CHAPTER 6 — FUTURE DIRECTIONS


This study has concentrated on producing distributional estimates by using existing data collected and compiled under two different frameworks: household surveys of income, expenditure and wealth; and the ASNA national accounting framework. The production of the statistical building blocks for distributional analysis is only a first step. Distributional analysis to date has addressed largely social policy questions. By linkage to the national accounts, distributional data can be employed also to address macroeconomic policy questions not otherwise possible


IMPROVEMENTS

Outputs

For an economy that occupies a large and diverse economic territory with policy responsibility distributed in a federation, it could be expected that there would be use for distributional data presented by state or smaller geographical areas. Unfortunately the production of such data is hampered by the current lack of suitable data at finer geographic levels.

We note that geographic distribution was not a priority in the OECD work, but that need is largely served by country level detail.


Conceptual Consistency

The study melded data collected and compiled under two different frameworks. Ideally both the household surveys and the national accounts should use the same concepts. In practice this cannot be achieved because of differing requirements of users and indeed the practicality of collecting data about some concepts with a given method. However, there are some ways to improve conceptual convergence of the two datasets:

• Elimination of unnecessary differences. This includes removal of NPISH from the household sector in ASNA; the treatment of unincorporated businesses, their gross mixed income and their property income and expense in the household surveys to enable the derivation of SNA-compliant aggregates; and convergence in the treatment of owner occupied housing in both datasets. The treatment of Social Transfers in Kind to households from governments and NPISH should be made consistent.
• Where conceptual differences cannot be eliminated due to differences in user requirements and measurement practicalities there are some methods that could be used to show how the two datasets differ.

Methods and Sources

In addition to the methodological questions raised in chapter 4, such as the underlying assumptions used to distribute the macro estimates and the models used to distribute the out of scope population from micro surveys, there needs to be some methodological development, including internationally, on the best practice for income and wealth equivalisation.


Developments

The current exercise has been conducted on the basis of available data and within existing frameworks. Discussed above are improvements that could be made towards better coherence between the available datasets. What is not clearly understood, is the utility of distributional metrics within a macroeconomic framework, and therefore what directions and developments there should be to better serve the uses. It is also possible, that in the absence of frequent distributional data, their relevance is not well formed in the user community. This section of the paper makes some suggestions based on academic studies and work done in the USA statistical system.

Academic work

There is a significant body of academic literature devoted to income inequality, see references. Much of this literature concentrates on income shares at the very top end of the distribution (from taxation data), for example the top percentile. Work at this fine level of detail at the extreme of the distribution has, inevitably, come across data limitations in the data sources used, such as “top coding” (the data reaches a cut-off where incomes are reported as “above the cut-off”, and not enumerated, usually for privacy reasons). Some of the literature is therefore about data quality and the analytical techniques or alternative data sources to navigate around these. Unfortunately, this may have resulted in undue concentration on specific datasets and techniques in past studies. Recent work by Burkhauser, Hahn and Wilson suggest some interesting distributional stories may have been missed, for example the distributional impacts of changes to the tax regimes, and changes in economic behaviour.

This has led to suggestions about what national accountants would call alternative concepts of income and their distributions, see Armour, Burkhauser and Larrimore. These range from before and after tax “market income”, before and after cash social benefits, before and after social benefits in kind, and before and after capital gains, possibly in two steps, realised and unrealised. Comparative distributions along this income continuum contain information about the outcomes of the tax and social benefits policies of governments. The four-step comparison presented in chapter 3 in this paper is based on this suggestion. However, we have constrained that analysis to components that are clearly within the 2008 SNA definition of income. The academic work indicates that there may be analytical insights into extending beyond this boundary.

As an example of an extended analysis, a five-step comparison is shown in the following graphs by adding in superannuation benefit payments (referred to in the 2008 SNA as pension benefits) to the distribution. Superannuation benefit payments are withdrawals of accumulated past saving in the 2008 SNA, and not income from current GDP, a conceptually coherent treatment. However, withdrawals of accumulated saving are available for consumption expenditure, and in the case of superannuation beneficiaries will assist explanation of their behaviour.




If appropriate data were available, further steps could be added to show the impact of realised capital gains and capital taxes, and unrealised capital gains. Note that this does not imply redefining 2008 SNA income. Rather it adds to the analysis components of the 2008 SNA non-income flows compiled in the financial, revaluation and other volume changes accounts.

Work by Other National Statistical Agencies

The OECD project and general interest in the subject of distribution has resulted in a number of national statistical agencies undertaking work in this field. For example, the OECD project was based on pioneering work by INSEE in France, see references. Of interest is a theme emerging from the statistical work that is similar to that emerging from academic work: the uses and boundaries of distributional analysis are largely unexplored because the main concentration has been on data production, not analysis. An example of this is the work by Fixler and Johnson in the USA, who ask the question “where are the applications?” As an example of a macroeconomic applications for their dataset, they sketched an analysis of the impact that distribution had on the marginal propensity to consume, an interesting question given various policy initiatives in the USA at that time.


CONCLUSION

The production of the statistical building blocks for distributional analysis is only a first step. Distributional analysis to date has addressed largely social policy questions. By linkage to the national accounts, distributional data can be employed also to address macroeconomic policy questions not otherwise possible. If distributional data were part of the standard economic data outputs, policy analysis could proceed without being hampered by the burden of producing and refining the data.