When brokering knowlewdge on health equity in the public agenda relative inequalities are more appealing in the story telling; but when one moves to assessing the impact of actions that public health and politics should be accountable for, then absolute inequalities are the more appropriate measure; in fact, as underlined by Johan, relative inequalities are too sensitive to the decreasing trend of the background risk just for arithmetic reasons however in the same assessment we have to take into account that relative measures maintain a crucial role in evalutating the potential modification of effect of the action in the intensity of health disadvantage moreover one should alco consider the different mechanism of generation of health inequalities we are trying to assess; in this case the question could be simplified as follow: 

• given that health inequalities come from two fundamental mechanisms a) inequal exposures to risk factors and b) inequal vulnerability to health effect of the same factor
• and that the actions tackling health inequalities belong to two fundamentalctegories a) allocating interventions proportionally to the distribution of exposure to risk factors and b) modifying the design of action in order to make them equallly effective for more vulnerable people
• then the evaluation of impact should look at absolute measure in the case of a) and to both relative and absolute measures in case.

The debate on the relevance of relative vs absolute health inequalities has a parallel in economics, and this latter debate may be useful in understanding the contrast between absolute and relative health inequalities. For many decades, poverty experts have maintained a discussion on whether we should measure the impact of policies based on a measure of ‘absolute  poverty’, which uses an absolute threshold or a fixed standard of what households or individuals should have in order to meet their basic needs, e.g., living on less than US$1.90 dollars a day; or a measure of ‘relative poverty’, which uses a cut-off point in relation to the overall distribution of income or consumption in a country at a given point in time.

Most countries measure poverty using an absolute threshold. This threshold or fixed standard, however, is not common for all countries, but it is individually determined by each country’s Government, and it is accepted by international agencies such as the World Bank as long as the threshold and estimation can be backed by sound methodologies. Countries draw a line based on this threshold and publish, for example, the national poverty headcount ratio, i.e., the percentage of of the population living below the national absolute poverty lines.

The parallel between measures of health inequalities vs poverty is not perfect, if only because the former is based on a measure of association between two variables (e.g., education and health), while the latter is established based on the distribution of a single variable, e.g., the fraction of households below a certain threshold of income. Nevertheless, the parallel is useful because it represents a common theoretical debate on whether we should measure progress based on an absolute standard (e.g., $US1.90 a day) or whether measures should assess poverty based on the living standards of individuals relative to the overall levels of income in a given population, e.g., median income at a given point.

Given that most countries use absolute poverty measures, one would be tempted to conclude that this must be the right thing to do. A few countries, however, have chosen to use measures of relative poverty. Interestingly, setting relative poverty lines is exceptionally prevalent in Europe, particularly in recent years, as many of the middle-income countries in the region would find it difficult to define a common set of good and services that would represent an absolute poverty line and that would be perceived as valid by an international audience. Eurostat, for example, publishes a measure of poverty based on a relative threshold, i.e., those living below 60% of the national median equivalised disposable income after social transfers.  This would seem to suggest that while absolute poverty measures may be particularly useful for low- and middle-income countries, where a large share of the population does not have access to what is commonly accepted as basic living standards, this concept loses value in populations where the majority or all citizens have access to this living standard, and so the definition of poverty becomes more ‘relative’.

This suggests that despite the indisputable value of absolute measures of inequality, there is a particular value in using relative measures. Although the latter does not necessarily imply a low standard of living, it measures low income in comparison to other residents in a country, whatever that might be. In other words, relative measures would view health as an outcome dependent on the social context, rather than as an absolute that may mean different things across different countries or periods. Because an ‘absolute standard’ is likely to change from time to time, as well as from country to country, it would seem almost impossible to establish a level of absolute inequality that would be useful or valid when comparing countries or periods. Relative measures of inequality, on the other hand, are non-dimensional, and gives us a sense of the the health of a socially disadvantage group relative to that for the rest of the population; thus it is conditional on the levels of overall health in that population, at a particular point in time.  

The limits of overly focusing on absolute measures of inequality can be best illustrated with an extreme example similar to the one posed by Mackenbach: Suppose we have two populations (countries): Population A has a rate of mortality for disease X of 100/1,000, while population B has a rate of the same disease X of 20/1,000). Suppose that both populations A and B have an absolute inequality (e.g., difference between primary vs tertiary educated) of 10. Can we conclude from this that population A is as unequal as population B? While there is some more information needed to assess this, there are clearly some problems in defending this: 10 deaths in a population with an underlying rate of 100/1,000 is unlikely to mean the same as 10 deaths in a population with an underlying rate of 20/1,000. The same comparison can be made for comparisons over time within the same country: A rate difference is so dependent on the underlying rate of disease at a given period that we cannot compare the level of absolute inequality of England in 1,800 with that in 2,010 and say that we are doing very well in terms of inequality just because the latter number is much smaller (we may be doing very well in terms of overall health improvements, but that is a different question). In other words: absolute measures of inequality are even more dependent on the ‘underlying’ mortality or morbidity rates than relative measures, the main argument against the latter.

The concept of inequality has thus the necessity of a comparison to whatever the level of human capital is for other members of the population at that point in time, and as such a measure of relative inequality seems essential, even if some of its ‘mathematical properties’ might be deemed undesirable. That is not to say that absolute levels of mortality are irrelevant: In fact, Governments and international agencies constantly monitor absolute levels of mortality and health in their statistics. The added value of a measure of inequality, however, is to provide additional information of the distribution of this mortality rate in the population, whatever that might be; rather than telling us something about the underlying level of mortality, something we can get from overall health statistics.

In conclusion, while absolute inequalities may be useful in some ways, relative measures would seem at least equally important because they are non-dimensional and possibly more useful for comparing populations as well as time periods, particularly when underlying rates change dramatically. Perhaps we should encourage policy makers to interpret relative inequalities side-by-side total mortality or morbidity rates, but that is a different strategy than using absolute measures of health inequalities as our gold standard to measure progress in health inequality alone.