(This essay was published in Hong Kong Economic Journal on 9 October 2013)
Public interest in income inequality is quite recent even though income dispersions have been rising for many decades. This interest is in part due to increasing sensitivity to normative concerns regarding distributive justice, but it is also a form of political expression as society becomes more open and contentious. The growing gap between rich and poor has been blamed on many different things, involving a lot of strong convictions and political emotions, but there has not been much analysis and evidence. I shall try to fill this gap.
Any sensible public policy to tackle income inequality has to be informed by a correct understanding of what accounts for rising income dispersion and whether this dispersion represents social, economic and political realities. This is to a considerable degree a scientific question and has to be correctly answered before society can address properly the policy issues. In this essay, I present an empirical account of the drivers of income dispersions, identify their different underlying causes, and consider what should be the appropriate policies for addressing them. In particular, I draw attention to an important distinction that should be made between understanding the spread of individual and household income dispersions.
I use a subset of the population census datasets from 1981 to 2011 to estimate Gini-coefficients for household and individual income dispersions. The Gini-coefficient is a popular measure of income dispersion. The coefficient varies between 0, which reflects complete equality and 1, which indicates complete inequality (one person has all the income, all others have none). I should note there is some discrepancy between my estimates and those made by the Census and Statistics Department (C&SD) mainly because my sample sizes are smaller and the income variables are coarsened, but these are I hope minor differences. I hope my study will encourage others to undertake further analysis using more comprehensive and detailed datasets.
Breakdowns of Individual Income Dispersions
Household and individual income dispersions have generally increased over time with a few temporary exceptions (see Table 1). The rate of growth in these dispersions has slowed down somewhat in recent years, but the fact of the matter is that inequality is still continuing to grow. My estimated Gini-coefficients show increases in household income dispersion over the periods 1981-2011, 1991-2011 and 2001-2011 of 0.074, 0.048 and 0.031, respectively. The corresponding estimates for increases in individual income dispersions are 0.090, 0.053 and 0.021, respectively. Over these three decades, the pattern of change in household income dispersion is quite different from that of individual income dispersion. Household income dispersion has grown faster over time, while individual income dispersion has risen more slowly. The difference in the pace of growth between the two is a key issue in understanding the overall increase in income inequality.
Table 1: Gini-coefficients for individual and household income dispersion
| Gini-coefficient by Year | Change in Gini-coefficients (for various Time Periods) | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1971 | 1981 | 1991 | 2001 | 2011 | 1971-2011 | 1981-2011 | 1991-2011 | 2001-2011 | |
| Individual income dispersion | – | 0.397 | 0.434 | 0.466 | 0.487 | – | 0.09 | 0.053 | 0.021 |
| Household income dispersion | – | 0.459 | 0.485 | 0.502 | 0.533 | – | 0.074 | 0.048 | 0.031 |
| Household income dispersion according to C&SD estimates | 0.43 | 0.451 | 0.476 | 0.525 | 0.537 | 0.107 | 0.086 | 0.061 | 0.012 |
Note: C&SD denotes Census and Statistics Department
The income of a household is composed of the individual contributions of different members. It is the product of multiple decisions made within a household regarding many interrelated choices – employment and retirement, marriage and divorce, fertility, living in or moving out, and so on. All these factors contribute in a complex and intricate way to determine household income and its dispersion in society.
To understand why the dispersion of household income has changed over time one must be able to unpack all these intricate relationships and their effects on observed income dispersion. This will be a very challenging if not an impossible task. It will certainly need a richer body of data than what a population census can offer. A simple strategy that relies on the available data is to decompose the dispersion of income into various underlying component factors, to compare how the various components change over time, identify the major sources of income dispersion, and then examine what remains unexplained and why it has changed. This will not reveal the intricate underlying relationships, but it will provide some informative ideas on the important drivers of change in household income dispersion over time.
In performing an empirical decomposition analysis of income dispersion we need to use another tool – the Theil-index, instead of the Gini-coefficient. The latter is not suitable because its formula cannot be mathematically decomposed in a simple manner. The Theil-index can be additively decomposed to yield the share contributions of different factors or groups within the population and it is widely used for this type of analysis. In practice the Theil-index gives estimates that are quite similar to the Gini-coefficient.
According to the Theil-indices, household and individual income dispersions have risen, respectively, from 0.452 to 0.564 and from 0.407 to 0.522 over the period 1976-2011. Household income dispersion therefore increased by slightly less – 0.112 – than individual income dispersion which increased by 0.115 over this period. This pattern is reversed over the period 1991-2011, when the increase in household income dispersion wass 0.077 as against 0.066 for individual income dispersion. The pattern reversals are similar to those found with Gini-coefficient estimates. But the question remains: why is this the case? To understand this phenomenon we need to examine the underlying sources of individual and household income dispersions.
Education and Equality of Opportunity
Individual income dispersion can be decomposed into five factors or groups: age, gender, education, marital status, and recent immigrant status (see Table 2). In the three decades of 1976-2011, age accounted for 3.8% of the change in income dispersion, gender for -12.4%, education for 73.6%, marital status for -0.1%, and recent immigrant status for 11.5%. The corresponding figures for the most recent decade 1991-2011 show that these trends have altered somewhat. Age accounted for 6.2% of the change in income dispersion, gender for -8.4%, education for 65.6%, marital status for 0.2%, and recent immigrant status for 15.2%.
Table 2. Factors of Individual Income Dispersion
| Year | Overall Theil Index | Between group variations | Within group variations | ||||
|---|---|---|---|---|---|---|---|
| Component due to age | Component due to gender | Component due to education | Component due to marital status | Component due to recent immigrant status | Residual Component | ||
| 1976 | 0.407 | 0.026 | 0.027 | 0.1 | 0.006 | 0.005 | 0.243 |
| 1981 | 0.4 | 0.029 | 0.025 | 0.1 | 0.009 | 0.01 | 0.227 |
| 1986 | 0.429 | 0.033 | 0.019 | 0.12 | 0.005 | 0.003 | 0.249 |
| 1991 | 0.456 | 0.027 | 0.018 | 0.141 | 0.006 | 0.008 | 0.256 |
| 1996 | 0.503 | 0.024 | 0.012 | 0.162 | 0.006 | 0.017 | 0.282 |
| 2001 | 0.482 | 0.027 | 0.015 | 0.17 | 0.005 | 0.012 | 0.253 |
| 2006 | 0.482 | 0.032 | 0.011 | 0.143 | 0.005 | 0.012 | 0.279 |
| 2011 | 0.522 | 0.031 | 0.012 | 0.185 | 0.006 | 0.018 | 0.27 |
| 1976-2011 | 0.115 | 0.004 | -0.014 | 0.084 | 0 | 0.013 | 0.027 |
| 3.80% | -12.40% | 73.60% | -0.10% | 11.50% | 23.60% | ||
| 1991-2011 | 0.066 | 0.004 | -0.006 | 0.043 | 0 | 0.01 | 0.014 |
| 6.20% | -8.40% | 65.60% | 0.20% | 15.20% | 21.20% | ||
| 2001-2011 | 0.04 | 0.004 | -0.002 | 0.014 | 0.001 | 0.006 | 0.017 |
| 10.30% | -5.60% | 36.30% | 2.30% | 14.60% | 42.20% | ||
Note: Only individuals aged 18 or over who were not students are included. Domestic helpers excluded.
Nonetheless, it is evident that by far the single most important driver of individual income dispersion is education, and that education has been the dominant factor over three decades. This is not surprising because income dispersion among working individuals largely reflects variations in individual productivity, which is primarily captured by education attainment. But education’s effect on individual income dispersion has also diminished in the past decade to a considerable extent. In two separate published essays I have argued that the failure to adopt a sound population policy and expand educational opportunities, especially post-secondary education, in the 1990s has adversely impacted both income inequality and economic productivity growth (see HKEJ 18 April 2012 and 2 October 2013).
Population policy and education are important forms of human capital investment and I have found it to be a critical reason why economic productivity in Hong Kong has been higher than Singapore for 50 years. However, that gap has narrowed considerably in the past decade. Given education’s important historical role in driving individual income dispersion, education policy has to be the primary policy tool for tackling individual income inequality.
Irrelevance of Other Factors in Income Dispersion
The effects of other factors on individual income dispersion are relatively minor compared to that of education. The contribution of age is negative and reflects the effect of large numbers of baby-boomers getting older over time. The share of high-income baby boomers in the working population increased during the three decades studied and this helped to lower income dispersion. But this effect is ending as the cohort retires and exits the workforce.
The contribution of gender to individual income dispersion is negative and falling over time. Women have become more educated and therefore have formed a growing part of the working population. This has contributed to lowering income dispersion within the working population as income differences between men and women have narrowed over time.
The contribution of marital status to individual income dispersion is minimal over time. This probably reflects the offsetting effects of a rising share of lower income single and divorced individuals in the workforce and a declining share of higher income married workers, but the precise nature of these effects have not been fully investigated.
The contribution of recent immigrants to income dispersion has been positive. This reflects the fact that their numbers have been rising over time and they tend to have a higher proportion of lower productivity individuals. Immigration policy is of course another form of human capital investment policy. In a free society the decision to emigrate is an individual right over which government should not exercise control. But in Hong Kong whom we allow to emigrate here is a matter of government policy. Family reunion cases are also related to human rights, although what constitutes family is subject to policy delineation.
To sum up, then, the increase in individual income dispersion is foremost affected by education. Other factors contribute, too, but none as significantly as education. Education policy therefore must be at the centre of efforts to tackle the rising inequality in individual incomes.
As for the dispersion of household income, I shall explore that topic in next week’s essay.
