married-couple households without children also had substantial gains in median income between 1969 and 1996, and again the increased importance of the earnings/income of the wife was very evident. Among households with a householder under 40 years old, median income increased by 34 percent over the period but only by 17 percent when the earnings of wives were excluded from the income definition. Households with a householder between 40 and 64 years old had an increase of 34 percent when the earnings of wives were counted, but an increase of 16 percent when the earnings were excluded.
The median income of households with a female householder with children and no spouse rose by 10 percent between 1969 and 1996, but the median income of households with a male householder with children and no spouse fell by 8 percent.
Married-couple households with a householder 65 years old or over and one-person households with a householder 65 years old or over showed very substantial gains in median income. The married-couple households had a 57 percent gain over the 1969 to 1996 period (34 percent when the income of wives was excluded from the income definition). The income of these households rose 28 percent from 1969 to 1979, a period when there was a gain in the median income of all households, and rose 17 percent from 1979 to 1983, a period when there was a decline in the median income of all households. The median income of these households fell by 3 percent between 1989 and 1993 and rose 3 percent between 1993 and 1996. Among one-person households with a householder 65 years old or over, the median income of both males and females rose 63 percent over the full period. The median income of single male householders 65 years old or over rose 26 percent from 1969 to 1979 and rose 21 percent between 1979 and 1983. The median income of single female householders 65 years old or over rose 34 percent and 17 percent during the same two subperiods. More recently, the median income of single male householders 65 years old and over failed to show a significant change during either of the two subperiods 1989 to 1993 and 1993 to 1996. Over the same two subperiods, the median income of single female householders 65 years old and over showed a decline of 5 percent and an increase of 6 percent.
Table 6 shows data on the proportion of households with low or high relative incomes. The relative income measure was developed as a means of providing information on the shape of the income distribution at the low and high ends. The measure uses an equivalence scale to adjust household income for the number of persons in the household and determines the relative income measure by comparing each household's adjusted income amount to the median adjusted income amount (for a more complete description of this measure see "Trends in Relative Income: 1964 to 1989," Series P60-177). /2 Households with relative income levels of less than 0.5 are considered to have low relative incomes, and households with relative income levels of 2.0 or more are considered to have high relative incomes.
There are at least two items of considerable interest in the Table 6 data. One striking result is the growth in the proportion of households with high relative incomes. The proportion was 12 percent in 1969 and reached 17 percent in 1993. There was no further rise in the proportion from 1993 to 1996. For eight of the twelve household types, the proportion with high relative incomes was higher in 1996 than in 1969. Two household types --households with a male householder with children and no spouse; and households with two or more persons, a male householder, and no spouse or children -- showed a decline in the proportion with high relative incomes. Men under 65 years old and living alone and women 65 years old and over and living alone had no statistically significant change.
The proportion with low relative incomes was 21 percent in 1969, 23 percent in 1993, and 22 percent in 1996. A second striking result that can be seen in the Table 6 data is the sharp reduction between 1969 and 1983 in the proportion of "elderly" households with low relative incomes. Among married-couple households with no children and with a householder 65 years or over, the proportion with low relative incomes was 34 percent in 1969, 19 percent in 1983, and 18 percent in 1996 (the 1983 and 1996 figures are not statistically different). Among single male householders 65 years or over, the proportion with low relative incomes was 54 percent in 1969, 39 percent in 1983, and 35 percent in 1996. Among single female householders in this age category, the proportion with low relative incomes was 68 percent in 1969, 49 percent in 1983, and 51 percent in 1996.
The time series changes in the proportion of households with low or high relative incomes means that there was a fairly sharp decline, from 67 percent in 1969 to 61 percent in 1996, in the proportion of households with "middle" incomes.
Tables showing data for each of the years 1969 to 1996 for each of the 12 household types can be found at the web site address given below. The tables show data on the number and percent distribution of households by type, the median income of nonmarried-couple households, the median income of married-couple households including and excluding the earnings or income of wives, and the percent of households with low or high relative incomes. Tables showing standard errors for each median income figure are also included.
The Census Bureau's time series on median income is a powerful economic statistic that tracks the progress (or lack of progress) of households in the middle of the income distribution. But the past 30 years have seen dramatic changes in the overall distribution of income while the series on median household income has been remarkably stable. An adequate analysis of trends in consumer income must go beyond median household income to consider a range of income statistics. Time series data on median household income may, in fact, be misleading if substantial changes have occurred in the characteristics of households. This report points out that substantial changes have occurred and urges people interested in long-term income trends to include data on specific types of households as part of their analysis. /3
1/ Data from the Survey of Current Business, August, 1997, Bureau of Economic Analysis, U. S. Department of Commerce.
2/ There is an extensive literature on the use of equivalence scales. The scale used here was chosen for its simplicity and is described by Patricia Ruggles in her 1990 book, Drawing the Line -- Alternative Poverty Measures and Their Implications for Public Policy, The Urban Institute Press, Washington, D.C. The factor used to identify equivalent income levels for households of different size is equal to the square root of the number of persons in the household.
For example, to reach an equivalent level of income, a four-person household would need twice as much income as a one-person household, and an eight-person household would need twice as much income as a two-person household.
3/ For further information on household income statistics see "Money Income in the United States: 1996," Series P60-193. Users interested in a comprehensive analysis of recent changes in income, wages, and labor force participation should also see The State of Working America, 1996-97, Lawrence Mishel, Jared Bernstein, and John Schmitt, Economic Policy Institute Series, Armonk: M. E. Sharpe, 1997.
All sample-based statistics are subject to sampling error, as well as nonsampling error such as survey design flaws, respondent classification and reporting errors, data processing mistakes, and undercoverage. The Census Bureau has taken steps in the form of quality control and editing procedures to reduce errors made by respondents, coders, and interviewers. Ratio estimation to independent age-race-sex population controls partially controls for bias attributable to to survey undercoverage. However, biases exist in the estimates when missed people have characteristics different from those of interviewed people in the same age-race-sex group.
Analytical statements in this report have been tested and meet statistical standards. However, because of methodological differences, use caution when comparing these data with data from other sources.
Contact Geneva A. Burns, Demographic Statistical Methods Division, at 301-763-4214 or on the internet at <Geneva.A.Burns@ccmail.census.gov> for information on (1) the source of the data, (2) the accuracy of the estimates, (3) the use of standard errors, and (4) the computation of standard errors.
Last Revised: Friday, 30-Sep-2011 17:20:23 EDT