(This essay was published in Hong Kong Economic Journal on 16 October 2013)
The key factor driving the dispersion of individual incomes is the distribution of productivity in the working population. The role of education is very important for increasing productivity and is by far the largest investment a person makes in human capital. Other factors like age, gender, marital status, and recent immigrant status are less important.
The dispersion of household income is, however, altogether a different matter.
The central characteristic of household income is that it is the sum of each individual member’s income. Household income, unlike the strong correlation between individual income and own education, may only be weakly correlated with the education of the household head. The dispersion of household income therefore depends on the membership and composition of the household, and the complex interactions of members. The Theil Index decomposition analysis discussed below can reveal only a limited static dimension of their interactions, but it cannot uncover the deep intricate relationships.
The sources of household income dispersion are decomposed into five groups: age, gender, education, marital status, and recent immigrant status of the household head (see Table 1). In the period 1976-2011, age accounted for 31.7% of the change in income dispersion, gender for -4.1%, education for 59.4%, marital status for 12.1%, and recent immigrant status for 11.3%. The corresponding figures for the more recent period 1991-2011 show age accounted for 34.2%, gender for 0.4%, education for 31.4%, marital status for 16.4%, and recent immigrant status for 12.0%.
Table 1: Decomposition of household income dispersion
| Year | Overall Gini Coefficient | Overall Theil Index |
Between group variations | Within group variations | ||||
| Compo- nent due to age |
Compo- nent due to gender |
Compo-nent due to education |
Compo-nent due to marital status | Compo-nent due to recent immigrant status | Residual Compo-nent | |||
| 1976 | 0.442 | 0.452 | 0.005 | 0.01 | 0.077 | 0.008 | 0.004 | 0.348 |
| 1981 | 0.459 | 0.48 | 0.011 | 0.008 | 0.077 | 0.013 | 0.008 | 0.363 |
| 1986 | 0.447 | 0.424 | 0.011 | 0.005 | 0.09 | 0.008 | 0.002 | 0.308 |
| 1991 | 0.485 | 0.487 | 0.014 | 0.005 | 0.119 | 0.009 | 0.007 | 0.332 |
| 1996 | 0.489 | 0.491 | 0.021 | 0.004 | 0.12 | 0.011 | 0.016 | 0.319 |
| 2001 | 0.502 | 0.503 | 0.033 | 0.006 | 0.134 | 0.014 | 0.009 | 0.307 |
| 2006 | 0.515 | 0.53 | 0.033 | 0.006 | 0.114 | 0.017 | 0.014 | 0.346 |
| 2011 | 0.533 | 0.564 | 0.041 | 0.005 | 0.144 | 0.022 | 0.016 | 0.336 |
| 1976-2011 | 0.075 | 0.112 | 0.035 | -0.005 | 0.066 | 0.014 | 0.013 | -0.012 |
| 31.70% | -4.10% | 59.40% | 12.10% | 11.30% | -10.50% | |||
| 1991-2011 | 0.048 | 0.077 | 0.027 | 0.000 | 0.024 | 0.013 | 0.009 | 0.004 |
| 34.20% | 0.40% | 31.40% | 16.40% | 12.00% | 5.50% | |||
| 2001-2011 | 0.031 | 0.061 | 0.008 | -0.001 | 0.01 | 0.008 | 0.007 | 0.03 |
| 12.50% | -0.80% | 16.10% | 12.70% | 11.30% | 48.30% | |||
Note: Only individuals aged 18 or over who were not students are included. Domestic helpers excluded.
Education continues to be an important source of income dispersion among households, but its role has apparently diminished over time compared to other factors, like age, marital status, and recent immigrant status. Why is this the case? And how do we interpret these results?
Impact of Differences in Household Composition
Let us consider a number of different situations. First, not all members of a household join the labor force even if they can find work. Retired individuals choose not to work, but their expenditures are not zero and their living standards are either supported by savings or income transfers. The recent “houseman” or “otaku” (宅男) phenomenon shows that some male household members choose not to work. The number of women who choose not to work is higher.
Second, household income typically rises with the age of the head of the household and peaks before falling off in old age. If the age composition of the population is changing then observed household income dispersion would also change. The generation of baby-boomers in Hong Kong is rapidly ageing and their numbers are very large. This is creating a very large impact on observed household income dispersion as their retirement age approaches. This is evident from figures in Table 1 where age has contributed 31.7% of the rise in income dispersion since 1976.
Third, in a married household many women with young children choose not to work even if they have a previous work history and plan to return to work in the future when their children grow up. When a woman withdraws from the labor force one would not assume that the consequent decline in household income necessarily implies a decline in household expenditures and standards of living. If the child was the result of a planned family choice then the consequent fall in household income should have been foreseen and well prepared for. Marital status has contributed 12.1% of the rise in income dispersion since 1976.
Fourth, gender is not likely to be an important source of variation in household income dispersion because men head most households. Those headed by women tend to be either single or divorced and their effects are probably captured by marital status already. The results in Table 1 shows that this has indeed been the case since 1976.
Fifth, households with recent immigrants are overwhelmingly those from the Mainland. The stock of human capital available in these households is generally lower so recent immigrant status is a source of household income dispersion. Recent immigrant status has contributed to 11.3% of the rise in income dispersion since 1976, which is about the same as marital status but significantly less than either age or education.
Income Dispersion Arising from Changing Household Size
Households vary in their sizes, and their size also changes over their lifecycle. Households make many decisions, like marriage, divorce, fertility, work, retirement, emigration, and so on that impact household size. The increase of small households over time has reduced the average size of households in society.
This growth in small households has demographic, social, and economic origins. When large numbers of adult children of the baby boomer generation leave their parents’ household, many small households are formed as a consequence. When the baby boomer generation enters retirement age, their households also become smaller. These two related changes have been an important factor behind the growth of single-person households over time. Their share in the population has risen rapidly since 2001 (see Table 2). Measured household income dispersion is likely to increase because of these demographic changes. This accounts for the very large contribution of age to increasing household income dispersion.
Table 2: Number of elderly, divorced individuals per thousand households, and number of single-person households
| 1976 | 1981 | 1986 | 1991 | 1996 | 2001 | 2006 | 2011 | |
|---|---|---|---|---|---|---|---|---|
| Number of elderly individuals over age 65 per 1000 households | 225 | 249 | 276 | 292 | 319 | 344 | 343 | 363 |
| Number of divorced individuals per 1000 households | 15 | 18 | 27 | 31 | 49 | 70 | 91 | 107 |
| Number of single-person households per 1000 households | 144 | 152 | 148 | 148 | 149 | 156 | 165 | 171 |
In addition, the growth of divorced individuals has also led to the growth of small households as larger households split in half, into smaller units. Divorce is therefore a new and increasingly significant source accounting for the rise in household income dispersion. The number of divorced individuals per household has increased dramatically from 49 per thousand in 1996 to 107 per thousand in 2011 (see Table 2). The total number of divorced individuals in the population was about 90,000 in 1996 and it increased to 260,000 in 2011. Both the rise of divorced individuals and single persons wanting to live on their own increase the number of small households. It accounts for why marital status is also a significant source of observed household income dispersion over time.
An increase in the number of small households increases the proportion of households in the lower income segment of the income distribution, so income dispersion naturally increases. The question is whether this a problem. Is greater household income dispersion a good thing, a bad thing, or a non-thing?
When is Rising Inequality Bad?
We shall consider three situations. The first involves two households that are completely identical in every respect. Over their lifecycles both households make identical choices at every stage, and they are economically equal. There is only one difference between them – every member in one household is two years older than a corresponding member belonging to the other household.
If we now compare their incomes in a particular year, they may look different. In the younger household the wife is a homemaker who will return to work in the following year, but in the older household the wife has already returned to work the year before. Their observed incomes in that particular year would not be the same because they are at different stages of their respective lifecycles. We would then observe economic inequality when in fact it couldn’t be because these two households are by assumption identical.
The implication of this example is that some of the observed dispersion in household incomes is simply noise that exists because households are being observed at different stages of their lifecycle. This kind of income dispersion does not matter. It is a reflection purely of when we observe them (or when we collect the data) and not of any worsening of genuine economic inequality. Over 30% of the measured increase in household income dispersion is due to changes in the age structure of the population. These increases in income dispersion should not be a policy concern.
Second, consider the case of a member in a large household who wants to leave and live separately. If housing were sufficiently cheap, he would do so and a new small household would be formed. So if housing became cheaper, then one would observe an increase in observed household income dispersion. But should one then conclude that things had actually gotten worse?
In this situation, the member who leaves the large household would be better off – his housing consumption would improve and he would enjoy more privacy. The remaining members of the original household may miss his presence, but they now live more spaciously. The measured increase in household income dispersion would therefore not be a symptom of a bad thing. What would be bad would be to have high housing prices that kept people stuck in their original homes and prevented from improving their living conditions. In Hong Kong the rise of sub-divided units in the housing market is the response of a tight housing market to growing demand from individuals wanting to move out of their parents’ home.
The two situations described above are not indicative of a worsening of income dispersion. But the third situation is different. Growing household income dispersion due to rising divorce rates is not a good thing. Divorced families have adverse effects on the development of children as parental investment in children often declines. One of the most alarming facts about divorce is that it is more prevalent among poor families. Rising divorce rates reduce intergenerational mobility especially for poor families. In this situation public policy should aim to reduce the incentive of couples to divorce. This would produce positive effects on reducing income dispersion and enhancing intergenerational mobility.
Is rising income inequality a bad thing? Sometimes, but not always.
Education is the most important determinant of income dispersion among individuals and indirectly among households. Government policy should place human capital investment at the center of its policy strategy to reduce poverty and enhance intergenerational mobility. Rising divorce rates should be given far more important policy attention as a growing source of poverty that impedes intergenerational mobility. Investing in the children of poor and broken families is the best policy to reduce long-term income inequality.
Society should provide additional subsidies and support to students from poor and broken families as an investment in their human capital. Students with ability should be offered scholarships to study with the best schools. The government does not have to fund everything, but it should take a lead to encourage private contributions targeted at this purpose.
