Measuring Inequality of Opportunity in more periods than ever - PowerPoint PPT Presentation

Measuring Inequality of Opportunity in more periods than ever before: the capital income approach Hugo del Valle-Incl´ an Cruces RGEA, ECOBAS LIS research visit, January 2019 H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 1 / 48

We are going to talk about... What is Inequality of Opportunity? 1 The Ferreira and Gignoux’s method The outcome of choice The limitation of relying on parental education IOPK: the capital income approach 2 Capital income as a proxy of family background Validation of IOPK Taking advantage of IOPK Robustness and future analyses H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 2 / 48

H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 3 / 48

What is inequality of Opportunity? The Inequality of Opportunity approach considers that any outcome (such as income, wealth, scientific achievement...) is a function of circumstances and efforts: I = Φ( C, E ) - Circumstances are factors that might influence a given outcome but cannot be chosen by individuals. These include gender, race, geographical origin, family background... - Effort is the intensity with which individuals devote themselves to work, and can, on the contrary, be decided by them Thus, the field of Inequality of Opportunity distinguishes between inequalities that are due to personal responsibility —for which individuals can be held responsible— and may therefore be considered ethically acceptable ; and those that are not —for which individuals cannot be blamed—, which are therefore deemed unjust . H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 4 / 48

The Ferreira and Gignoux (2011) method We employ the parametric procedure proposed by Ferreira and Gignoux (2011). It consists on regressing an outcome y i on a set of circumstances C i ln y i = ψC i + u i Actually, that is a reduced form of ln y i = αC i + βE i + e i where E i is the influence of circumstances on the level of effort. Then, to “remove” the effect of effort E i on the outcome, we predict its values ˆ y i (based only in the circumstances set C i ) and apply an inequality measure to these fitted values, usually the Gini index or the GE(0) (Mean Log Deviation). Finally, we get an absolute measure of IOP IOP abs = I ( { µ k i } ) and a relative one IOP rel = I ( { µ k i } ) I ( y ) H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 5 / 48

The outcome of choice The outcome considered, y i , can be wealth, health status, educational attainment... but income is the one most frequently studied. Yet, once we have decided to study income, we must choose between different concepts. For example, should we consider personal income or household income? In the study of income inequality the standard practice is to consider household income, but should it also be the case in the study of opportunity inequality? H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 6 / 48

The outcome of choice Shapley value decomposition of the role of circumstances. Source: LIS and author’s calculations. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 7 / 48

The outcome of choice Shapley value decomposition of the role of circumstances. Source: EU-SILC and author’s calculations. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 8 / 48

The limitation of relying on parental education In the EU-SILC database we have information on parental background in 2 of the 13 waves, those corresponding to 2004 and 2010, which is around a 15%. In the LIS database we have this information in 63 of the 344 datasets, approximately an 18%. Implications of this limitation are, for example, that we do not know how the level of IOP is nowadays or how has evolved in most countries, or if its evolution is linked to income or wealth inequality, or any other economic or social phenomena, at a cross-country level. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 9 / 48

How can we overcome the limitation of relying on parental education? We propose a new empirical strategy that is not limited by the scarcity of data on parental education. It consists of using another variable, widely available, as a proxy of socioeconomic origin. Capital income of households may be a good proxy of socioeconomic origin because positive returns on investment are linked to financial literacy, cognitive abilities and education ( Gaudecker2013 ). Also, stock market participation increases with income and education (Mankiw and Zeldes, 1991) and it is influenced by behavioral biases and cognitive abilities (Andersen and Nielsen, 2010). Furthermore, individuals with high financial literacy and high cognitive abilities face lower cost of acquiring information and thus lower participation cost than individuals who know little about financial markets and have low cognitive abilities (Grinblatt et al., 2010). We use data on the capital income of households to construct a discrete variable that classifies individuals into levels according to the importance of capital income in respect to their household’s total income. This variable is built such that its Cumulative Distribution Function tries to resemble that of the variable parental education. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 10 / 48

Grouping individuals according to the relative importance of their households’ capital income First, we look at the distribution of parental education in the periods and countries for which we have this information. We have previously constructed a variable of parental education with three levels (primary or less, secondary, and tertiary or more), and hence we can see how the education of parents is distributed across individuals. We look at the values of its CDF, and store them Second, we generate a variable on the relative importance of households’ capital income in respect to their total income (a simple ratio) for every individual in the sample. This is: Kinc i = Capital income of individual i ’s household Total income of individual i ’s household Third, we find the exact values of this variable at which its CDF takes the same values of the parental education’s CDF Finally, we create a discrete variable with three groups of a size determined by the parental education’s CFD H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 11 / 48

Grouping individuals according to the relative importance of their households’ capital income: an example A fictitious distribution of parental education Frequency Percent Cumulative Parental education Primary or less 6,000 60 60 Secondary 3,000 30 90 Tertiary or more 1,000 10 100 Total 10,000 100 The highest education level attained by any of the parents is considered. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 12 / 48

Grouping individuals according to the relative importance of their households’ capital income: an example A fictitious distribution of levels of capital income relative importance K income to total income Frequency Percent Cumulative Kinc i ≤ 0 . 01 6,000 60 60 0 . 01 < Kinc i ≤ 0 . 03 3,000 30 90 Kinc i > 0 . 03 1,000 10 100 10,000 100 Total H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 13 / 48

Grouping individuals according to the relative importance of their households’ capital income: an example � 1 if Kinc i ≤ 0 . 01 Capital income levels = 2 if 0 . 01 < Kinc i ≤ 0 . 03 (1) 3 if Kinc i > 0 . 03 where: Kinc i is the ratio of capital income to total income of the household of individual i H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 14 / 48

Validation of IOPK: methodology To assess the validity of the capital income approach we make a strong assumption: that the estimates returned by the standard approach (including parental education in the set of circumstances) are accurate. Assuming that, we will compare them with the results returned by the capital income approach, and conclude that the latter are valid only to the extent that they are similar to the former. We will use the LIS and the EU-SILC database, and employ both the Gini index and the Mean Log Deviation as inequality measures. We will mostly show results using Personal Income as outcome, but we will refer to Equivalized Household Disposable Income at some points as well. We restrict our sample to individuals... aged 25 to 64 who are either active in the labor market (employed full or part time or unemployed) or fulfilling domestic tasks and care responsibilities As the set of circumstances, we consider binary gender, 4 age groups, immigrant status/ethnicity, population density and parental education/capital income. This means that the number of types can be up to 96. H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 15 / 48

Validation of IOPK: evolution of IOP and IOPK in Germany using the LIS database H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 16 / 48

Validation of IOPK: evolution of IOP and IOPK in Italy using the LIS database H. del Valle-Incl´ an Cruces IOPK: the capital income approach LIS, January 2019 17 / 48