Inequality of opportunity at the household level

Details

Economic inclusion

To what extent do circumstances at birth explain household wealth and individual educational attainment in the transition region?

To answer this question consistently for as many countries as possible, the analysis in this chapter is based on the 2010 round of the Life in Transition Survey (LiTS). This contains data for 38,864 households from 35 countries – 29 transition countries in Europe and Central Asia (but excluding Turkmenistan), as well as Turkey and the five western European comparator countrie.¹ The data were collected by interviewing randomly selected household members, of whom about 39 per cent (15,106 individuals) identified themselves as the head of the household.

For each outcome variable – either an index of household wealth² or a variable indicating whether the respondent had obtained a tertiary degree³ – an econometric model is estimated that establishes the extent to which circumstances at birth contribute to the variation in outcomes (see Annex 5.1 for details). This contribution, which in the case of the wealth index is simply the “fit” of the regression, is referred to as the (estimated) inequality of opportunity (IOp) with regard to either household assets (IOp^wealth) or educational attainment (IOp^edu).

A complication arises from the fact that the LiTS contains only information about the circumstances of the respondent member of the household, not those of other household members. By contrast, the asset index refers to the household as a whole. This is addressed by conducting the analysis of IOp^wealth using a subsample of households for which the respondent was the head of the household. Consequently, this analysis looks at whether the circumstances of the head of the household explain inequality in household wealth. Because spouses, domestic partners and other adult household members are often from similar backgrounds,⁴ IOpwealth should be a good proxy for overall inequality of opportunity with regard to household assets and adequate for the purposes of cross-country comparison.

One important limitation applies: because spouses or domestic partners are usually of a different gender, it makes no sense to measure the influence of gender on household wealth. While gender is always a characteristic, or “circumstance”, of the head of the household, it is rarely a circumstance of the household. Hence, it is not considered in the statistical analysis estimating IOp^wealth.

Gender is, however, considered in the estimates of IOpedu, because these address a different question – whether an individual’s circumstances or characteristics explain inequality in his or her educational attainment at tertiary level. In this context, gender is a potentially relevant circumstance. In addition, the analysis of IOp^wealth is undertaken separately for male and female-headed households to see if this affects the results.

Besides gender (for IOp^edu only), the analysis also considers the following circumstances.

• Whether a person was born in an urban or rural area: This investigates a possible source of inequality of opportunity due, for example, to geographically-determined differences in the quality of schooling or – since a person’s place of birth and place of residence as an adult are highly correlated⁵ – differences in job opportunities. It can also reflect access to basic services, such as roads, waste removal, indoor plumbing and electricity, which can directly and indirectly impact an individual’s economic opportunity.
• The level of educational attainment of the respondent’s father and mother: This may capture the influence of parental education on the quality and extent of a child’s education and act as a proxy for the individual’s social background and/or parental networks, which can provide opportunities for a child later on.
• Whether the individual’s parents were members of the communist party: In former communist countries party membership was often required for admission to specific schools and professions. In many cases, people serving in such professions received payment in assets in addition to income, which may have had an impact on the distribution of assets for the older generation.⁶ In addition, a parent’s membership of the communist party may act as a proxy for parental networks.

Other circumstances and characteristics, such as ethnicity, mother tongue, sexual orientation, religious background or physical disability, were not considered, either because of data constraints or because the categories in which these variables would have to be expressed vary greatly across countries. For example, most of the transition and Western countries studied in this chapter have no single ethnicity or mother tongue.

To illustrate how the circumstances considered affect the two outcome variables (household assets and tertiary education) in transition and comparator economies, Charts 5.1 to 5.4 plot a set of intra-country correlations. In Charts 5.1 to 5.3 the length of the left-hand bar (or axis) in each pairing represents the effect of a specific circumstance – being born in an urban (rather than rural) area, being born to parents with a level of educational attainment that is one notch higher,⁷ or having a parent who was a communist party member – on the household asset index.⁸ The right-hand bar denotes the impact of each circumstance on the probability of having completed tertiary education.⁹ Chart 5.4 shows how being male affects that probability.

As expected, the impact of parental education on the assets and tertiary education of children is positive almost everywhere, with particularly large impacts on the asset index in south-eastern Europe. The effect of an individual’s birthplace is more heterogeneous: being born in an urban area is generally a predictor of superior wealth and education. There are exceptions, however, particularly with regard to wealth; in France, Slovenia and the United Kingdom a rural birthplace is a statistically significant predictor of higher levels of household assets.

Having a parent who was a communist party member generally puts individuals in transition economies at an advantage. In regard to household assets, the effect is small and generally statistically insignificant, but for tertiary education it can be quite large (comparable to that of an urban birthplace) and is often statistically significant. In addition, men are more likely than women to have a tertiary degree in western Europe and most countries in eastern Europe and the Caucasus (EEC), while the reverse is true in most central European and Baltic (CEB) countries.

Having described country-level correlations between individual circumstances and outcomes, the next step in the analysis is to examine the extent to which circumstances at birth explain variations in household assets and tertiary education in transition countries.

1. Unfortunately, such data are not yet available for the SEMED countries, although a few studies have looked at inequality of opportunity in the SEMED region: see El Enbaby (2012), Belhaj Hassine (2012) and Salehi-Isfahani et al. (2011). [back]
2. The analysis focuses on household wealth because the LiTS lacks reliable income data. An asset index was constructed using principal components analysis, which yields a weighted average of the assets owned by a household. The technique is used extensively in the literature to capture "wealth"; see Filmer and Pritchett (2001), McKenzie (2005), Sahn and Stifel (2003), Vyas and Kumaranayake (2006) and Ferreira et al. (2011). LiTS-based inequality is correlated positively, although far from perfectly, with measures of income inequality (the coefficient of cross-country correlation with Gini coefficients taken from the Standardized World Income Inequality Database is about 0.25) [back]
3. Consistent with Chapter 4 of this Transition Report, this refers to university education only. Note that although studies in other regions use educational attainment at secondary level as a measure of economic advantage, this is not as meaningful in the transition region because virtually every transition country has achieved high rates of secondary school completion, comparable to rates achieved in advanced economies. In contrast, the completion rates for tertiary education in the LiTS range from 5.4 per cent in Kosovo to 38.5 per cent in Belarus. The median completion rate is 18.2 per cent. [back]
4. In particular, parental wealth is highly correlated within households. This relationship holds when parental wealth is instrumented using parental education. See Charles et al. (2013). [back]
5. The correlation in our dataset is 0.63, which is significant at the 1 per cent level. [back]
6. See Heyns (2005). [back]
7. Parents' educational attainment is measured as a discrete, ordered variable. A mother or father with no degree is given a value of 1 for this variable, and one who has completed primary education is given a value of 2. Secondary and post-secondary degrees are counted separately. Postgraduate tertiary education is assigned a value of 6. [back]
8. The asset index is centred on 0. Its distribution varies from country to country, but it typically runs from about -4 to +4, with a standard deviation of about 2. [back]
9. The impact on assets is based on country-by-country ordinary least squares (OLS) regressions of the asset index on circumstances; the impact on tertiary education is based on an analogous set of probit regressions. The impact on assets is measured in terms of index points, whereas educational impact is measured in terms of the probability of having completed a tertiary degree. For example, an impact of 0.6 on both assets and education means that a person born in an urban area will, on average, have an asset index 0.6 point higher than someone born in a rural area and is 60 per cent more likely to have completed tertiary education. Note that since the impacts on assets and education are measured in different units, they should not be directly compared with each other. However, impacts on assets and education can be compared (separately) across countries within each chart and across charts. (In light of possible omitted variable bias, the bar heights should only be taken as a rough guide.) [back]
10. This variable was omitted for regressions involving the western European comparator countries and Turkey. Including it for Germany does not make a qualitative difference to the results. [back]

• Prev
• Next