In sum, the approach to valuing the benefits of public education and health care services amounts to asking the following question: How much would the income of a household have to be increased if it had to pay for the free or subsidized public service (or the insurance value in the cases in which this applies to health care benefits) at the full cost to the government? The conventional answer to this question is to look at production costs. This approach, however, raises a number of issues: it does not take into account variations in needs across income or age groups, nor does it consider service quality, and may not reflect the actual valuation by beneficiaries.37 Teachers may not show up at local schools, and the quality of the schooling services provided may be a fraction of what households would deem as adequate given the amount of taxes that they have paid. Distributional analysis of in-kind transfers may reveal that poorer households gain larger shares of particular categories of public spending than higher-income households. However, this result may owe to the fact that the middle classes and the rich opt out of public education and health care because of their poor quality. Given the limitations of available data, however, the cost-of-provision method is the best one can do for now.38
Consumption Taxes and Subsidies
A second element that is typically excluded from assessments of people’s consumption possibilities is the impact that consumption taxes and production subsidies have on what—following the naming convention established in the Commitment to Equity project—we can call “consumable income,” i.e., the actual consumption of goods and services by people.39 To illustrate this point, let’s consider two countries (or the same country but at two points in time) and assume that final consumption expenditures are the same in the two cases, but that in one the value added tax (VAT) is 10% while in the other it equals 20%. Obviously, for a given amount of money income, what households can actually consume will be higher in the first case than in the second.
Consumption taxes can increase poverty. In a sample of 28 low- and middle-income countries, the Commitment to Equity project found that for Armenia, Bolivia, Ethiopia, Ghana, Guatemala, Honduras, Nicaragua, Sri Lanka, and Tanzania, the headcount poverty for consumable income, based on a poverty line of $2.50 per day (in 2005 PPP), is higher than the headcount for market income (before personal income and consumption taxes, cash transfers, and consumption subsidies), i.e., consumption taxes increase the prevalence of income poverty. In Ghana, Nicaragua, and Tanzania, net payers to the fiscal system begin in the income range $0–1.25/day in purchasing power parity (i.e., the ultra-poor) when consumption taxes are included. In Guatemala, Ethiopia, and Armenia, net tax payers begin in the income group of extreme poor with $1.25–2.50/day. In Sri Lanka, Peru, El Salvador, Dominican Republic, Honduras, and Bolivia, net payers to the fiscal system begin in the income category $2.50–4/day, i.e., in the group classified as moderately poor.40
Beyond these effects on the absolute level of consumption, consumption taxes may also impact on distribution. When annual income is used as a measure of economic well-being, consumption taxes are regressive, i.e., relatively more of them are paid by low-income groups of the population, as poor people spend a greater share of their income on consumption than rich people. Conversely, when lifetime income is used as a metric of economic well-being, consumption taxes could be proportional (or even progressive), under the assumption that today’s savings will be spent on consumption goods in the future. Even in a life-course perspective, however, consumption taxes may have regressive effects when considering that accumulated savings may be used to finance the future purchase of capital goods (e.g., housing) rather than consumption goods; that this purchase may be effected abroad rather than domestically; that different consumption goods may be subject to different levies; and that the structure of consumption may differ across the income distribution. In all these cases, consumption taxes will have redistributive effects that are generally ignored by studies of fiscal redistribution, in addition to those operating through the general level of prices.
While it is acknowledged that household consumption possibilities are reduced and increased by, respectively, consumption taxes and production subsidies passed on to the prices households pay for goods and services, taking this impact into account has not been part of the conventions typically used for analyzing disparities in households’ economic well-being.41
Conclusions
Since the turn of the 21st century, both policy-makers and the public at large have paid growing attention to the distribution of household economic resources. This has been accompanied by a growing number of micro-data sets becoming available in individual countries (notably on wealth), a growing focus on the top end of the distributions, the mobilization of additional data sources such as tax records, steps to bring closer together macro- and micro-data streams, and a growing attention to the “global distribution of income,” which has led to the construction of large international data sets combining information from different national sources. These developments have changed significantly the landscape since as recently as 2009, when the report of the Stiglitz-Sen-Fitoussi Commission was published. In particular, returning to the use of tax data and, especially, combining them with data from (income and wealth) surveys and from national accounts has generated a number of seminal contributions, and helped focus attention on top incomes in an unprecedented manner.42
While there has been progress, major issues remain in achieving the goal of measuring the distribution of household economic resources across countries and over time. Different international data sets feature important discrepancies in terms of both levels and changes of inequality for the same country and time period; inconsistent narratives on inequality levels and trends among micro- and macro-sources are notable and, in some cases, have become larger over time; inequality indicators tend to reflect only partially the true extent of inequality due to the underreporting and noncoverage of rich individuals in household surveys; measuring the correct income concept is still challenging; international conventions remain incomplete; and data on wealth inequality, while more common than before, still remain scarce.
In this context, a number of recommendations are put forward:
• Defining and measuring a more comprehensive income concept. As discussed above, more analytical and empirical work is needed to accurately reflect in a broader income concept the value of in-kind benefits such as education and health care services provided to households by governments and nonprofit institutions. In addition, the measurement of consumption possibilities must consider the impact of the services that households produce for their own consumption as well as the impact of consumption taxes and subsidies. The international convention proposed by the Canberra Group Handbook acknowledges the need to broaden the conceptual definition of household income to consider benefits in-kind, but remains silent on how to achieve this in practice, while excluding both household services produced for own use and consumption taxes and subsidies. This needs to change, in ways that do not compromise the quality and comparability of existing measures of other income streams. This could be achieved by complementing existing measures of household disposable income (which largely follow international guidelines) with experimental measures based on broader concepts (e.g., measures that integrate the value of benefits in kind, services produced by households for their own use, and consumption taxes). Clearly, consumption possibilities are different depending on, for example, VAT rates: two individuals with the same disposable income (or adjusted final income, for that matter) but with different structure of their consumption expenditure would have different consumption possibilities when the VAT rates applied to goods and services differ.
• Correcting for under-reporting and noncoverage of the rich. Assessing the extent to which there is under-reporting at the top (and bottom) end of the distribution and whether rich (and poor) people are “missing” from income, consumption, and wealth distributions should be a common practice in the measurement of economic i
nequality. “Rich lists” (reporting the number and the income/wealth values of very wealthy individuals and households) exist for many countries, and tax records (when of good quality) provide an important resource for implementing that correction. Proposals for adjustments, where appropriate at the national level, for underrepresentation and noncoverage by surveys should be made. All of this will require considerable investment in improving and developing statistics. Of prime importance is for governments to make the information from (anonymized) tax records available and allow for the linking through personal identification numbers between surveys and registries.43 The scholarly community working on inequality should undertake a thorough and systematic assessment of the various methods to contend with under-reporting and noncoverage, and come up with recommendations of best practices, including some key robustness checks.
• Increasing the availability of data on the distribution of wealth. There are a series of sources to obtain information on the distribution of wealth: dedicated household surveys on wealth; administrative data on investment income, capitalized to yield estimates of the underlying wealth; lists of large wealth-holders, such as the annual Forbes Richest People in America List, or the Sunday Times Rich List for the UK; population censuses, which in some cases and years included questions on household wealth; administrative data on individual estates at death, multiplied-up to yield estimates of the wealth of the living; and administrative data on the wealth of the living derived from annual wealth taxes. Greater international efforts should be devoted to assess the availability and quality of data on wealth distribution and make recommendations so that the necessary data are periodically collected in as many countries as possible, and in ways that make the information comparable across countries and over time.
• Addressing inconsistencies in international data sets. Growing interest in the “global distribution” of income or wealth (i.e., the distribution that would obtain when all people of the world are considered as citizens of the same country) has recently led to the proliferation of international data sets combining information from a large array of national sources. While the quality of these data sets is generally a function of the underlying national data, the agencies and researchers initiating these data sets often make various assumptions to fill data gaps or to increase the ex post comparability of these estimates. Even when these international data sets are limited to parts of the world where country-level data are more readily available, different data treatments applied to national data, and differences in data collections (across countries and over time), may not be visible to users. Given that global inequality analyses are so sensitive to the choice of database, data set users should acquire a thorough understanding of the assumptions and methodological choices embodied in the data they are about to use, and undertake systematic robustness checks to determine if their results are sensitive to the use of a particular data set. Data set producers should document all assumptions clearly and thoroughly; make the data, programs, and results publicly available to allow for replicability whenever it applies; compare their methods and results with one another; and, eventually, agree on conventions and best practice when calculating inequality indicators from micro-data, secondary, and imputation-based sources. Finally, the international community should devote greater financial resources to allow poorer countries to put in place the statistical infrastructure that is needed to fill the gaps and provide the information needed to gain a better understanding of national and global inequalities. Providing a better picture of the global income distribution is a global public good (needed, for example, to assess the impact of globalization on people in all countries of the world), and the onus is on rich countries to provide part of the necessary resources for this to happen.
• In line with one of the main recommendations of the Monitoring Global Poverty report, an international organization should take the lead in setting up a standing Statistical Working Group on economic inequality, with a remit to set guidelines for the measurement of household income, consumption, and wealth; to examine the relation between the three; to investigate the relation between household survey, national accounts, tax records, and other data sources; and to make proposals on how consistency among them can be enhanced. The latter would be important to address the issue of sometimes inconsistent narratives among sources on inequality levels and trends.
• To integrate or not to integrate? Undoubtedly, the life of users of economic inequality data would be made much easier by the existence of one integrated data source on the distribution of household income, consumption, and wealth, compiled from various sources: household surveys, administrative registries, statements provided by financial institutions, and national accounts. However, we are still far away from this ideal: individual data sources are compiled with different goals, according to different conventions and definitions. The assumptions made by national accounts statisticians when integrating counterpart information from various institutional sectors may be less palatable to survey statisticians. While several initiatives are currently underway in developed countries (both to integrate micro- and macro-statistics for the household sector, and to integrate various types of micro-statistics), in low- and middle-income countries the questions about the quality of data makes integration exceedingly difficult. When survey income aggregates are between 40% and 60% of national accounts aggregates, for instance, one wonders whether the problem is really the existence of under-reporting and noncoverage in the surveys or rather with the accuracy of national accounts. In such a context, there is considerable value in a multi-source approach to investigate the distribution of income, consumption, and wealth. No single method is sufficient on its own, and it is necessary to draw attention to their strengths and weaknesses (Alvaredo, Atkinson, and Morelli, 2016). In these situations, rather than choosing one alternative, one should probably pursue both the integrated data approach as well as the dashboard approach (Bourguignon, 2016). The dashboard approach would entail reporting estimates from household surveys and tax data (and possibly other distribution data) separately as they describe different segments of the distribution; integrating both through, for example, the DINA (Distributional National Accounts) methodology described in Chapter 6, as well as other methods described above; and using national accounts and administrative data to investigate sources of inconsistency and to assess their implications for inequality results.
Addressing all these issues will require more investment of resources (both financial and intellectual) on the part of governments, statistical offices, multilateral organizations, philanthropic foundations, and researchers alike. It will also require cooperation among these constituencies to generate international conventions where they are lacking, and implementation guidelines where needed. Finally, accurate measurement of economic inequality will require a political commitment. Governments, international organizations, and the scholarly community need to be committed to transparency and to make information publicly available in ways that facilitate the measurement and analysis of economic inequality while protecting the identity of respondents to preserve confidentiality.
One final word: While the discussion here has emphasized the shortcomings, problems, and limitations of existing statistics on economic inequality, we have adopted in this chapter the same view underpinning the Report of the Commission on Global Poverty (Atkinson, 2016). We should be aware of the uncertainty that surrounds inequality indicators, and be conscious that both levels and changes in inequality are measured with a considerable margin of error. Different sources are, however, affected by different problems and biases, and by crossing different perspectives and information sources we can get a better and richer understanding of the underlying reality. Hence, rather than taking the position that nothing can be said, we want to encourage the research and statistical communities to identify different potential sources of error, to develop methodologies to address these problems, and to attach an indication of their possible size, as well as propose
ways to introduce more robustness in measuring such a crucial indicator as the extent of economic inequality and how it changes over time (Atkinson [2016], p. 15).
Notes
1. See, for example, Alvaredo and Gasparini (2015), Anand and Segal (2015), Atkinson (2015), Bourguignon (2015a), Bourguignon and Morrisson (2002), Bourguignon, Ferreira, and Lustig (2005), Cornia (2014), Deaton (2013), Ferreira et al. (2012), Ferreira et al. (2016), Lopez-Calva and Lustig (2010), Milanovic (2016), Piketty (2014), and Stiglitz (2012). See also Klasen et al. (2018) and other chapters of the report by the International Panel on Social Progress (2018).
For Good Measure Page 11