African Development Bank Group
Relative Deprivation and
Well-Being of the Rural Youth
Working P
aper Series
Tekalign Gutu Sakketa and Nicolas Gerber
Improve the Quality of Lifefor the People of Africa
n° 296 June 2018
5
theQ
Working Paper N
o296
Abstract
Relative income deprivation is one mechanism through which income or wealth inequality is hypothesized to affect human behaviour, with consequences on well-being. The study checks these effects against multiple self-identified reference groups using a unique rich panel data set from Ethiopia, enabling us to examine a broader range of questions related to youth well-being than in previous studies in developing countries. In doing so, the study extends the standard analysis of relative deprivation (RD) from income per se, to consider social relative deprivation as well as assets (non- monetary) relative deprivation. Since the effects of relative deprivation on well-being are also sensitive to the kind of measurements
employed, the study employs two measurements of relative deprivation: objective and subjective, and compere the results from both. Our empirical results indicate that objective relative income deprivation has a ‘’signal effect’’ or a ‘’positive externality’’—
higher income of others in the reference group indicate higher prospects for youth (that induce motivation), whereas the subjective income RD has a ‘’status effect’’—higher income of others in the reference group reduces life satisfaction. Both objective and subjective measures of social and non- monetary RD have a ‘’status effect’’. Our findings are robust to different specifications. The policy implications of the results are discussed.
Rights and Permissions All rights reserved.
The text and data in this publication may be reproduced as long as the source is cited. Reproduction for commercial purposes is forbidden. The WPS disseminates the findings of work in progress, preliminary research results, and development experience and lessons, to encourage the exchange of ideas and innovative thinking among researchers, development practitioners, policy makers, and donors. The findings, interpretations, and conclusions expressed in the Bank’s WPS are entirely those of the author(s) and do not necessarily represent the view of the African Development Bank Group, its Board of Directors, or the countries they represent.
Working Papers are available online at https://www.afdb.org/en/documents/publications/working-paper-series/
Produced by Macroeconomics Policy, Forecasting, and Research Department
Coordinator Adeleke O. Salami
This paper is the product of the Vice-Presidency for Economic Governance and Knowledge Management. It is part of a larger effort by the African Development Bank to promote knowledge and learning, share ideas, provide open access to its research, and make a contribution to development policy. The papers featured in the Working Paper Series (WPS) are those considered to have a bearing on the mission of AfDB, its strategic objectives of Inclusive and Green Growth, and its High-5 priority areas—to Power Africa, Feed Africa, Industrialize Africa, Integrate Africa and Improve Living Conditions of Africans. The authors may be contacted at workingpaper@afdb.org.
Correct citation: Sakketa T. G. and N. Gerber (2016). Relative Deprivation and Well-Being of the Rural Youth,Working Paper Series N° 296, African Development Bank, Abidjan, Côte d’Ivoire.
1
Relative Deprivation and Well-Being of the Rural Youth
Tekalign Gutu Sakketa and Nicolas Gerber1
JEL classification: D31, D63, I31, Z13
Keywords: Objective relative deprivation, subjective relative deprivation, income, non-income, social capital, subjective well-being, rural youth, Ethiopia
1 Tekalign Gutu Sakketa and Nicolas Gerber are staff of the Department of Economic and Technological Change, Center for Development Research, University of Bonn. Walter-Flex str.3, Bonn, Germany. Tele: +49 15222 975295.
Email: tekalign2003@gmail.ocm/ s7tesakk@uni-bonn.de
2 1. Introduction
Concern for status (or relative deprivation) is one mechanism through which income or wealth inequality is hypothesized to affect population well-being, such as health, happiness, or human capital formation (Stark and Taylor 1991; Graham, Nikolova, 2015). Extensive research carried out in developed countries has shown the importance of relative deprivation (RD)for individual well-being and behavior, since the time of Adam Smith (Alpizar, Carlsson, and Johansson- Stenman 2005; Smith 2010). More recently, this analysis has extended to empirically testing the importance of both subjective (stated) and objective (revealed) RD (Easterlin 1995; Clark et al. 2008; Akay and Martinsson 2011).
Surprisingly, there is only scarce empirical evidence of the impact of household relative income (RD), relative consumption (consumption deprivation), and relative social prestige (social deprivation) on well-being in developing countries (Salti 2010). Existing literature on RD focuses on the adult or mature population. In the context of sub-Saharan Africa (SSA), where more than 65 percent of the population is 25 years old or younger, this is a clear knowledge gap. The recent Arab uprisings and emerging empirical evidence also suggest that a concern for RD is more pronounced among the youth, and that this is not limited to richer countries (Pingle and Mitchell 2002, for the latter). What is evident across existing studies is that people do have status concerns that, in turn, affect their well-being and aspirations.
Social comparison among youth (as individuals or as groups) is at the heart of RD.
Relative deprivation may affect the well-being of people, in general, and youth, in particular, in several ways. “First, well-being is maximized when people live under conditions that mimic those under which humans evolved’’ (Chen 2015). For instance, hunter-gatherer societies punished those who deviated from customary practices of equal sharing of food (Deaton 2001).
Second, studies have shown that RD undermines the protective role of the biochemical system of stress response against wide range of human diseases. Third, rank, rather than absolute possession of resources (money) itself, may determine power and access to (exclusion of) material goods and services (Eibner and Evans 2005). A good example here is the occupational status, which may determine the degree of control people have over others. Fourth, empirical evidence has shown that RD affects health and happiness—the two most common indicators of well-being (Kondo et al. 2008; Subramanyam et al. 2009). Finally, RD can foster life satisfaction by promoting a stronger pursuit for status. Stark (2004) showed that increase in inequality of wealth prompts a stronger quest for status, which, in turn, fosters the accumulation of wealth. Thus, such feelings of RD diminish or enhance an individual’s well-being. Youth
3 population groups are usually responsive to such feelings of RD. Such behavioral responses often force individuals to shift their allocation of resources from meeting basic needs to the purchase of positional goods, such as mobile phones or expensive clothes, even though their absolute income remains low. It may also induce individuals to work harder in order to achieve the higher living standard that others in the reference groups have achieved.
Despite increasing coverage of research on RD, four issues remain unclear in the literature that tests the relative deprivation hypothesis in relation to population well-being: (1) the choice of an indicator (which indicator or object of social comparison—income, consumption, wealth, housing facilities, social capital, political connections) that an individual or a group uses to compare himself/herself against their reference group; (2) the choice of reference group (whether to use geographic proximity, such as village or region; demographic characteristics, such as age, gender, ethnicity, relatives, workplaces, peers, and religion; or economic reference groups, such as occupation, size of land holding, number of livestock holdings); (3) efforts to establish stronger causal designs (need for longitudinal studies with careful control for confounding by individual or household income and other indicators of socioeconomic position); and (4) inadequate research to advance innovative approaches to operationalize measurement of RD, including the measurement of RD in dimensions other than income (Adjaye-Gbewonyo and Kawachi 2012). To our surprise, the existing literature in economics relies on limited measure of relative income deprivation—a “unidimensional”
measure of RD (those derived uniquely from income); and no adequate empirical work has been carried out beyond the income-based approach to measure RD in other dimensions. The objective approach widely used in economics literature uses income as an object of social comparison to test the RD hypothesis while explicitly, or implicitly, assuming individuals compare themselves with individuals within the same reference group (comparison groups).
Measuring RD is not easy and an elusive issue (Adjaye-Gbewonyo and Kawachi, 2012). There are two analytical approaches to test the effect of RD on well-being. The first approach uses a revealed preferences approach based on survey data, employed here, using subjective and objective measures. In this paper, we employ both measures by disaggregating RD along different dimensions, and compare the results of the two measures. The objective measures of RD is common in economics literature, while the second measures is extensively used in sociology and anthropology literature. The difference between the two comparison measures of revealed preferences employed in this paper is that objective measures of RD of different dimensions is computed using the Yitzhaki Index, which takes into account the distances in incomes and non-income, while subjective measures of relative position or self-
4 rated deprivation relate the self-reported level of an individual’s relative standing in the income (wealth) distribution of the reference groups. The subjective measures of RD can be taken as the rank or informative only of the position of individual in the income (wealth), non-income, and social capital scales. The two measurement approaches are similar in terms of analytical approaches, but differ in terms of measurement, which we turn to later, discussing in greater detail. The second approach is based on the stated preferences theory. This approach is common in experimental economics.
Similarly, the critical challenge in the computation of RD is the identification of the
“reference group,” within which individuals and groups make economic and social comparisons (Subramanyam et al. 2009). Literature on this identification is inconclusive. It varies from individual to individual, from group to group, or even from society to society. The common practice in existing studies is for researchers to determine or assign a reference group to each individual or group (Stark and Taylor 1991; Stark and Wang 2000; Stark, Micevska, andMycielski 2009; Quinn 2006; to mention a few). The choice of reference groups critically affects the outcomes and policy implications. In this study, we therefore use 12 different self- reported reference groups, identified through structured questionnaire, focus group discussions, and observational experiences.
To test the effect of RD on subjective well-being (SWB) of youth, regression analysis has been used in the literature. In regression analysis, a variety of definitions and measures of RD are used, and whether these measures are valid in terms of their correspondence to conceptualizations of RD and their ability to capture the essence of the relative deprivation hypothesis, as it applies to well-being remains an open question (Adjaye-Gbewonyo and Kawachi 2012; Salti 2010). The traditional objective measures used to quantify RD are measures of relative income, employing the Yitzhaki Index (Yitzhaki 1979), Deaton formulation (Deaton 2001), and other income-based measures, such as Log-normal formulations and percentile rank (Eibner and Evans 2005; Li and Zhu 2006; Subramanyam et al. 2009), used with absolute level of income and ownership of basic needs, together with other relevant variables, including observed and unobserved individual characteristics. In doing so, the sign and significance of the relative income parameter are then used as approximation for RD among the population of interest. What is evident, as a general conclusion in the literature, is that relative deprivation is negative and significant in both economic and statistical senses, which implies that RD or relative concerns, on average, result in welfare loss (Akay and Martinsson 2012; Clark and Oswald 1996, 1998; Stark 2010;Ferreri-i-Carbonell 2005; Clark et al. 2008, among others).
5 Recent evidence has shown the importance of multidimensional relative deprivation—
RD that uses multiplicity of indicators or dimensions, such as social capital (social relative deprivation), ecological status (ecological relative deprivation), non-monetary indicators (e.g., consumption), and political status (political relative deprivation) beyond the space of income or monetary approach (Barnett et al. 2004; Sweet, 2011). Income or monetary indicators are inadequate alone in capturing the prevalence of deprivation, especially in developing countries because of the fact that (1) income is often underreported (understated) in most poor countries;
(2) rural livelihood or well-being is equally derived from non-income and social capital, since a significant portion of well-being is derived from non-income, such as in-kind transfers, gifts, material possessions, household amenities, and item ownership. For instance, Sweet (2011:260) suggested that ‘‘because individuals may not know others’ incomes, symbolic capital—the material possessions or consumption patterns one uses to display one’s social status—may be a more salient basis for social comparisons.’’
Likewise, Barnett et al. (2004) noted that it is not clear that income alone is most relevant to individuals, when they compare themselves with others.Another justification to move beyond the money metric measurement of poverty is that the poor people most commonly define poverty in terms of insecurity, rather than low income (Barrett and Carter 2000; World Bank 2001).Social deprivation or social relative concerns refers to deprivation in terms of the inability to fulfill the expectations and pressures of family, neighbors, and tutors, say school teachers, and the failure to participate in customary community events (Townsend 1987). It is best captured by measures of social capital, such as trust and cooperative relations between individuals who share social identity, e.g.,ethnicity and connections with others of comparative status and power (Subramanian and Kawachi, 2004). It points to power and prestige dimensions of RD.
The novelty of our research is fivefold. First, we extend the standard analysis of relative concerns (or RD) to income per se, and consider social relative deprivation, as well as non- monetary (non-income) relative deprivation. We investigate their effects on the life satisfaction of rural youth. We show that strictly monetary indicators of deprivation can be misleading, in terms of identifying the well-being impacts of deprivation. Second, using different types of reference groups, we show how critical the choice of reference groups can be; how it impacts the robustness of the results and identifies different avenues for policy intervention. Third, we rely purely on self-reported reference groups (as opposed to assigned reference groups), thus increasing the level of information about each reference group. In other words, we rely on respondents’ own classifications of reference groups in defining and operationalizing relative
6 deprivation measures, rather than our own classifications. Fourth, our paper is the first in the literature that combines these three innovations with objective and subjective measures of relative deprivation. We provide evidence that RD is a key determinant of life satisfaction for rural youth, and show that our results are robust to the different analytical approaches. Finally, our panel data set allows us to control for omitted variable bias and confounding factors that can lead to possible endogeneity. By doing so, we contribute to the literature, heretofore in conclusive and lacking of empirical evidence on the importance of relative concerns among young people and the relevance of multiple choice of reference groups in the analysis of subjective well-being of rural youth.
Overall, our results consistently imply that, although the magnitude of RD depends on the measurement of RD employed, the reference group definitions, and type of youth, the results converge such that RDs (of different dimensions) have statistically and economically meaningful impacts on SWB of rural youth. An increase in individual relative income deprivation prompts a stronger desire for hard work, which, in turn, fosters the accumulation of wealth and/or income and hence, enhances life satisfaction. Our findings also suggest the variation of the impacts of RD of different dimensions on youth SWB, for sons and daughters across reference groups. In addition, the effect of absolute income on SWB of male youth is consistent across the two measurement approaches employed, but vary for female youth members; and the results are robust across various measures of RD and reference group definitions. Furthermore, the results indicate that decomposing the contributions of each RD along different dimensions would help to avoid the averaging of positive and negative income and non-income RD, and SWB relations (by reducing problems of aggregation of RD). The same is true with splitting of male and female youth, as well as youth who are members and household heads.
The remaining parts of the paper are organized as follows. The next section reports those reference groups that are perceived to be important for youth respondents. Section 3 presents the results from objective measures of RD, including the data set, empirical framework, estimation techniques, and the impact of objective relative deprivation of different dimensions on subjective well-being. Section 4 presents the results from the subjective measurement of RD based on different specifications, and after presenting the empirical framework, elements of the subjective data set and the estimation techniques employed. In addition, the section discusses the impact of absolute income on subjective well-being for the different specifications. Finally, in Section 5, we discuss the results and make conclusions.
7 2. Relative deprivation and the choice of reference groups: Relative to whom?
The first step in computation of RD is the determination or choice of relevant “reference group”
with whom individuals or groups make comparisons about their life situations. The concept of the reference group was first explored in the 1960s in the field of social psychology (Runciman 1966). Runciman noted that the choice of reference groups is important for the study of status concerns. Akay et al. (2014) presented a brief history and explanation of the different reference groups.
The most difficult task in empirical tests of the RD hypothesis is the choice or determination of reference group. What is common in economics literature is that reference groups are chosen a priori by the author(s) based on geographic proximity; demographic characteristics, such as age, education, race, or gender; or other economic comparison groups, undermining the fact that individuals do not necessarily share the same reference group (Eibner and Evans 2005; Kondo et al. 2009; Subramanyam et al. 2009). Additionally, individuals may have multiple, simultaneous reference groups that affect their well-being in different ways.
Furthermore, unlike in developed countries, the presence and reliance on informal insurance systems in low-income countries adds to the complexity of reference group structures. For instance, an income increase of others in their network could either positively or negatively affect the utility of an individual belonging to that network or reference group.
Hence, the choice of reference group is consequential in that the effect of RDvaries, depending on how the reference group is specified and defined. Unlike previous surveys, our survey contained questions that asked people directly to whom they compare themselves. Like our study, we are aware of only a few studies which have employed this approach (Clark and Senik 2010a; Senik 2009; Knight et al. 2009; Carlsson and Qin 2010; Carlsson et al. 2007a;
Mayraz et al. 2009). Akay et al. (2014) asked how respondents perceive themselves, relative to others, by presenting multiple reference groups. They did not ask about the reference groups against which people compare themselves. The insignificant effect of relative income deprivation in poor societies could be partly due to the way the reference groups are defined and RD is measured, as well as operationalized. To overcome these inherent limitations, we propose the use of multiple reference groups identified via a structured questionnaire and focus group discussions (FGD) with different groups of societies, such as adult women, adult men and, most importantly, the youth, by explicitly asking respondents to whom they are more likely to compare their life situations or socioeconomic status.
8 In other words, respondents were asked to identify their most relevant reference groups,1used separately and in combination, to whom they make economic, social, and political comparisons. The responses are summarized in Table 1. Questions relating to the same multiple reference groups followed directly afterward:“How would you describe/position your gross income, non-income (material assets), and social capital[indicators listed below] in comparison to the youth in [the stated reference group].” Respondents were offered a seven- point scale, ranging from “the highest/richest (1)’’ to “the lowest/poorest (7),” the scales in a reverse order. Finally, the standard life satisfaction/dissatisfaction question was asked using a pictorially supported, “Suppose there are nine steps on this ladder. Suppose the very top (the 9th step) represents the best possible life for you and the bottom represents the worst life for you. Where on the ladder do you personally feel you stand?” The respondents were also requested to position themselves, compared to others, four years ago, three years ago, and one year ago, after stating their current standing.
This kind of approach gives respondents the freedom to make their own comparisons, thus avoiding the problem of reference group identifications. The FGD held with different groups of youth, adult men, and adult women, and also suggests that the youth and their parents do care about relative concerns. We examine in detail later which scope of comparison (within, upward, or society-wide) matters for youth well-being. Given the observed distribution of income and consumption intensities across the reference groups, we wanted to know to which reference group youth and their parents compared most intensely. First, we discuss the multiple reference groups identified. Later, for simplicity and empirical reasons, we use some of the identified reference groups mainly for the remainder of the analyses.
Rural youth may be slightly comparable, in terms of observed attributes such as age and education. It is not logical to assume that a youth who is 16 years old can compare himself/herself with someone 32 years of age. Hence, we narrow the reference group categorizing the youth into similar age groups (into four categories). Accordingly, we divide age brackets into 13–19, 20–24, 25–29, and older than 30. Education is divided into four different categories, according to the number of years of education: illiterate to barely literate, 1–6, 7–10, and more than 10 years of education. Of the sociodemographic reference groups, 88 percent and 58 percent of youth respondents indicated similar age groups and youth of the same religion, respectively, as relevant comparison groups to their subjective well-being. With regard to geographic area reference groups, 94 percent and 67 percent of youth respondents regard other youths in their village and worerda, respectively, as their relevant comparison groups, when they compare their economic, social, and political status with that of others. As
9 for the economic reference groups, for instance, focusing on the work-related reference group, 94 percent of respondents regard economic comparisons (such as size of landholding, number of livestock, and other youth in the same occupation) in their occupation as important when they compare their socioeconomic status with that of others. Since the educational infrastructure in our study areas is more or less similar, direct comparisons can be established.
Hence, it may be the case that youth in rural areas compare themselves to the entire rural youth population.
Thus, in this analyses, unlike the conventional approach of choosing reference groups a priori, we identify self-selected, multiple reference groups, used separately and in combination, based on geographic areas, as well as demographic and economic statuses.
Table 1. Self-identified reference groups (as reported by respondents)
Reference group Youth Father Mother
Yes (%) No (%) Sociodemographic reference groups
The same age group 88 12 ** **
Birth order/siblings 31 69
Other youth in the same ethnic group 55 45
Other youth in other ethnic groups 39 61
Educational level of other youth in the same village *** ** ***
Housing facilities of other youth in the same village *** ** **
Other youth in the same religious group 58 42
Other youth in other religious groups 41 59
Peers’ behavior (alcohol, and other) ** ***
Geographical areas reference groups
Other youth in the same Village (neighbors) 94 6 *** ***
Other youths in the same kebele 93 7 *** ***
Other youth in the same woreda 67 33 ** **
Economic reference groups
Size of land holdings 92 8 ** **
Number of livestock holdings 87 13 *** **
Other youth in the same occupation (all the same-type workers) 94 6 ** **
Source: Survey results and own compilation from FGD responses.
Note: *** and ** indicates the intensity of comparisons (or relevance of reference groups) for the different indicators (income, economic, and social indicators) is highly important, and important also for socioeconomic status comparison, as identified during the FGD. This reference groups were not included in the survey questionnaire and identified through FGD. Some reference groups overlap, for instance, ethnic group with the whole sample used in combination with others while other reference groups used in their own. Especially, geographic area reference groups are highly used in combination with other reference groups. The use of the same religion as reference group was so strong in districts where Muslims and Protestants are dominant; whereas, the use of different religion is so strong in dominant Muslim areas.
Based on the responses, we decided to use the following reference groups: (1) sociodemographic reference groups (other youth of similar age group, education level, housing
10 facilities); (2)geographical areas reference groups (at the levels of village, kebele, and woreda) (indicator of neighbors as comparison groups or reference groups); 3) economic reference groups (land size, number of livestock owned measured in Tropical Livestock Unit (TLU), and youth doing a similar occupation or youth having similar experience); and combinations of the three categories. The concept of neighborhood was approximated using the entire village (enumeration area), entire kebele, and entire woreda.
In this study, some reference groups are also used as a comparison indicator/variable.
For instance, educational level, housing facilities, and number of livestock owned could serve also as comparison variables.As shown previously, religion—which is often ignored, even as religious adherence is on the rise in developing countries—is becoming an important factor for economic and social inclusion (or exclusion), and poverty alleviation (or aggravation)is, hence, identified as one of the reference groups. The 2015 Global Attitudes Survey looked at how people around the world feel about religion in their lives. The study found that 98 percent of Ethiopians consider religion as a very important part of who they are (Pew Research Center 2017).
3. Objective measures of relative deprivation and subjective well-being of rural youth 3.1. Data
Our main dataset for this study is drawn from the novel Ethiopian Agricultural Growth Program (AGP) survey, covering agricultural potential areas of Oromia, Ethiopia. The AGP survey, a detailed agricultural panel survey carried out in 2010–11 in four major regions (Oromiya, Amhara, Southern Nations Nationalities People (SNNP), and Tigray), and in the Oromiya region in 2014–15 in a subsample of households and youth. The first wave of survey was implemented jointly by Central Statistical Agency (CSA) and the Ethiopian Strategic Support Program (ESSP) of the International Food Policy Research Institute (IFPRI) during July, 3–22, 2011. The second wave was carried out during the months of December 2014 and January 2015. Out of the four regions of AGP sites, our study focuses on Oromiya region and exclusively on youth (both as household members and as household heads) and their families (mainly, fathers and mothers) subsampled from the first wave in 2014–15. The data collection process for this research has three phases to capture youth well-being dynamics at the household level. First, participatory rural appraisal techniques were employed to understand the context and how relative concerns are perceived in the society. In the second stage, pretests
11 of the survey questionnaires were administered to a few respondents. Finally, socioeconomic surveys were addressed to the main respondents.
Multi-stage sampling techniques were employed to subsample households with youth members and youth-headed households during the second wave. A subsample of 660 youth from 521 randomly selected households, included during baseline, were selected randomly during the second wave and participated in the survey. Households who qualified, but were not available due to death or migration or difficult to track, were replaced from the contingency list. In this analysis, youth has been defined as persons within the age interval of 13 to 34. The survey collected detail information on youth characteristics, household characteristics, father’s characteristics, mother’s characteristics, wealth of households, as well as youth’s separate wealth, income,2 employment, social networks, life events, reference groups, and subjective well-being.
The community questionnaire included location and access to basic infrastructures;
institutions and youth-related projects and interventions in the areas. After eliminating some questionnaires, due to missing information, non-response, and inconsistency, we obtained a sample of 1,162 observations, based on 620 youth individuals covered in 2010–11 and 2014–
15. The subjective well-being measure is obtained by asking the subjects [the Cantril ladder questions] “How satisfied are you at present with your life as a whole?” and having them respond using a9-point scale from 1 to 9; where 1 represents the worst possible life and 9 represents the best possible life (Cantril 1965). The standard life satisfaction/dissatisfaction questions were pictorially supported with the picture of a ladder. The questions were along different time horizons. For instance, we stated to the respondent that, “There are nine steps on this ladder. Suppose the very top (the 9th step) represents the best possible life for you and the bottom represents the worst possible life for you. Where on the ladder do you personally feel you stand at the present time, three years ago, and one year ago?”
In the econometric specifications, we follow up the main approach in the analyses of SWB and regress income and other socioeconomic factors such as health and marital status, among others, on youth SWB (Van Praag and Ferrer-i-Carbonell 2008). We control for both absolute and relative income of the different dimensions (as described later in this paper), where income is entered in the analyses as the logarithm of the income per capita from all sources,3 including government transfers, in order to control for differences in household size together with relative income deprivation. In order to take account of inflation, all income measures are deflated to 2011 prices.
12 In the objective measures to relative deprivation along different dimensions, we categorize RD into income, non-income (both called economic relative deprivation), and social relative deprivation. We construct separate indices for each dimension of RD using the Yitzhaki Index across the different reference groups. Computation techniques are presented in the following section. Finally, we relate these indices with youth well-being, along with other control variables. The scales of the responses for non-income and social indicators show the weighting of indicators implicitly done by the respondent.
We follow a series of steps (in line with Elgar et al. 2016) in computing objective RD from the scores provided for social and non-income indicators. In this paper, income used for the computation of the relative deprivation index, using the Yitzhaki Index, is yearly household income, which is calculated as the sum of sale of crops, off-farm income, sale of livestock and livestock products, oxen rentals, gifts, transfers, and remittances.
3.1.1. Measuring subjective well-being
Like RD, the measurement of well-being has always been a controversial issue and subject of debate in economics. In this study, we use self-reported measures of well-being (that is, self- reported subjective metrics—subjective well-being) as proxies for youth well-being, captured using specially designed survey questions. This approach has been widely used in literature (Ferrer-i-Carbonell and Frijters 2004; D’Ambrosio and Frick 2007; Akay and Martinsson 2011; Easterlin 1995, among others).
Subjective well-being is captured in two ways: by interviewing youth in surveys using a single-occasion, self-reported question and from the work satisfaction questions of the general Edenred-Ipsos Barometer. The first approach is based on the standard life satisfaction measures on a 9-point scale, ranging from 1 (indicating the worst possible life) to 9 (the best possible life). In this approach, each youth respondent was requested to evaluate his/her well- being as a reflective assessment of one’s life, as a whole rather than a description of an emotional state. The second approach,which employs the general Edenred-Ipsos Barometer, uses sum of the answers to 17 questions of the Edenred-Ipsos Barometer, categorized into three pillars scores from 1 to 5 (1–strongly disagree and 5–strongly agree). Judgments about life satisfaction specific to life domains, such as work, health, community (and/or social networks), and relationships were applied here to capture satisfaction with life. Since all other variables of interest, such as income and other explanatory variables in the model, are captured during baseline data, we used a recall method to capture SWB of individuals at the time of baseline.
13 This technique enabled us to make full use of the panel nature of our data set, since panel data allows for control of otherwise unobserved individual characteristics. As such, respondents were asked to rate their satisfaction with life over the last five years: four years ago, three years ago, and one year ago, compared to their current situation during the second wave of the survey. This exercise also enabled the respondents to assess their personal well- being over time, based on their own opinion. After all, SWB is subjective to the individual in question and it is completely left up to the individual to explain his SWB or level of satisfaction (McBride 2001). On the other hand, these scales of measures of satisfaction may result in panel effects (learning effects), that is, respondents may tend to use these scales differently after
“getting used” to them. For instance, respondents may stay away from the extreme values such as “9”.
Figure 1 shows the distribution of subjective well-being across one of the different relevant reference groups (woredas). As indicated below, the distribution of SWB is not similar across the reference group, using woreda, for instance.
Figure 1. Distribution of subjective well-being across woredas
Source: Authors.
0.1.2.3 0.1.2.3 0.1.2.3
0 5 10 0 5 10 0 5 10 0 5 10
1 2 3 4
5 6 7 8
9 10 11 12
Density
SWB across woredas
14 3.2. The empirical framework and econometric methods: Objective measures
The theoretical foundation of this paper is based on the concept of RD put forward by Runciman (1969) and operationalized by Yitzhaki (1979) and Stark and Yitzhaki (1988). We conceptualize that youth satisfaction in life depends on absolute and relative income (or wealth) of their own or parents, their social capital, and the social capital of their peer groups.
As stated earlier, “objective” measures of RDare decomposed along three different dimensions.
Relative income deprivation (RD), which is the common income-based RD measure of the Yitzhaki Index, is defined as the gaps between the individual’s (or household’s) income or wealth and the incomes or wealth of all individuals or households richer than the individual (or household) within a reference group. According to this measure, individuals or households within the same reference group and with identical income, Y, all experience the same level of RD.
The same is true with other dimensions of RD, such as social capital deprivation (SD) and non-monetary relative deprivation (NID),that is, deprivation in material assets. With some extension of Stark and Zawojska (2015), we model the link between RD of the different dimensions,in which income, non-income, and social capital are the objects of relative comparison, and youth well-being, using objective measures in the following way: Consider a youth population of n, in which every member of n has a positive level of income, Yi. The income distribution among youth are given by 𝑌1 < 𝑌2 < ⋯ < 𝑌𝑛; where 𝑌𝑖 denotes the income of the household to which youth i belongs. In the same manner, the social score distribution, like the income distribution,is given by 𝑆1 < 𝑆2 < ⋯ < 𝑆𝑛; where Si denotes the social score (or capital) of the ith youth. Likewise, the non-income items or material possessions one uses to display one’s social status, compared to those generally owned in his or her reference groups, are also given by 𝑁𝐼1 < 𝑁𝐼2, … < 𝑁𝐼𝑛; where 𝑁𝐼𝑖 denotes the non-income items or patterns of the ith youth belonging to a household.
Thus, the utility (well-being) function, Ui, of youth i belonging to population n, can be defined as follows:
𝑈𝑖(𝑌1, … , 𝑌𝑛) = 𝛽𝑖𝑌𝑖− (1 − 𝛽𝑖)𝑅𝐷𝑖𝑟(𝑌1, … , 𝑌𝑛) +𝜃𝑖𝑆𝑖− (1 − 𝜃𝑖)𝑆𝐷𝑖𝑟(𝑆1, … , 𝑆𝑛)
+ 𝛿𝑖𝑁𝐼𝑖 − (1 − 𝛿𝑖)𝑁𝐼𝐷𝑖𝑟((𝑁𝐼1, … , 𝑁𝐼𝑛), (1)
where 𝑅𝐷𝑖𝑟(. ) is a measure of relative income deprivation; 𝛽𝑖Ͼ (0,1) expresses the weight accorded by youth i to his/her parents’ income and (1-𝛽𝑖) expresses the intensity of concern that youth i attaches to relative income; 𝑆𝐷𝑖𝑟 (.) and 𝑁𝐼𝐷𝑖𝑟 (.) are measures of relative social
15 deprivation and relative non-monetary deprivation, respectively; θiϾ (0,1) is the weight accorded by youth i to his/her social capital; (1 − 𝜃𝑖) and (1-δi) express the intensity of concern that youth i attaches to relative social capital and relative non-monetary income, respectively;
and r denotes types of self-identified reference groups presented in Table 1. The measure of relative income, relative non-income, and relative social deprivation of youth i, who is a member of a reference group of n individuals (the subpopulation of all individuals belonging to the same reference group (r) such that i=1, 2,…, n), is defined as the weighted sum of the excesses of incomes, non-incomes, and social capital (connections) higher than Yi, Si, NIi, such that each excess is weighted by its relative incidence.
To operationalize objective measures of relative deprivation, we calculated RD for each youth within identified reference groups using the Yitzhaki Index (Yitzhaki 1979). For instance, the relative income deprivation function of youth i with household income Yi, who is a member of a self-identified reference group (r) of n individuals, is given as follows:
𝑅𝐷𝑖𝑟(𝑌1, … , 𝑌𝑛) = 1
𝑛∑𝑛𝑗=𝑖+1(𝑌𝑗− 𝑌𝑖) , (2)
where 𝑌𝑗>𝑌𝑖; noting that for any j≤i, max {𝑌𝑗−𝑌𝑖, 0} = 0; j are individuals whose income are greater than i; n is the subpopulation of all individuals belonging to a comparison group or reference group, r, i.e, the number of individuals who are in the r reference group. Note that n varies with the kinds of reference groups used. The Yitzhaki Index is an “upward-looking”
index of deprivation by construction. Based on this construction, we model and calculate “r=
12” estimates of RD for each youth i. One of the prominent findings in this study that deserves special attention is the direction of the effect of relative income deprivation on SWB. If the economic success (income in this case) of other individuals or households in the reference groups depresses youth welfare, it means that the coefficient of RD is negative and interpreted as ‘’status effect.’’ On the other hand, youth well-being can be positively affected by the income of the relevant peer groups.
Under such conditions, we expect that the coefficient of RD is positive and can be an indication of a ‘’signal effect or positive source of information’—higher income of others in the reference group indicate higher prospects for youth and, hence, shapes future expectations.
It means also that youth build aspirations based on the achievements of other peers, such as based on the standard of living of other youth of similar age, occupation, etc. Positive effects of relative deprivation on SWB could be also related to pure ‘’economic externalities,’’ where relative income (deprivation) acts as a proxy for the benefits of living with rich(er) people or in wealthier neighborhoods (Ferrer-i-Carbonell 2005).
16 We compute the social relative deprivation by summing up the responses of the 17 questions of social capital indicators that use the scores from 1 to 5 (1–highly deprived or strongly disagree and 5–least deprived or strongly agree) and defined as the weighted sum of the excess of social scores higher than 𝑆𝑖, such that the excess is weighted by its relative incidence:
𝑆𝐷𝑖𝑟(𝑆1, … , 𝑆𝑛) = 1
𝑛∑𝑛𝑗=𝑖+1(𝑆𝑗− 𝑆𝑖), (3)
where 𝑆𝑗 > 𝑆𝑖; noting that j≤i, max {𝑆𝑗−𝑆𝑖, 0} = 0.
A similar approach is used in Elgar et al. (2016). Mathematically, the same approach is also employed to construct relative non-income (non-monetary) deprivation (NID) from the non- income scores/items (NI) as follows:
𝑁𝐼𝐷𝑖𝑟(𝑁𝐼1, … , 𝑁𝐼𝑛) = 1
𝑛∑𝑛𝑗=𝑖+1(𝑁𝐼𝑗− 𝑁𝐼𝑖), (4) where𝑁𝐼𝑗 > 𝑁𝐼𝑖 and j≤i, max {𝑁𝐼𝑗−𝑁𝐼𝑖, 0} = 0
We expect that relative social deprivation and non-monetary relative deprivation are negatively associated with youth well-being. However, large social networks improve well- being significantly (Akay et al. 2012). Generally, a utility function encompassing the three dimensions of relative deprivations can be expressed in the following relation:
𝑈(𝑖, ℎ) = 𝑆𝑊𝐵(𝑅𝐷𝑖𝑟, 𝑆𝐷𝑖𝑟, 𝑁𝐼𝐷𝑖𝑟, 𝑌𝑖, 𝑆𝑖, 𝑁𝐼𝑖 𝑋), (5) where X is a set of other control variables, such as individual, household, and community characteristics determining the well-being or life satisfaction of an individual i who belongs to household h.
Alternatively, the above relationship can be expressed as follows, where the different dimensions of relative deprivation are the function of the respective income, social capital, and non-income of the reference groups:
𝑈(𝑖, ℎ) = 𝑆𝑊𝐵(𝑅𝐷𝑖𝑟(𝑌𝑖, 𝑌𝑗), 𝑆𝐷𝑖𝑟(𝑆𝑖, 𝑆𝑗), 𝑁𝐼𝐷𝑖𝑟(𝑁𝐼𝑖, 𝑁𝐼𝑗), 𝑌𝑖, 𝑆𝑖, 𝑁𝐼𝑖 𝑋) (6)
Given the ordinal nature of the dependent variable (SWB), the ordered probit specification would be an appropriate method employed in regression. In order to make full use of the panel nature of our data, controlling for otherwise unobserved individual characteristics, and potentially different use of the underlying satisfaction scale (running from 1 to 9) across individuals, an ideal approach would be to employ a fixed effects estimator.
Unfortunately, such a fixed-effects ordered probit estimator does not exist in standard statistical software packages. As an approximation, we use linear fixed-effects regression models, in
17 addition to the use of random-effects ordered logistic regression models. The first alternative approximation has been commonly used in literature (Ferrer-i-Carbonell and Frijters 2004;
D’Ambrosio and Frick 2007).
Our default model specification considers SWB as latent:
𝑆𝑊𝐵𝑖𝑡∗ = 𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒log(𝑌𝑖𝑡,ℎ) + 𝛾′𝑍𝑖𝑡+ 𝜎𝑘+ 𝑢𝑖𝑡, (7) where 𝑆𝑊𝐵𝑖∗ is the self-reported SWB of youth i on a subjective scale; 𝑌𝑖 is absolute per capita income (PCI) of youth i that belongs to household h in year t in log form; Zi is a vector of observable individual, household, and community characteristics, which affect well-being; 𝜎𝑘 is district level and other individual fixed effects (unobservable time invariant)that captures unobservable differences; and 𝑢𝑖 is an error term, which is assumed to be normally distributed with mean zero and a variance of one. This is the benchmark model against which we compare our results using multiple reference groups. To test the impact of RD of different dimensions on the well-being of youth, we extend our specification in (7) above as follows:
𝑆𝑊𝐵𝑖𝑡∗ = 𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒log(𝑌𝑖𝑡,ℎ) + 𝛽𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 log(𝑅𝐷𝑟(𝑌𝑖𝑡)) + 𝜃𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑆𝑖𝑡+ 𝜃𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑆𝐷𝑟(𝑆𝑖𝑡)
+ 𝛿𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑁𝐼𝑖𝑡 + 𝛿𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑁𝐼𝐷𝑟(𝑖𝑡) + 𝑍𝑖𝑡′𝛾 + 𝜎𝑘+ 𝑢𝑖𝑡, (8) where 𝑅𝐷𝑟(𝑌𝑖) is the income RD of youth i with respect to different reference groups, r; 𝑆𝑖 is an index constructed from different indicators of social status (social capital), using principal component analysis (PCA) (different indicators used to compute the social index is presented in appendix); 𝑆𝐷𝑟(𝑆𝑖) is social relative deprivation of youth i in the reference group r, as defined in the same way as previously; 𝑁𝐼𝑖 is non-income index computed from the different scores of non-income, which are economic indicators (see Appendix Table A.1); 𝑁𝐼𝐷𝑟(𝑁𝐼𝑖) is non-income relative deprivation of youth i who belongs to reference group ras previously defined; (. )𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 and (. )𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 are parameters for absolute and relative income, non- income, and social capital to be estimated. In the estimations, we employ a number of different specifications to test the robustness of our results. For instance, we separately estimate the different specifications presented here for youth members and youth household heads; and for male and female youth. We also include the fathers’ and mothers’ characteristics (Equation 9), and interaction terms (Equation 10) to the specifications here, expressed as follows:
𝑆𝑊𝐵𝑖𝑡∗ = 𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒log(𝑌𝑖𝑡,ℎ) + 𝛽𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 log(𝑅𝐷𝑟(𝑌𝑗𝑡)) + 𝜃𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑆𝑖𝑡+ 𝜃𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑆𝐷𝑟(𝑆𝑗𝑡)
+ 𝛿𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑁𝐼𝑖𝑡 + 𝛿𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑁𝐼𝐷𝑟(𝑁𝐼𝑗𝑡)
+ 𝜌𝐹𝑖𝑡+ 𝜇𝑀𝑖𝑡+ 𝑍𝑖𝑡′𝛾 + 𝜎𝑘+ 𝑢𝑖𝑡 (9)
18 𝑆𝑊𝐵𝑖𝑡∗ = 𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒log(𝑌𝑖𝑡,ℎ) + 𝛽𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 log(𝑅𝐷𝑟(𝑌𝑗𝑡))
+ 𝜃𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑆𝑖𝑡+ 𝜃𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑆𝐷𝑟(𝑆𝑗𝑡) + 𝛿𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝑁𝐼𝑖𝑡 + 𝛿𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑁𝐼𝐷𝑟(𝑁𝐼𝑗𝑡)
+ 𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒log(𝑌𝑖𝑡,ℎ) ∗ 𝐸𝑑𝑢𝑚𝑜𝑚
+𝜌𝐹𝑖𝑡+ 𝜇𝑀𝑖𝑡+ 𝑍𝑖𝑡′𝛾 + 𝜎𝑘+ 𝑢𝑖𝑡, (10) where F and M denote father’s and mother’s characteristics.
As stated earlier, we expect that absolute income, non-income, and social networks or social capital affect SWB positively (𝛽𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒> 0; 𝜃𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 > 0; 𝛿𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 > 0). However, the effects of relative income deprivation, non-income deprivation, and social deprivation are a priori undetermined, that is, their effects could be negative or positive. If the effects are negative (𝛽𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 < 0; 𝜃𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 < 0; 𝛿𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 < 0), it implies a ‘’status effect’’ of relative deprivation. However, it is possible that these relative deprivations of income, non-income, and social capital could affect the well-being positively (𝛽𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 > 0; 𝜃𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 >
0; 𝛿𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 > 0), implying a‘’signal effect’’of RD.
3.3. Results and discussions: Evidence from objective [measures of] relative deprivation
We first present the main descriptive results of the characteristics of our respondents. Next, we present the econometric results of the various SWB functions, with different specifications presented earlier.
3.3.1. Descriptive results from the objective measures
Table 2 summarizes the basic characteristics of our youth sample. The average age of our youth sample is about 18 and 22 years in 2010 and 2015, respectively. The youth sample contains mostly male (63 percent of youth respondents are male). The majority of them live with their parents (73 percent and 71 percent in the years 2010 and 2015, respectively). The average years of education was about 2.83 years during the baseline survey. The average years of education increased to 4.20 years during the follow-up survey. About 72 percent of youth samples were single in 2010–11. This percent has decreased to 64 percent in 2014–15. The average farm size per own capita is about 0.56 hectares in 2010–11, and this has decreased to 0.53 hectares in the 2014–15 agricultural production seasons.
19 Table 2. Characteristics of the youth respondents
Category 2010–11 2014–15
Age in years (mean) 18.45 21.54
Gender (%)
Male 62.97 62.97
Female 37.03 37.03
Years of education 2.83 4.20
Marital status (%)
Single 72.08 63.79
Married (single spouse) 27.92 36.21
Other - -
Family size 7 6.00
Farm size per capita (ha) 0.59 0.53
Birth order 3.0 3.00
Occupation (main) (%)
Part-time/ student/domestic worker 43.47 33.55
Full-time farmer 54.46 59.52
Non-farm worker 2.07 6.94
Youth type (%)
Head 27.00 29.00
Member 73.00 71
Source: Survey results
Table 3 summarizes the main descriptive results for the different dimensions of RD, computed using Yitzhaki Index for the year 2014–15. The lowest income relative deprivation index is obtained when youth compare themselves with that of their peers (when youth of similar type is used as a reference group). The highest income RD is obtained when youth compare their per capita income with that of other youth of similar age groups, the same ethnic groups, similar education, and youth from the same occupation. As to the non-income RD, the highest and the lowest deprivation score is obtained with respect to reference groups (land size,education, ethnicity), and other youth in the same kebele, respectively. The three dimensions of relative deprivation indices indicate that youth are feeling more deprived when the scope of geographic reference groups increases from village to woreda level. Hence, reference group definition and selection matters in well-being analysis. Summary statistics of other relevant variables used in the regression models are presented in Appendix A.
20 Table 3. Descriptive statistics of the indices across multiple reference groups for objective measures: Median values reported for social and non- income relative deprivation
Sociodemographic reference groups
Geographical areas reference groups
Economic reference groups Composed reference group Age Education Youth
type
Ethnicity* Village level
kebele Woreda level
Land size
Livestock holding
Occupa- tion
Age and woreda
Age and occupation Economic
Relative deprivation
RD (birr) 1156.22 1.04 0.45 1155.98 829.63 840.96 1052.27 1020.86 990.48 1124.38 1132.18 344.19
ERD (years) 1.61 - 0.001 1.52 1.33 1.26 1.36 1.50 1.40 1.41 1.38 1.51
NID 1.13 1.16 0.001 1.16 0.98 0.97 1.02 1.16 1.14 1.12 1.13 1.14
Social relative deprivation
SD 2.12 2.11 0.001 2.13 2.05 2.07 2.09 2.12 2.10 2.13 2.11 2.11
RD (birr) 659 659 659 659 659 659 659 659 659 659 659 659
Source: survey results
Note: * Overlaps with the whole sample, as all belong to the same ethnic group. Kebele reference group overlaps with religion as reference group.
21 We asked also respondents how important are the comparisons of different income, non-income or wealth (livestock ownership, housing facilities, clothing), and social capital (specifically, connections or networks) to them, in affecting their life satisfaction compared to their most relevant groups (for instance, we ask: which of the following comparison indicators are the most relevant to you for comparing your own life situation, to that of others?).
Respondents were requested to respond either “important” or “not important” to the list of comparison indicators (the object of comparison) provided to them. This question was proceeded by a question that asked respondents whether they compared their living conditions with others and whether these kind of comparisons are relevant for their life satisfaction. If a respondent was indicateda ‘’yes’’ response, a follow-up question was asked to find out which reference group(s) or comparison group(s) individually, or in combination, is (are) at stake in comparing one’s own socioeconomic status with that of others (that is, positioning one’s life situation) using dichotomous questions until the list is exhausted.4
The summary results of the most important objects of comparison, in addition to income, are presented in Table 4. Most respondents regarded cash income (92 percent), livestock ownership (89 percent), living environments (access to basic infrastructure such as water, road, and electricity, 76 percent), education (80 percent), housing facilities (86 percent), clothing (85 percent) and connections (75 percent) as important (relevant) to their assessment of well-being (life satisfaction) with that of others. Hence, youth regard not only cash income but also non-income and social comparisons essential for their life satisfaction. The respondents indicate also that these comparison indicators (criteria) change over time.
