«Corrigenda to OECD publications may be found on line at: © OECD 2014 You can copy, download or print OECD content for ...»
The relevant dimensions of Inclusive Growth may vary across countries depending on their level of development, social preferences, specific conditions and circumstances. In particular, other dimensions may be relevant for developing countries, such as social connections, civic engagement and the quality of institutions, including: personal and professional networks, citizens’ influence on collective decisions, and trust in public institutions. Economic growth may affect social connections by shaping people’s opportunities to meet and socialise in workplaces (e.g. through hours worked, atypical workers, etc.) but also in the broader community, (e.g. through migration, longer commuting times). Weak economic growth can bring a decline in trust in public institutions. Another dimension to consider is personal security, including safety in conflict and post-conflict situations, violence against women and children, and the problems associated with criminality, which impact on economic growth and human development. Infrastructure and housing, including access to decent homes, sanitation, safe water, transport and other basic infrastructure could also be relevant to emerging economies and developing countries. This is also the case for social protection, including the availability of and access to services, such as social security, assistance programmes, and unemployment insurance, which tend to be more developed in more advanced societies.
How can distributional aspects of Inclusive Growth be measured?
The second key pillar of Inclusive Growth – emphasis on distribution – requires an innovative method for comparing multidimensional well-being outcomes across the population.
This is a complex undertaking, but can be approached by identifying a “representative household” and aggregate monetary and non-monetary outcomes in a single, multidimensional, measure of “overall
living standards” that can be used to assess policy alternatives (Boarini et al., 2014). In particular:
The “representative household”. In conventional analysis of the effects of structural policies on economic growth, emphasis is placed on an “average” individual or household, and outcomes are described as the mean values of conventional indicators. This is the case of empirical evidence of how specific policies, for example, in the area of innovation, education, environmental regulation, affect per capita (mean) income, which has been the approach taken in Going for Growth. However, as shown in Chapter 1, well-being outcomes are very unequally distributed, which calls for gauging the effects of policies not only on the mean, but also at other points of the distribution. One such point is the median, representing the value that cuts the distribution into two equal parts. Thus, the income of a median household or person is much more representative for the income of the population at large or of the “middle class” than a simple average. OECD work is increasingly shedding light on the policy impacts on different segments of the distribution. In many developing and emerging countries, it may also be important to focus on additional segments of the distribution of income, such as on those individuals with income at the poverty line.
The notion of “overall living standards”. In a multidimensional concept of Inclusive Growth, different dimensions must be measureable in a comparable way. One way of doing this is to use the principle of equivalent income, which is defined as the hypothetical income that would make an individual indifferent between their current situation and a benchmark situation (typically, the best possible outcome in non-income dimensions). The contribution – positive or negative – of the non-income components to multidimensional living standards is measured by the monetary value that households are willing to pay to enjoy, or get rid of, those components. The empirical computation of equivalent income is complex, and important challenges need to be addressed, but conventional techniques can be used, as in the illustration below.
The emphasis placed by Inclusive Growth on different points of the distribution of outcomes makes it flexible for policy analysis. For example, Inclusive Growth could be used to assess the effects of pro-growth policies on different income groups. In contrast to conventional propoor growth studies (e.g., OECD, 2006), where the analysis focuses on the incomes of a particular low-income group, the approach detailed below identifies if growth has been especially favourable to particular income groups, and it can be extended to cover non-income dimensions. This approach relies on the use of the Generalised Mean Framework (Boarini et al., 2014) (Box 2.5).
Box 2.5. Using Atkinson's generalised means to aggregate welfare across individuals General means are grounded in Atkinson’s (1970) framework for inequality and welfare analysis and belong
to the family of “equally distributed welfare” functions. Formally, general means are defined as follows:
where the vector w=(w.1,…w n) measures the welfare distribution, w.i0 is the welfare of the i-th person, and n is the population size.
The general mean reduces to the standard mean when τ =0 and to the geometric mean when τ =1. The general mean of individual welfare places greater weight on higher welfare individuals and less weight on lower welfare individuals as the parameter rises. Hence τ is sometimes interpreted as a measure of the level of inequality aversion.
In the simple case where individual welfare is defined in income alone, the general mean is called “income standards” (Foster and Szekely, 2008). An ongoing OECD project uses income standards to look at the evolution of income growth across the whole distribution (Causa et al., 2014 forthcoming). This approach, which builds on Foster et al. (2013) and is also used by the World Bank for tracking income inclusiveness has been applied to all OECD countries from mid-1990s to the end of 2000s.
Figure 2.3 for Belgium and Finland, two countries that experience opposite trends in growth of income standards.
In Belgium, between 1995 and 2009, incomes grew faster among households in the lower half of the distribution and particularly so among the poorest. By contrast, Finland recorded a marked increase in income growth among households in the upper half of the income distribution.
Income growth has benefitted different social groups: the cases of Belgium and Finland
Source: OECD calculations based on Causa et al. 2014.
How can a multidimensional approach to Inclusive Growth become policy-actionable?
The third key pillar of Inclusive Growth – policy relevance – calls for careful “mapping” of policies to multidimensional outcomes. As noted above, conventional economic growth theory focuses essentially on average material living standards, whereas Inclusive Growth looks at the distributional effects of policies along the entire distribution of outcomes. The task ahead is therefore more complex and requires a better understanding of the causal links between policies and outcomes in various areas. For instance, extensive work has been undertaken to explain the policy determinants of growth of GDP per capita and of labour market participation. However, much less is known about policy levers of household income specifically or of health status. Indeed, conventional analysis looks at the effects of policies on selected outcomes separately, whereas what is proposed here is a joint assessment of the impact of policies on a set of income and non-income outcomes that matter for wellbeing. For instance, tax and benefits policy has a direct influence on household disposable income but has also a complex impact on unemployment and an indirect influence on health status (if for instance taxation finances health expenditure). These aspects are further discussed at the end of this chapter.
2.4. An illustration of the Inclusive Growth Framework in practice A three-dimensional measure of Inclusive Growth A simple exercise can be used to illustrate the concept of Inclusive Growth. This illustration uses three dimensions: unemployment and health status as the non-income dimensions to be considered, along with household income (Box 2.6).24 In essence, inclusiveness is captured by relating to three dimensions of well-being (income, jobs, health) and by taking into account distributions of outcomes along these dimensions and across different population groups. Attention is placed on OECD countries and over the period 1995-2011, distinguishing before (1995-2007) and after the crisis (20017-2011 or last year available).
There are three steps to take when measuring multidimensional living standards at an aggregate level:
-- measuring income-related living standards (captured for instance by consumption or real income) at the individual level. The approach here uses household real disposable income as the relevant measure, although from a conceptual viewpoint, a measure of household real net adjusted disposable income would be preferable (see Stiglitz et al., 2009).; it is then necessary to bring the chosen non-material dimensions, i.e. unemployment and health, into the picture and to measure these dimensions at the individual level in order to combine them with measured material dimensions;
-- expressing the non-income dimensions in a monetary metrics, using estimated shadow prices of nonincome dimensions and
-- computing the broader living standard measure for the ‘representative’ household.
There are specific reasons for choosing unemployment and health status in this illustration.
The unemployment rate is a strong determinant of subjective well-being. Unemployment is also the variable that has repeatedly been used in the literature on the measurement of multidimensional living standards and well-being (Fleurbaey and Gaulier, 2009). At the same time, not all types of unemployment are equally relevant from the perspective of well-being. For instance, long-term unemployment and weak prospects of returning to work following a lay-off seem to be more detrimental than short spells of unemployment between jobs. The average rate of unemployment, which is used in this analysis, cannot distinguish between these features of the labour market.
However, analysis that has been carried out to consider alternative formulations (e.g., long term versus short term unemployment; unemployment turnover; employment) and their impact on the measure of Inclusive Growth shows that, overall, results are robust for these various alternative formulations.
Life expectancy, the measurement variable for health, is a standard gauge of longevity and is one of the best available measures of health status. Life expectancy, however, does not measure healthiness or quality of life (OECD 2011b). In many studies the proxy used for healthiness is often a “self-reported health” variable, although this variable is subject to some measurement errors and is available for most countries only since the mid-2000s. Despite weak variability among countries, the number of years necessary to gain one extra year of life expectancy has varied significantly among OECD countries during 1995-2009, including among high-income OECD countries.25 OECD work has already documented links between life expectancy and environmental and life-style variables, as discussed below (Joumard et al. 2010).
Mortality measures have the twin advantage of being widely available for large sets of countries and of having long time series. They are very well documented and available by age, gender and, in some countries, by educational attainment (Sen, 1998; Mackenbach et al., 2008). Also, there are large and persistent inequalities in longevity within countries that tend to be correlated with the socio-economic background of individuals. Furthermore, the socio-economic determinants of inequality in longevity, such as the education gradient of mortality, are very different across OECD countries. The implication is that life expectancy is expected to play a significant role as determinant of multidimensional inequality and as a driver of cross-country differences in the level and evolution of multidimensional living standards. That said, it will be of interest to examine how simple measures of mortality compare with measures of morbidity or quality-adjusted measures of mortality once such data become available for a larger number of countries and time periods.
Two main findings: an illustration of multidimensional living standards
Multidimensional living standards can be compared across countries as a function of household disposable income. The representative household is the median household, and equivalent income for health reflects the monetised value of differences in outcomes relative to the sample’s reference country with highest life expectancy: Japan. For the jobs dimension, absence of unemployment has been taken as the reference value. Equivalent income then presents the loss in multidimensional living standards that a representative household in a particular country suffers by experiencing unemployment, an unequal distribution of household income, and shorter longevity than the reference country. On average, the total welfare loss associated with the three components represents as much as 45% of disposable income, with almost equal contributions of income inequality, health and unemployment (Figure 2.4).26 If the representative household is set not as the median household but as a household whose income corresponds to the bottom quartile of the income distribution, the estimated welfare loss would amount to as much as 63% of disposable income (with the loss due to income inequality amounting to 33% of disposable income).
Note: Aversion to income inequality is chosen so that living standards reflect the median household income.
Source: OECD calculations based on OECD Annual National Accounts, OECD Income Distribution Database and OECD Health Data Base.
Growth in multidimensional living standards and in GDP per capita may differ considerably, highlighting the relevance of multidimensionality for policy analysis on Inclusive Growth. Focussing on a set of 18 OECD countries between 1995 and 2007, and taking median equivalent income as the reference for the analysis, we observe that all measures of multidimensional living standards show improvements over the period of analysis and would thus point to persistent Inclusive Growth based on our definition. This has to be put in perspective, however. The growth rate of multidimensional living standards for the median household is almost certainly biased upward, because our current measure of inequality only reflects inequality of disposable income, not inequality of equivalent income. Unemployment and life expectancy are distributed unequally across individuals (see for instance OECD, 2011b) and enhance the (mostly negative) effects of increasing inequality.
This would in turn reduce the measured change in multidimensional living standards for the median household. Developing the data needed to capture inequality in jobs and health for all countries is therefore important in future work.