Local inequality as a determinant of subjective well-being

(1)

0

Faculty of Economics and Business

Amsterdam School of Economics

Master Thesis

Pía Arce

August, 2016

Student Number: 11123796

Supervisor: Prof. Menno Pradhan

Local inequality as a determinant of subjective

well-being

(2)

1

Statement of Originality

This document is written by Student Pía Arce who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

Abstract

A great deal of research exists on the topic of happiness and inequality comparing different countries and regions in the world. Nevertheless, the relationship between local inequality and life satisfaction has often been neglected from economic literature. Filling this knowledge gap is important because it allows to understand better the impact of inequality on well-being within a country, and not only at a macro level. In this study, I linked a Life Satisfaction variable from the Ecuadorian Survey of Living Standards 2013 to local inequality at the level of Parroquia extracted from the 2010 Ecuadorian Poverty Map. To estimate this relation, I regressed an ordered probit model with cluster correction of errors. I found that local inequality has a negative effect on life satisfaction and that households at the bottom of the distribution are more negatively affected than other households. Finally, households at the top of the consumption distribution do not seem to be affected by local inequality. I conclude that Ecuadorians, mainly the poorest, are negatively affected by local inequality.

(4)

3

I. Introduction

The study of happiness and subjective well-being (SWB) has been gaining strength for several years now. Many studies have been devoted to studying the socio-economic determinants of happiness.

Inequality is one of the variables that have recently been studied as a relevant determinant of SWB. So far, inequality has shown mixed results as a determinant of SWB in literature. In some studies, income inequality seems to increase life satisfaction, reporting higher well-being than people from more equal countries (Rözier & Kreeykamp, 2013). However, other studies show that it decreases life satisfactions in other countries (Dolan, Peasgood & White, 2008; Hajdu & Hajdu, 2014; Alesina, di Tella & MacCulloch, 2004; Helliwell, Layard & Sachs, 2012).

I want to contribute to the literature that links SWB and inequality by analyzing the effect of inequality at the level of local community on life satisfaction. I want to determine whether local inequality affects well-being and whether this effect varies depending on the economic status of the household relative to their reference group. I chose Ecuador to conduct this study because of two reasons: First, Ecuador presents a Household Survey of very good quality, including questions about life satisfaction. Also, it has a publicly available Poverty map created with this Household Survey. Secondly, Ecuador is placed in a region where poverty has declined but inequality remains high, making the reduction if inequality an important policy target. We will argue that Ecuadorians are negatively affected by local inequality.

People dislike inequality for several reasons: it creates lack of social cohesion (Wilkinson R. G., 1997) and changes people’s perception of their own opportunities and aspirations. One of the more salient reasons for their dislike is that people think it is unfair (Helliwell, Layard, & Sachs, 2012; Hajdu & Hajdu, 2014).

Studies have often found that relative income is an important determinant of SWB. Relative income theories state that it is not only income or consumption that matters for individual happiness, but also whether they are higher or lower in the income ladder compared to a certain group of reference. The reference group can be any group to which the individual chooses to compare with in terms of consumption or income (Wilkinson, 1997; Schwarze & Härpfer, 2007). In this study, I will argue that Ecuadorians tend to compare themselves to other people living in their municipality of local community, which makes these people their relevant reference group. I also argue that, since one of the households’ reference groups is its local community, local inequality should be studied as a determinant of SWB.

(5)

4 This topic is important because most of literature and studies about inequality and SWB focus in the comparison between countries, using indexes at a country level scale. Thus, most researchers have overlooked the effect of local income/consumption inequality on happiness. However, literature that I have collected on this topic shows that local inequality may be also relevant. In this study, I provide more evidence about the effect of local inequality in SWB and thus contribute to the knowledge in this area.

Another important feature in this study is that I employed Poverty Maps as a tool of analysis. Poverty maps are meant to be instruments for policy targeting and allow us to measure the inequality at local level in a particularly precise manner. This research brings a new use to Poverty Maps, using them as input for analysis.

My first hypothesis is that local inequality decreases life satisfaction. One explanation for this is inequality aversion, which states that people have an intrinsic preference for fairness. Another explanation is the decreasing marginal utility. The change in consumption that comes from a reduction in inequality will mean a higher increase for the well-being of the poorest compared to the decrease in well-being for the richest; this will have a positive net effect. The third explanation is that high levels of inequality have a negative impact on societal outcomes that, in turn, negatively affect social cohesion and decrease households’ life satisfaction. In this study, I measure the net effect.

My second hypothesis is that households at the bottom of the distribution are more negatively affected by local inequality. I want to test whether inequality has different effects on SWB depending on household’s status. Research shows that those at the bottom of the distribution tend to be more negatively affected by inequality (Schwarze & Härpfer, 2007) and I believe this is also true when measuring the effect of status within their local community, since this is the relevant comparison group for this study.

For this purpose, I used an ordered probit estimation with clustered standard errors to measure the impact of local Gini on SWB. In this model, I also included the relative consumption, an interaction between local inequality and relative consumption, household consumption per capita, and a set of personal characteristics and other characteristics of the local community to control for common factors.

I found that local inequality affects life satisfaction negatively and that households at the bottom of the distribution are more negatively affected by the increase of local inequality than those in the middle. However, I found that households at the top of the distribution, contrary to my predictions,

(6)

5 are not affected by local inequality. I also demonstrate that local inequality has a robust negative effect at the level of local community (or municipalities), but this effect is less robust at higher levels of disaggregation.

The next sections are organized as follows: section II presents a brief literature review about the existing evidence of the relationship between consumption/income, inequality and life satisfaction. Section III presents contextual information about the case of Ecuador in Latin America. Section IV states this paper’s hypothesis in detail and explains the empirical strategy. Section V presents the data and variables used in this model. Section VI are the preliminary and empirical results. Finally, section VII presents the conclusions, discussions, further research and improvements.

II. Theoretical considerations and empirical precedents

SWB literature has increased over the last years in various fields of study, such as medicine, psychology and economics. In this paper, I will focus on the study of social and economic determinants of SWB.

Several economists argue that SWB can be used as an alternative measure of utility function (Frey & Stutzer, 2002; Kingdon & Knight, 2006; Dolan, Peasgood & White, 2008). Other measurements of well-being, such as the Human Development Index and Poverty Measures, often require a value of judgment about the weight of each component in the final index. Self-reported SWB allows to quantify a broader concept of what constitutes an adequate standard of living, beyond income or consumption levels, that includes no judgment of value in the weighting of input (Kingdon & Knight, 2006).

Typical social and economic determinants of SWB found in the literature are income or consumption, including relative income and income inequality; personal characteristics, such as age, gender or ethnicity; and socially developed characteristics, such as education or health.

Much research has been conducted in the area of income and has produced mixed results. Early studies have shown that, in Western developed countries, average happiness has not increased along with real income growth (Easterlin, 1995). Easterlin et al. (2010), in a more recent and comprehensive research, show that these results also hold for Eastern European countries transiting from socialism to capitalism, as well as for several other developing and developed countries. This study suggests that this null relation holds in the long run, defined as at least 10 years in the study. However, the same study found that income and happiness correlate positively in the short run. Moreover,

(7)

6 estimating the relationship between income and happiness in cross-sectional data from a single country often brings a positive relationship in both developed and developing countries, with a bigger slope in the latter.

Studies that have included relative income variables suggest that well-being is affected by relativities. Additional income may not increase well-being if those in the relevant comparison group also increase their income (Dolan, Peasgood & White, 2008). Moreover, relative income affects SWB through two channels: a comparison to others in a relevant reference group (social comparison) and to a previous income status (adaptation)1_{(Clark, Frijters & Shields, 2008).}

Furthermore, Easterlin (1995) noted that the effect of relative income would explain why cross-sectional data shows that wealthier individuals within a society are happier, but societies are not happier when the average income increases over time. Clark et al. (2008) propose a model to explain these findings. They suggest that, at a point in time, the marginal utility of extra consumption remains always positive because those with higher income enjoy higher status and higher consumption. However, over time, everyone’s income increases; therefore, the utility increases only because of extra consumption and not because of status. Thus, the marginal utility of consumption is positive but approaches zero, and the marginal utility of status always remains positive2_.

Literature about what constitutes a reference group is much more developed in the area of psychology than economics. Some researchers argue that reference groups consist of a wide variety of concepts. Wilkinson (1997) mentions that, when asked, people said they tend to compare themselves to “people like themselves”. Sachs et al. (2012) report that reference groups may change depending on the country. In a representative sample of rural Chinese people, participants said they mainly compare themselves to others in the same village, while in the European Social Survey the most common answer was to mention “colleagues” as a reference group (Helliwell, Layard & Sachs, 2012). Another reference group often mentioned is themselves in the past; this is, how their life satisfaction compares to how they used to be (Deaton, 2008). Other studies find that the reference group can be defined as a group with similar characteristics such as age range, education group, geographical area, etc. (Van Praag, 2011). Moreover, Schwarze and Härpfer (2007) noted that it is

1_{“Adaptation” refers to the fact that individuals get used to their circumstances, so a change in income or}

consumption has only temporary effects.

2_{This is, income becomes irrelevant when countries get richer. Nervetheless, people are always better-off}

(8)

7 reasonable to assume that individuals are affected more by inequality within their own region than by nationwide inequality.

From the previous review done so far, the importance of estimating the relation between income inequality and SWB is clear. Results of several international studies on income inequality have shown mixed results. Alesina et al. (2004) find that, in USA and Europe, the effects of inequality on SWB vary depending on how inequality is interpreted. Therefore, in countries with high mobility, income inequality is perceived as an “opportunity”, while in countries with low mobility, income inequality is found to have a negative impact. Latin American countries have a fairly unequal income distribution, but they tend to be happier than former communist countries, which are more equal. Nevertheless, within Latin America, inequality decreases happiness, because it is perceived as persistent “unfairness” (Graham & Felton, 2006).

Local inequality is rarely addressed in the literature, which shows the need for more research. Schwarze and Härpfer (2007) link life satisfaction data to inequality of the pre- and post-government income distribution at a regional level. Results indicate that Germans are inequality averse, but redistribution does not increase life satisfaction. The authors also state that the position in the income distribution has no effect on life satisfaction, except for those on the lowest incomes. Compared to other individuals, those at the bottom of the income distribution are less satisfied with their lives.

This paper aims to contribute to the literature on the effect of local inequality on SWB. In this study, I use the smaller geographical unit available in data as reference group. Therefore, I will consider that individuals compare themselves with their local community or municipality in terms of relative income/consumption and inequality.

III. Context: The case of Ecuador

Latin America has improved in many developmental goals such as poverty reduction and education coverage in the last few decades. Nevertheless, Latin America remains one of the most unequal regions in the world. This makes the study of inequality and SWB especially relevant in the region, and aims to be a tool for development. In this research I focus on the case of Ecuador, during the years during 2010-20133_.

3_{I chose these years because I use data from 2010 and 2013 in the reseach. One of the assumptions made in}

the analysis is that economic conditions of Ecuador did not present neither big nor abrupt changes during this period. Therefore, it is important to analyze if this condition holds.

(9)

8 Ecuador is the eight biggest economy in Latin America, with its economic resources mainly based on exports of oil, bananas, shrimp, gold, and other primary agricultural products. Ecuadorian economy growth averaged 4.4% between 2006 and 2013, mainly driven by the high price of oil and important flows of foreign investment. The average growth between 2010 and 2013 was 5.4% (see Figure 1 in Appendix). The poverty headcount ratio has shown a remarkable improvement since 2006, reducing from 37.6% to 25.5% in 2013 (Figure 2 in Appendix). The poverty gap at national lines has also shown a reduction, going from 15.3% in 2007 to 9% in 2013. The rural poverty gap index is higher than the urban poverty gap index. However, rural poverty gap has shown an improvement, from 29.1% in 2006 to 16.4% in 2013 (Figure 3 in Appendix).

Overall, Ecuador is a medium income country in Latin America, with a GDP per capita similar to Peru and Dominican Republic4_{in 2013. In terms of inequality, Ecuador is positioned as a}

mid-inequality, mid-income country in the context of Latin America (Figure 5 in Appendix) and presented a moderate reduction of inequality from 2010 to 2013 (Figure 4 in Appendix). Ecuador is also placed below the average life satisfaction in Latin American countries according to Latinobarómetro Survey (Figure 6 in Appendix).

IV. Hypothesis and Empirical Strategy

The general hypothesis is that inequality affects well-being negatively through decreasing marginal utility of consumption, inequality aversion and social distress. I also posit that this effect varies depending on the economic status of the household relative to their reference group.

IV.1 Hypotesis 1

Before presenting the first hypothesis, it is necessary to understand the theoretical reasons for which a change in local inequality might affect SWB. For this, I proceed to explain three potential mechanisms as discussed in the literature.

 Decreasing marginal utility of consumption

A decrease in local inequality, holding average consumption at the Parroquia constant will, at the same time, increase the consumption of the poor and decrease the consumption of the rich. Therefore, a decrease in inequality will have a positive effect in average utility due to a decreasing marginal utility of consumption. This means that the increase of well-being of those on the bottom of the

(10)

9 distribution will overcome the decrease in well-being of the people on top, resulting in a higher average utility. In this analysis, the terms utility and well-being can be consider a synonyms, as previously explained.

Secondly, at the individual level, a concave utility function in consumption represents the choices of a self-interested individual with risk aversion preferences. Therefore, self-interested, risk-averse people value a lower level of uncertainty, which translates into a lower level of inequality (Schwarze & Härpfer, 2007).

In order to consider the decreasing marginal utility as a valid explanation for the effects of inequality on reported SWB, I need to discuss whether the measure of life satisfaction has indeed the same properties as the concept of “utility”. That is, whether consumption or income increases SWB at decreasing rates.

Evidence of U.S. studies shows that, in the short run, the relationship between income and subjective well-being is positive at decreasing rates. Nevertheless, this evidence also states that there is no relation between income and well-being in the long run (Frey & Stutzer, 2002).

In this research I analyze a single country in a cross-section setting. According to literature, I should find a positive and decreasing relation between consumption and well-being. Therefore, the first mechanism is that an increase in local inequality will have a either negative or null effect on SWB depending on the behavior if SWB on consumption in data. This relationship will be studied in section VI.1.

 Inequality aversion: Altruist individuals

The term inequality aversion means that people have preferences for fairness. That is to say, inequality can have a direct and negative effect on an individual’s well-being in a society where people are likely to empathize with the suffering of others, especially if these people belong to their reference group (Hajdu & Hajdu, 2014). According to this theory, people dislike inequality because is directly influenced by the suffering of their poorest neighbors.

In the context of this research, inequality aversion implies that individuals intrinsically prefer a less unequal income distribution irrespective of their own position on the income distribution. The effect of inequality regarding income distribution placement will be discussed in Hypothesis 2.

(11)

10 According to Schwarze and Häpfer (2007), this theory means that people are intrinsically inequality averse and, therefore, inequality enters individual utility function: “individuals are altruistic or prefer a more equal income distribution, which then becomes something of a ‘public good’” (p. 234). Thus, under this theory, higher local inequality would directly and negatively affect SWB.

 Social distress

Another possible mechanism is that local inequality is likely to cause social distress. Social distress happens when an individual perceives some social interactions (such as trust, or political engagement) or societal outputs (such as crime, health or employment) negatively.

Researchers have found that inequality has a negative effect in social interactions and social outcomes. Income inequality have a negative effect on pecuniary crime rates (Scorzafave & Soares, 2009; Brush, 2007), working hours (Bowles & Parks, 2005), health (Pickett & Wilkinson, 2010) (Kawachi, Kennedy, Lochner, & Prothrow-Stith, 1997), and mobility (Wilkinson & Pickett, 2009), amongst others. Therefore, these outcomes will affect negatively SWB indirectly through social distress (Hajdu & Hajdu, 2014).

In this study, I measure how inequality affects SWB. For this purpose, I could differentiate between the three potential mechanisms acting simultaneously. It would be possible to discern whether inequality affects SWB through social distress, or directly through inequality aversion, if I could control for potential societal outcomes that cause distress. However, I do not have the required data to do so. Also, controlling for these variables and including local inequality in the same regression would lead to identification problems and inconsistent estimators because of the inclusion of outcome variables that are also affected by inequality.

The effect of decreasing marginal utility is also difficult to separate from the other two effects. If I observe a positive, but not diminishing marginal utility of consumption, then the negative effects of an increase in inequality on SWB could be attributed only to inequality aversion and social distress. However, if the data shows a positive and decreasing relation between SWB and consumption, then I cannot distinguish of any of the three mechanisms proposed. Therefore, the three mechanisms would suggest a negative effect of inequality on SWB.

(12)

11 Based on the previous argumentation, the first hypothesis is:

H1: An increase of local inequality has a negative impact on subjective well-being.

IV.2 Hypothesis 2

I also want to examine whether the effect of inequality in SWB is higher in disadvantaged households within the reference groups. However, some clarifications are needed to correctly explain the second hypothesis.

Firstly, it is necessary to define the concept of reference group. As mentioned previously in the literature, a reference group is the group of individuals to which households compare in terms of consumption. Reference groups are very diverse (colleagues, co-workers, neighbors, ethnic groups, and others) and change accordingly to the country or age group the individual belongs to.

Schwarze and Häpfer (2007) argue that it is reasonable to assume that individuals are more affected by inequality within their own region than nationwide, and that regional inequality is observed by people at least as good as nationwide inequality. In this study, I want to measure the effect of local inequality; therefore, I chose the municipality or local community where the household resides as the relevant reference group. In Ecuador, this geographic area is named Parroquia.

Another important point to discuss is how households compare to their reference group. Households could compare themselves to the consumption of the Parroquia in two ways: looking at the average consumption of the Parroquia, or looking at their position in the consumption distribution within the Parroquia (quantile distribution or ranking). Here I will introduce these two possible mechanisms:

First, households can compare themselves to the average consumption of the Parroquia. I will refer to this as an “absolute” comparison. In this scenario, an increment in the average consumption of the Parroquia, holding household consumption constant, would reduce the relative position of household consumption within the Parroquia. This would reduce household life satisfaction. Nevertheless, a change in local inequality, keeping average consumption of the Parroquia and household consumption constant, would not affect household placement in the distribution. Therefore, this mechanism is insensitive to changes on inequality and it would be useless to include it in this analysis since the aim is to measure the potential effects of an increase in local inequality.

(13)

12 The second mechanism is a comparison between household consumption and the distribution of the Parroquia’s consumption. Households place themselves in a certain quantile of consumption; if local inequality increases, people of different quantiles would be affected differently by this change. There are many reasons why this is the case. For example, people in a low quantile, living in a

Parroquia with high inequality, would perceive their own income as being “too far” from the top

quantile. Nevertheless, this may not be the case for those households at the top of the distribution in the same Parroquia. This could have detrimental effects in life satisfaction because inequality is perceived more negatively for those in the bottom and increases happiness for those at the top (Kingdon & Knight, 2006; Helliwell, Layard & Sachs, 2012). In this analysis, I will focus on this mechanism.

Thus, the second hypothesis can be described as follows:

H2: Households at bottom of the distribution are more negatively affected by local inequality.

I predict that the SWB of households at the bottom of the consumption distribution are more negatively affected by inequality than households in the top of the distribution. Some evidence from USA and Europe shows that people at the bottom of the income/consumption distribution are negatively affected by inequality. People at the top of the distribution seem to be indifferent or negatively affected depending on the country (Alesina, Di Tella, & MacCulloch, 2004; Schwarze & Härpfer, 2007).

(14)

13

IV.3 Empirical Strategy

To test these hypotheses, I elaborated a design based on two previous studies, Schwarze & Härpfer (2007) and Alesina et al. (2004). I replicated the main design of these papers: an ordered probit with cluster correction of errors, and the main control variables (set of personal characteristics and social characteristics). Both papers use panel data and therefore include time series analysis and fixed effects. The current analysis was developed with a cross-section database, requiring the inclusion of local level characteristics.

The model can be represented as follows:

𝑆𝑊𝐵𝑖𝑗= 𝛽0+ 𝛽1𝑍𝑖𝑗+ 𝛽3𝑋𝑗+ 𝛽4𝐺𝑗+ 𝛽5(𝑇𝑖𝑗∗ 𝐺𝑗) + 𝛽6𝐶𝑖𝑗+ 𝛽7𝑇𝑖𝑗+ 𝛽8𝐶𝑗+ 𝑒𝑖𝑗 (1)

𝑖 = ℎ𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑, 𝑗 = {𝑝𝑟𝑜𝑣𝑖𝑛𝑐𝑖𝑎, 𝑐𝑎𝑛𝑡𝑜𝑛, 𝑝𝑎𝑟𝑟𝑜𝑞𝑢𝑖𝑎},

where 𝑆𝑊𝐵𝑖𝑗 represents subjective well-being reported by the individual 𝑖 of the local area 𝑗, 𝑍𝑖𝑗 is

a vector of socio economic variables, 𝑋𝑗 is a vector of variables at a local unit level, 𝐺𝑗 is the local

inequality measure, 𝑇𝑖𝑗∗ 𝐺𝑗 is the interaction between household quantile of consumption and local

inequality, 𝐶𝑖𝑗 is per capita consumption of the household, 𝑇𝑖𝑗 is the household quantile of

consumption (this variable will be explained in detail on section V), 𝐶𝑗is the average consumption of

the Parroquia, and 𝑒𝑖𝑗 is the error term.

The first hypothesis is tested by the estimator 𝛽4, controlling by a set of local variables. This

parameter measures the net effect of local inequality on SWB that can be attributed (indirectly) to lack of social cohesion or (directly) to inequality aversion or decreasing marginal utility.

The second hypothesis is tested with the estimator 𝛽5. This parameter measures whether the effect

of local inequality on SWB depends on household consumption quantile. I will also pay attention to 𝛽7 to check if the quantile is still relevant after the inclusion of the interaction.

The method used for this model is an ordered probit estimation since SWB (𝑊𝑖𝑗) is a multiple

choice variable. The model also includes cluster standard errors at the local unit, in order to correct for possible correlation in the errors between Parroquias, Cantones or Provincias.

(15)

14

V. Data

For this research I used the Ecuador Poverty Map of 2010. From this database, I used inequality variables (Gini coefficient, Generalized Entropy Index, 90/10 ratio and Atkinson Index) and the average consumption at the level of Parroquia, Canton and Provincia5_{. The poverty map was}

obtained from the National Institute of Statistics and Census (INEC) and World Bank 67_.

The map was created using the 2010 Ecuadorian Census and the Ecuadorian Survey of Living Standards 2013 (ECV). The poverty map was constructed using the “Small Area Estimation Method” that allows for a better estimation of inequality indices, lower standard error, and higher precision than the one obtained directly from household surveys8_{. This makes my own estimations on inequality}

statistically stronger. This is a notorious advantage considering that measurement errors of explanatory variables could lead to inconsistent estimators.

Personal characteristic variables come from ECV 2013. These variables are: sex, age, number of children, area, employment status, health status, marital status and consumption (Further detail on this variables can be seen on Table 2 in the Appendix). Household consumption was calculated following the indications and methodology of INEC9_.

The main dependent variable is SWB. This was measured with the ECV 2013 variable “life satisfaction”. This variable asks people how they rate their satisfaction with their own life in a scale from 1 to 10 (Table 1 in Appendix). Average life satisfaction is 7.6 and 80% of the sample placed themselves between levels 7 and 10 (Figure 7 in Appendix).

It is important to clarify that only the head of the household or spouse reported a life satisfaction level. Therefore, all the personal characteristics included in regressions correspond to the head of the household, which means that neither children nor other members of the household were represented in the study. This is an important distinction that must be accounted for when interpreting the final results.

5_{Some of this Index include diferent specifications. Generalized entropy index is included several times with}

different values for parameters.

6_{Instituto Nacional de Estadística y Censos (INEC)}

7_{Source: INEC-BM Mapa de Pobreza y Desigualdad por consumo 2014}

8_{Further details in the small area estimation method used un the poverty map can be consulted in Molina et al.}

(2015) or directly in Elbers, Lanjouw and Lanjouw (2003)

(16)

15 Local level characteristics, such as total population, population density, total surface, average schooling years and percentage of migrants in the Parroquia, Canton and Provincia, were taken from INEC Census Tabulations. These variables aim to capture join local unit effects that may be affecting SWB in the households.

Finally, I created a variable called “TERCILE CONSUMPTION” that indicates whether per capita consumption of a household is located below the 40%, between 40% and 80% or above the 80% of the richest households of Parroquia, Canton or Provincia they belong to. In order to calculate this, I used Poverty Map estimations of the consumption of the 40% and 80% richest household per level of disaggregation according to the following rule10_:

𝑇𝐸𝑅𝐶𝐼𝐿𝐸 𝐶𝑂𝑁𝑆𝑈𝑀𝑃𝑇𝐼𝑂𝑁𝑖𝑗= { 1 𝑖𝑓 𝑝𝑐 𝑐𝑜𝑛𝑠.𝑖𝑗< 40% 𝑐𝑜𝑛𝑠. 𝑑𝑒𝑐𝑖𝑙𝑒𝑗 2 𝑖𝑓 40% 𝑐𝑜𝑛𝑠. 𝑑𝑒𝑐𝑖𝑙𝑒𝑗≤ 𝑝𝑐 𝑐𝑜𝑛𝑠.𝑖𝑗≤ 80% 𝑐𝑜𝑛𝑠. 𝑑𝑒𝑐𝑖𝑙𝑒𝑗 3 𝑖𝑓 𝑝𝑐 𝑐𝑜𝑛𝑠.𝑖𝑗> 80% 𝑐𝑜𝑛𝑠. 𝑑𝑒𝑐𝑖𝑙𝑒𝑗 (2) 𝑤ℎ𝑒𝑟𝑒 𝑖 = ℎ𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑; 𝑗 = {𝑃𝑎𝑟𝑟𝑜𝑞𝑢𝑖𝑎, 𝐶𝑎𝑛𝑡𝑜𝑛, 𝑃𝑟𝑜𝑣𝑖𝑛𝑐𝑖𝑎}

This variable reflects the relative position of the household within the Parroquia, Canton o Provincia in terms of consumption.

VI. Results

VI.1 Preliminary results

Figure 1 shows a slightly positive relation between Gini decile and Life Satisfaction. I calculated a variable called DECILE GINI. This indicates the decile of Gini where each Parroquia is placed. The positive correlation contradicts the first hypothesis and some of the evidence usually found in literature (as I previously mention in section II). This preliminary result suggests that either Ecuador is a country that perceive inequality as beneficial or that this correlation captures other non-observed effects.

10_{The selection criteria of 40% and 80% of the richest population was based on the symmetry of the data. That}

is, the categories were selected so the fraction of households in each category was fairly similar. Is also important to mention that, depending on the household tercile, the same household could follow in a diferent cathegory in a different level of disagregtion. This because TERCILE CONSUMPTION is a relative measure of consumption.

(17)

16

Figure 1. Gini deciles and Life satisfaction at Parroquia level.

Notes: y axis is scale from 7.3 to 7.9 for a better display of the fitted values; the real scale is from 1 to 10.

Consumption (or income) is usually mentioned as an important determinant of SWB. With the ECV 2013 database I observed that natural logarithms of per capita consumption and life satisfaction have a positive relation, as often mentioned in the literature (Figure 2). I also found that SWB increases at decreasing rates over consumption (Figure 8 in Appendix). This would confirm the assumption of decreasing marginal utility of consumption explained in section IV.1.

It is important to notice that the natural logarithm of per capita consumption and local Gini have a positive relation, as Figure 9 in the Appendix shows. This suggests that consumption and inequality should be analyzed altogether in order to study the causal effect of these variables.

(18)

17

Figure 2: Natural logarithm of per capita consumption household and Life Satisfaction

Hypothesis 2 states that the level of inequality has an effect on SWB that depends on the relative position of household consumption in the Parroquia. In the empirical strategy, this effect is reflected on the interaction between the tercile of consumption and the level of inequality (𝑇𝑖𝑗∗ 𝐺𝑗). Figure 3

shows the relationship between life satisfaction and tercile of consumption by four different levels of inequality measured with the Gini index (low, medium, medium high, high). This figure suggests that the relation between life satisfaction and tercile of consumption is positive at the four levels of inequality. Nevertheless, the slope of the relation is higher at higher levels of inequality.

I also compared the average life satisfaction of households at each tercile of consumption but living in Parroquias with different levels of inequality (from 1 being the “low inequality” to 4 being “high inequality”) (Figure 4). I found that households at the bottom of the consumption distribution (tercile 1) report higher levels of life satisfaction when they live in Parroquias with “low inequality” or “medium inequality” than households in the same tercile, but living in Parroquias with “medium high” and “high” levels of inequality. This difference is not statistically significant (p-value=0.180, T-test)11_{. Households at the second tercile of consumption report higher levels of life satisfaction}

when they live in Parroquias with “medium high” and “high” levels of inequality (p-value=0.000, T-test). Households in the third tercile of consumption also report higher levels of life satisfaction when

11_{In this analisis I made a t-test comparing the average life satistaction of housholds living in Parroquias with}

“low inequality” (average of 7.48) against housholds living in Paarroquias with any of the other inequality levels (average of 7.37).

(19)

18 they live in Parroquias with “medium high” and “high” levels of inequality (p-value=0.000, T-test) and this difference has a higher magnitude than the difference in tercile 2. These preliminary results are in line with the second hypothesis that inequality affects differently to each tercile. Nevertheless, in these graphs the effect of inequality in SWB is positive for terciles 2 and 3.

Figure 3: Relation between life satisfaction and tercile of consumption by level of inequality

Note: The slope of fitted lines increase at every level of inequality, with the exception of the highest level of inequality. The slope of the fitted line for Parroquias with low inequality is 0.16, for medium level is 0.12, for medium high level is 0.34 and for high inequality is 0.32.

(20)

19

Figure 4: Life satisfaction and Gini levels by tercile of consumption.

VI.2 Empirical Results

Empirical results of the main estimation are shown in Table 1. This table presents four progressive estimations consisting of an ordered probit (from the simplest to the most complex model). The main variables of interest are: Gini coefficient, natural logarithm of per capita consumption per household, tercile of consumption, natural logarithm of Parroquial consumption, and the interaction of tercile consumption and local Gini. I will refer to model 4 as the “preferred model”12_{. In this section, I only}

report the results of Parroquia. Nevertheless, same estimations of the preferred model for Canton and Provincia can be seen in detail in Table 4 and Table 6 in Appendix.

Results are consistent with literature and with my hypotheses. Local inequality (GINI) has negative effect on life satisfaction. This effect becomes statistically significant only when all variables are included in the model and after the inclusion of the interaction between tercile and Gini.

As expected, per capita consumption of the household (LN PC CONSUMPTION) increases life satisfaction. Parroquial consumption (LN PC CONSUMPTION PARROQUIA) is not statistically significant as a determinant of SWB. When I discussed the literature on reference groups, I introduced two mechanisms of comparison to the consumption of the local area: comparing with the average

12_{Model 4 is my preferred model because it has a higher Pseudo 𝑅}2_{and lower AIC and BIC than model 1 and}

2. Model 4 is preferred to model 3 because it includes all variables aimed to be measure in the study. Nevertheless, the bayesian information criteria differences between this two models is not big.

(21)

20 consumption of the Parroquia, or looking at their position in the consumption distribution within the

Parroquia. If people compared themselves to the average consumption of the Parroquia, an increase

in average income of the reference group, holding per capita consumption constant, would lead to a reduction of life satisfaction. In the results, Parroquial consumption does not have an impact on SWB and, therefore, this results do not provide evidence for the aforementioned mechanism.

The household's relative comparison with the consumption of the reference group (TERCILE CONSUMPTION) has a significant effect on the first three columns. When I include the interaction between tercile and Gini, TERCILE CONSUMPTION ceases to be significant and the interaction variable becomes significant. This means that the effect of relative consumption position is absorbed by the interactive term.

The effect of the interaction between tercile and Gini (TERCILE CONSUMPTION*GINI) is positive and statistically significant when included in model 4 (p-value = 0.000). This variable can be interpreted as the effect of a change in local inequality on SWB, depending on the tercile of consumption. The positive effect of the coefficient means that the effect of local inequality on life satisfaction becomes less negative for households placed in higher terciles, as I predicted in the section hypothesis. The effect of local inequality on happiness depends on household’s tercile, as suggested in Figure 4.

(22)

21

Table 1: Ordered probit of determinants of Subjective Well-being

(1) (2) (3) (4)

Life Satisfaction INEQULITY PERSONAL PARROQUIA INTERACIVE

INEQUALITY:

GINI -0.124 -0.388 -0.732 -1.632**

(0.750) (0.379) (0.103) (0.031)

TERCILES CONS. * GINI 0.449*

(0.067) CONSUMPTION: LN PC. CONSUMPTION 0.229*** _0.198*** _0.204*** _0.196*** (0.000) (0.000) (0.000) (0.000) TERCILES CONSUMPTION 0.002 0.028 0.028 -0.115 (0.924) (0.162) (0.153) (0.152) LN PC. CONSUMPTION PARROQUIA 0.019 -0.002 0.016 0.023 (0.608) (0.955) (0.751) (0.640) PERSONAL CHARACTERISTICS: WOMEN -0.008 0.011 0.011 (0.716) (0.621) (0.613) AGE -0.000 -0.001 -0.001 (0.906) (0.718) (0.693) AGE2 -0.000 -0.000 -0.000 (0.668) (0.902) (0.917) PARTNER 0.181*** _0.193*** _0.193*** (0.000) (0.000) (0.000) CHILD 0 TO 7 0.028*** _0.028*** _0.028*** (0.004) (0.004) (0.005) RURAL -0.017 0.004 0.006 (0.601) (0.911) (0.852) YEARS EDUCATION 0.019*** _0.019*** _0.018*** (0.000) (0.000) (0.000) EMPLOYED 0.025 0.035 0.036 (0.306) (0.163) (0.146) UNEMPLOYED -0.340*** _-0.332*** _-0.331*** (0.000) (0.000) (0.000) HEALTH -0.078*** _-0.083*** _-0.083*** (0.000) (0.000) (0.000) CRONICLE_HEALTH -0.120*** _-0.113*** _-0.113*** (0.000) (0.000) (0.000) PARROQUIAL VARIABLES: % MIGRANT PARROQUIA -4.336*** _-4.329*** (0.000) (0.000) POPULATION PARROQUIA 0.000*** _0.000*** (0.000) (0.000)

AVERAGE EDUCATION PARROQUIA 0.002 0.003

(0.885) (0.825)

SURFACE PARROQUIA -0.000 -0.000

(0.330) (0.332)

PARROQUIA POPULATION DENSITY -0.000 -0.000

(0.567) (0.561)

N 28576 26650 26456 26456

PSEUDO R2 0.008 0.015 0.016 0.016

AIC 103711.175 95218.092 94436.821 94434.484

BIC 103818.559 95414.665 94674.135 94679.981

Notes: p-values in parentheses: *_{p < 0.1,}**_{p < 0.05,}***_{p < 0.01}

All regressions include cluster standard error at the level of Parroquia. Regression (1) is an ordered probit including only consumption, inequality and poverty variables. Regression (2) also includes personal characteristics of the head of the household. Regression (3) also includes characteristics of the Parroquia. Regression (4) also includes an interactive term of the consumption tercile the household belongs to and the Gini coefficient in the Parroquia. The number of Parroquias remaining in the regressions was 1009 (140 where excluded after merging the Poverty Map and ECV 2013).

(23)

22

The interpretation of the magnitude of each coefficient is not simple in an ordered probit model. Therefore, further analyses must be made in order to correctly interpret the effect of a change in inequality and tercile of consumption on life satisfaction. For this purpose, I have ran a sensitive analysis for each tercile of consumption at the average value of Gini (0.33). Then, I increased the local Gini coefficient by 0.1 units in each tercile and compared the predicted probabilities of being in each level of life satisfaction and their confidence intervals (from 1 to 10) (see Table 6 in Appendix).

When local Gini increases by 0.1 units, households in tercile 1 show a higher probability of reporting low levels of life satisfaction (from 1 to 7) and a lower probability of reporting high levels of life satisfaction (from 8 to 10)13_{. Households in tercile 2 show the same behavior but with a slightly}

smaller variation. However, households in tercile 3 show marginal changes in the probabilities of reporting some levels of life satisfaction, with which I cannot derive strong conclusions (Figure 5).

Personal characteristics were included from column 2 onwards. Being married (PARTNER), having more children in the household (Child 0 to 7) and having more years of education (YEARS EDUCATION) increases life satisfaction. All these coefficient are statistically significant in all models.

On the other hand, being unemployed (UNEMPLOYED), being ill in the last month (HEALTH) and being ill the last year (CRO_HEALTH) decreases life satisfaction. These estimators are also statistically significant in all models.

The variables gender (WOMEN), age (AGE, AGE2), employment (EMPLOYED) and area (RURAL) were not statistically significant. More details about these variables can be found in Table 2 of the Appendix.

Local level characteristics were included from model 3 onwards. The results show that living in a

Parroquia with more population (POPULATION PARROQUIA) increases life satisfaction and

living in a Parroquia with higher percentage of migrant (%MIGRANT PARROQUIA) decreases life satisfaction. This effects are statistically significant. Other variables included in this category do not have a statistically significant effect.

13_{It important to clarify that the average life satisfaction reported in the survey is 7,6. Then, I considered low}

(24)

23

A. Tercile 1

B. Tercile 2

C. Tercile 3

(25)

24 In order to test for robustness between different measures of inequality, I regressed the preferred model with other local inequality indices instead of Gini (Atkinson index, General Entropy index and 90/10 ratio). Results are presented in Table 3 of the Appendix14_.

Local inequality have a negative and significant impact on SWB in 6 out of 7 regressions (all except Generalized Entropy index with a parameter of 2). The interaction between local inequality and tercile is statistically significant in 3 out of the 7 regressions; Gini, Atkinson index (1) and Atkinson index (2). As expected, per capita consumption is significant in all the regressions.

Comparing the effect of different measures of inequality is important because each measure tend to overrepresent different parts of the income distribution in population. The indices that put more weight on the extreme of the distribution (Generalized Entropy 2 and 1, and R90/10) are the least statistically significant15_{. On the other hand, the Atkinson index (2), which put more weight in the}

poorest, and Gini index, which put more weight in the middle class, are the most significant. From these results, it is possible to observe that local inequality is a very robust estimator of SWB.

VI.2.1 Canton and Provincia

As previously explained, this paper does not account for a detailed explanation of the result at the level of Canton and Parroquia. However, it is important to mention that local inequality (LOCAL INEQUALITY) is negative and statistically significant in 2 out of the 7 regressions at Canton level and Provincial level. Also, the interaction between inequality and tercile of consumption (INEQ* TERCILE) was not significant in the preferred model for Canton or Provincia (Table 4 and 5 in Appendix). These results suggest that inequality matters at smaller levels of local disaggregation, providing support to the relevance of the analysis at Parroquial level.

14_{I did the same analisys by Canton and Provincia. This can be seen in detail in Table 4 and 5 in the}

Appendix.

(26)

25

VII. Conclusions, Discussion and Further Research

VII.1

Conclusions

Results indicate that life satisfaction of Ecuadorians is negativly affected by local inequality. This can be explained either because Ecuadorians are inequality averse, there is decreasing marginal utility of consumption, or because higher inequality in the neighbourhood creates social distress. These three effects act in the same direction.

Preliminary regressions show that tercile of consumption might be a significant determinant of life satisfaction. Nevertheless, when I control for the interaction of Gini and tercile of consumption, relative position of consumption (TERCILE) does not show an effect on life satisfaction by itself.

I found that local inequality depends on the tercile of consumption (TERCILE *GINI) positively. This is, higher terciles of consumption make the effect of inequality less negative in life satisfaction.

Sensitivity analysis shows that people in a low relative position of consumption are affected negatively by local inequality. This effect holds only for households in the first two terciles. Richer households (tercile 3) are also negatively affected by inequality, but this effect is too small to be considered as conclusive evidence. This results match with the ones found by Schwarze and Härpfer (2007).

Per capita consumption (LN PC CONSUMPTION) increase life satisfaction and this effect is statistically significant in all regresions. Per capita consumption at Parroquial level (LN PC CONSUMPTION PARROQUIA) does not seems to affect life satisfaction.

When I investigated the impact of inequality on SWB at higher levels of dissagregation, I noticed that the effect of local inequlity is less robust. At the Cantonal level, neither the Gini coeficient nor the interaction variable were significant. Only the Atkison index and ratio 90/10 (out of 7 different measures of inequality) have a significant negative effect. At the Provincial level, the Gini coeficient is significant and the interaction variable is not. Here, only Gini and ratio 90/10 are significant (out of 7 different measures of inequality) and negative. Finally, local inequality at the level of Parroquia is significant (6 out of 7 different measures of inequality) and robust. The fact that we found so many significant estimators of local inequality at the Parroquia level and not at higher levels of dissagregation suggests that people are affected more by the inequality in their closest community than nationwide. This enhaces the relevance of examining local variables to better explain happiness.

(27)

26 VII.2 Discussion and Further Research

The analysis of local level inequality and its effect in SWB is barely developed in economic literature. I contributed to this discussion by testing whether local level consumption inequality is a relevant determinant for SWB.

The analysis at local level has been possible due to poverty maps. This study presents a new use for poverty and inequality maps and shows that the latter are a useful tool of analysis when mixed with household surveys.

Results are robust and in the same direction as the hypothesis. Nevertheless, even when estimators are statistically significant, I observe that changes in life satisfaction are small in magnitude. However, it is important to clarify that big changes in life satisfaction are not often observed. The scale of life satisfaction used in this study makes difficult to observe big changes in this variable16_.

This is supported by the fact that 80% of the sample shows levels of life satisfaction in between 7 and 10, which suggests that Ecuadorians are reluctant to report low levels of life satisfaction.

Being the first study of this kind in Ecuador, at the extent of my knowledge, there is plenty of room for improvement and further research around this topic. One possible improvement would be to replicate this research in a setting where all members of the household report their happiness or life satisfaction and not only the head of the household. This might trigger significant changes, because the head of the household bears particular responsibilities and privileges within household dynamics. Another improvement that is plausible in a shorter time period is to replicate the analyses, but with panel data. It would allow to improve the precision of estimations by including fixed effects who control for a more wide number of possible omitted variables. This analysis was not possible in this study because the Ecuadorian Poverty map of 2001 is not yet comparable with the one of 2010.

For further research, it would be interesting to reproduce this study in other Latin American countries, especially in those where local or global inequality is higher. It would also be interesting to recreate this research using other measures of SWB found in the literature. I believe that SWB and happiness are very complex concepts, and the way these concepts are measured could highly determine the results of the study.

16_{Even when I change per cápita consumption, one of the most significant variables in all regressions, the}

(28)

27 I further propose to investigate the effect of local inequality on SWB including variables that allow to separate the three theories named in Hypothesis 1 in order to not only measure the net effect of local inequality, but also to distinguish between the different mechanisms proposed by literature.

(29)

28

VIII. References

Alesina, A., Di Tella, R., & MacCulloch, R. (2004). Inequality and happiness: are Europeans and Americans different? Journal of public Economics, 2009-2042.

Blanchflower, D., & Oswald, A. (2004). Well-being over time in Britain and the USA. Journal of

Publi Economics, 1359– 1386.

Bowles, S., & Parks, Y. (2005). Emulation, Inequality, and Work Hours: Was Thorsten Veblen Right? The Economic Journal, F397-F412.

Brush, J. (2007). Does income inequality lead to more crime? A comparison of cross-sectional and time-series analyses of United States counties. Economics Letters, 264–268.

Clark, A., Frijters, P., & Shields, M. (2008). Relative Income, Happiness, and Utility: An

Explanation for the Easterlin Paradox and Other Puzzles. Journal of Economic Literature, 95-144.

Deaton, A. (2008). Income, Health and Wellbeing Around the World: Evidence from the Gallup World Poll. J Econ Perspect, 53–72.

Dolan, P., Peasgood, T., & White, M. (2008). Do we really know what makes us happy? A review of the economic literature on the factors associated with subjective well-being. Journal of

Economic Psychology, 94–122.

Easterlin, R. A. (1995). Will raising the incomes of all increase the happiness of all? Journal of

Economic Behavior & Organization, 35-47.

Easterlin, R., McVey, L., Switek, M., & Sawang, J. (2010). The happiness–income paradox revisited. Proceedings of the National Academy of Sciences, 22463-22468.

Frey, B., & Stutzer, A. (2002). What Can Economists Learn from Happiness Research? Journal of

Economic Literature, 402-435.

Graham, C., & Felton, A. (2006). Inequality and happiness: insights from Latin America. The

Journal of Economic Inequality, 4(1), 107-122.

Hajdu, T., & Hajdu, G. (2014). Reduction of Income Inequality and Subjective Well-Being in Europe. Economics Discussion Papers, No 2014-22, Kiel Institute for the World Economy.

(30)

29 INEC. (2006). Condiciones de Vida de los Ecuatorianos. National Bureau of Statistics and

Census.

Kahneman, D., & Deaton, A. (2010). High income improves evaluation of life but not emotional well-being. Proceeding of the national academy of science, 16489-16493.

Kawachi, I., Kennedy, B., Lochner, K., & Prothrow-Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 1491-1498.

Kingdon, G. G., & Knight, J. (2006). Subjective well-being poverty versus income poverty and capabilities poverty. The Journal of Development Studies, 1199-1224.

LATINOBAROMETRO. (2001). Latinobarómetro Report . Santiago, Chile.

Minot, N., Baulch, B., & Epprecht, M. (2003). Poverty and Inequality in Vietnam: Spatial Patterns and geographic determinants. International Food and Policy Research Institute and Institute of

Development Studies.

Pickett, K., & Wilkinson, R. (2010). Inequality: an underacknowledged source of mental illness and distress. The British Journal of Psychiatry, 426-428.

Rözier, J., & Kreeykamp, G. (2013). Income inequality and subjective well-being: A cross-national study on the conditional effects of individual and cross-national characteristics. . Social

indicators research, 1009-1023.

Schwarze, J., & Härpfer, M. (2007). Are people inequality averse, and do they prefer redistribution by the state?: evidence from german longitudinal data on life satisfaction. The

Journal of Socio-Economics, 36(2), 233-249.

Scorzafave, L., & Soares, M. (2009). Income inequality and pecuniary crimes. Economic Letters, 40–42.

Van Praag, B. (2011). Well-being inequality and reference groups: an agenda for new research.

The Journal of Economic Inequality, 111-127.

Wilkinson, R. G. (1997). Comment: income, inequality, and social cohesion. American Journal of

Public Health, 87 (9), 1504-1506.

Wilkinson, R., & Pickett, K. (2009). Income Inequality and Social Dysfunction. Annual Review of

Sociology, 493-511.

(31)

30

IX.

Appendix

Ecuador Time Context: 2006-2013

Figure 1: Ecuador’s GDP growth between 2006 and 2013 (annual %)

Source: World Bank

Figure 2: Poverty Headcount Ratio Ecuador 2006-2013

(32)

31

Figure 3: Rural, Urban and Total Poverty Gap Ecuador 2006-2013

Source: World Bank metadata

Figure 4: Ecuador’s Gini index 2006-2013

(33)

32 Ecuador Regional Context 2013

Figure 5: GDP per capita and Gini Index in Latin America 2013

Source: World Bank metadata

Figure 6: Average Life Satisfaction Latin America 2013

Source: Latinobarómetro 2013

(34)

33

Table 1: Life Satisfaction ECV Ecuador 2013

How do you feel about: your life in general? 1 (completely unhappy) to 10 (completely happy)

Figure 7: ECV 2013 Life Satisfaction

Source: ECV 2013

Figure 8: Per capita consumption and Life Satisfaction 2013

Note: Per capita consumption was trimmer for outliers, leaving individuals under US$1000 monthly out of the graph. Trimmed outliers are 1.99% of the survey.

(35)

34

Figure 9: Ln per capita consumption and Life Satisfaction 2013

Table 2: Description of variables.

Head of household

Personal Characteristics Consumption

Local level

Characteristics Interactive Inequality Index Age (years)

Ln per capita consumption (US$ per

month)† Total Population (habitant per 1000) Consumption Tercile*Inequality Gini Women (dummy: 1 if women, 0 otherwise) Consumption Tercile (1 if Tercile 1, 2 if Tercile 2, 3 if Tercile 3)†† Population Density (habitat per km2) Generalized Entropy Years of total education

(number of years)

Ln average per capita consumption local unit

(US$ per month)†

Surface (km2) Foster–Greer– Thorbecke Employed (dummy: 1 if employed, 0 otherwise) Average schooling years 90/10 ratio Unemployed (dummy: 1 if

unemployed, 0 otherwise) % of Migrants Atkinson

Rural (dummy)

Married (dummy: 1 if

married, 0 otherwise)

Child from 0 to 7 (number of children in household

between o and 7 years)

Health (dummy: 1 if ill in

the last month, 0 otherwise)

Cronical_Health (dummy: 1 if ill for more than 12

moths, 0 otherwise)

Source: ECV 2013

† Source: ECV 2013; †† Source: Variable created from ECV 2013 and INEC Poverty Map 2010

Source: 2010 Census tabulations INEC

Source: Variable created from ECV

2013 and INEC Poverty Map 2010

Source: INEC Poverty map

(36)

35

Table 3: Effect of local inequality, consumption, household and Parroquials characteristics on subjective well-being

(1) (2) (3) (4) (5) (6) (7)

Life Satisfaction GINI GE0 GE1 GE2 ATK1 ATK2 R1090

INEQ. PARROQUIAL -1.632** _-1.276* _-1.130* _-0.618 _-1.621** _-1.113** _-0.060*

(0.031) (0.055) (0.088) (0.153) (0.041) (0.018) (0.054)

TERCILES CONS. * INEQ. PARROQUIAL 0.449* _0.344 _0.340 _0.212 _0.436* _0.277* _0.014 (0.067) (0.108) (0.108) (0.130) (0.090) (0.079) (0.189) LN PC. CONSUMPTION 0.196*** _0.198*** _0.199*** _0.203*** _0.197*** _0.196*** _0.199*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) TERCILES CONSUMPTION -0.115 -0.031 -0.033 -0.022 -0.040 -0.048 -0.032 (0.152) (0.451) (0.419) (0.532) (0.368) (0.321) (0.514) LN PC. PARROQUIA CONSUMPTION 0.023 0.023 0.025 0.025 0.023 0.021 0.022 (0.640) (0.638) (0.617) (0.607) (0.643) (0.673) (0.666) PERSONAL CHARACTERISTICS

Yes yes Yes yes yes Yes yes

PARROQUIAL CHARACTERISTICS

Yes yes Yes yes yes Yes yes

N 26456 26456 26456 26456 26456 26456 26456

PSEUDO R2 0.016 0.016 0.016 0.016 0.016 0.016 0.016

AIC 94434.484 94436.295 94440.334 94444.367 94435.196 94430.146 94432.518

BIC 94679.981 94681.792 94685.831 94689.864 94680.693 94675.643 94678.016

Notes: p-values in parentheses: *_{p < 0.1,}**_{p < 0.05,}***_{p < 0.01.}

All regressions include cluster standard error at the level of Parroquia. All regressions are ordered probit. Regression (1) includes the average Gini coefficient in the Parroquia. Regression (2), (3) and (4) include the average General Entropy Index with the parameters 0, 1 and 2 respectively. Regression (5) and (6) include the Atkinson Index with parameter values 1 and 2 respectively. Regression (7) corresponds to the ratio 90% to 10%. The number of Parroquias remaining in the regressions was 1009 (140 where excluded after merging the Poverty Map and ECV 2013).

(37)

36

Table 4: Effect of local inequality, consumption, household and Cantonal characteristics on subjective well-being

(1) (2) (3) (4) (5) (6) (7)

INEQ. CANTONAL -0.875 -0.659 -0.578 -0.316 -0.860 -0.636* _-0.031*

(0.129) (0.129) (0.182) (0.249) (0.121) (0.079) (0.096)

TERCILES CONS. * INEQ. CANTONAL 0.242 0.171 0.178 0.117 0.227 0.155 0.006 (0.224) (0.264) (0.237) (0.219) (0.243) (0.225) (0.370) LN PC. CONSUMPTION 0.211*** _0.211*** _0.213*** _0.215*** _0.211*** _0.209*** _0.211*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) TERCILES CONSUMPTION -0.062 -0.012 -0.016 -0.013 -0.019 -0.026 -0.008 (0.346) (0.701) (0.608) (0.649) (0.595) (0.527) (0.828) LN PC. CANTON CONSUMPTION 0.025 0.024 0.027 0.030 0.024 0.021 0.018 (0.737) (0.752) (0.714) (0.691) (0.751) (0.782) (0.812) PERSONAL CHARACTERISTICS

yes yes Yes yes yes Yes yes

N 26539 26539 26539 26539 26539 26539 26539

PSEUDO R2 0.017 0.017 0.017 0.017 0.017 0.017 0.017

AIC 94669.666 94669.583 94671.587 94673.007 94669.374 94667.054 94666.634

BIC 94915.257 94915.174 94917.178 94918.598 94914.965 94912.645 94912.226

All regressions include cluster standard error at the level of Canton. All regressions are ordered probit. Regression (1) includes the average Gini coefficient in the Canton. Regression (2), (3) and (4) include the average General Entropy Index with the parameters 0, 1 and 2 respectively. Regression (5) and (6) include the Atkinson Index with parameter values 1 and 2 respectively. Regression (7) corresponds to the ratio 90% to 10%. The number of Cantones remaining in the regressions was 213 (8 where excluded after merging the Poverty Map and ECV 2013).

(38)

37

Table 5: Effect of local inequality, consumption, household and Provincia characteristics on subjective well-being

(1) (2) (3) (4) (5) (6) (7)

INEQ. PROVINCIA -1.300* _-0.842 _-0.799 _-0.415 _-1.148 _-0.839* _-0.035

(0.094) (0.132) (0.170) (0.279) (0.113) (0.078) (0.136)

TERCILES CONS. * INEQ. PROVINCIA -0.137 -0.104 -0.116 -0.078 -0.131 -0.074 -0.004 (0.623) (0.602) (0.580) (0.575) (0.615) (0.661) (0.633) LN PC. CONSUMPTION 0.220*** _0.221*** _0.222*** _0.224*** _0.221*** _0.220*** _0.221*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) TERCILES CONSUMPTION 0.072 0.044 0.048 0.043 0.047 0.047 0.042 (0.496) (0.388) (0.376) (0.358) (0.413) (0.466) (0.428) -0.019 0.001 0.004 0.029 -0.006 -0.010 0.013 LN PC. CANTON CONSUMPTION (0.866) (0.991) (0.970) (0.796) (0.959) (0.930) (0.909) PERSONAL CHARACTERISTICS

N 26650 26650 26650 26650 26650 26650 26650

PSEUDO R2 0.018 0.018 0.018 0.018 0.018 0.018 0.018

AIC 94934.964 94943.892 94946.919 94956.070 94940.941 94935.108 94945.815

BIC 95123.346 95140.465 95143.492 95152.644 95137.514 95131.681 95142.388

All regressions include cluster standard error at the level of Provincia. All regressions are ordered probit. Regression (1) includes the average Gini coefficient in the Provincia. Regression (2), (3) and (4) include the average General Entropy Index with the parameters 0, 1 and 2 respectively. Regression (5) and (6) include the Atkinson Index with parameter values 1 and 2 respectively. Regression (7) corresponds to the ratio 90% to 10%. The number of Provincias included in the regressions was 24.

(39)

38

Table 6: Sensitivity analysis: predicted values of Life satisfaction

Probability of reporting each level of life satisfaction with Gini of 0,33 and 0,43 / Tercile 1

Level of Gini=0,33 Gini=0,43

Life Satisfaction Probability Confidence interval

95% Probability Confidence interval 95% 1 0.0046 0.0037 0.0055 0.0064 0.0040 0.0088 2 0.0045 0.0035 0.0055 0.0060 0.0047 0.0074 3 0.0083 0.0068 0.0097 0.0107 0.0089 0.0125 4 0.0126 0.0111 0.0142 0.0159 0.0139 0.0178 5 0.0713 0.0659 0.0768 0.0849 0.0786 0.0912 6 0.0927 0.0878 0.0975 0.1043 0.0989 0.1096 7 0.2117 0.2031 0.2203 0.2240 0.2152 0.2327 8 0.3124 0.3014 0.3234 0.3045 0.2943 0.3147 9 0.1469 0.1416 0.1521 0.1323 0.1278 0.1369 10 0.1351 0.1211 0.1491 0.1110 0.0861 0.1359