Economics growth and the middle class in an economy in transitions the case of russia

Focusing on equivalent incomes, during the whole 1992–2008 period, it seemsthat while inequality did not change and average income increased, the rich groupmoved closer to the median in

Series Editor: Jacques Silber

Economic Studies in Inequality, Social Exclusion

Social Exclusion and Well-Being

Series editor

Jacques Silber, Ramat Gan, Israel

Department of Economics and Business

Management

Ariel University

Ariel

Israel

ISSN 2364-107X ISSN 2364-1088 (electronic)

Economic Studies in Inequality, Social Exclusion and Well-Being

ISBN 978-3-319-51093-4 ISBN 978-3-319-51094-1 (eBook)

DOI 10.1007/978-3-319-51094-1

Library of Congress Control Number: 2016962034

© Springer International Publishing AG 2017

This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part

of the material is concerned, speci fically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission

or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a speci fic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.

The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made The publisher remains neutral with regard to jurisdictional claims in published maps and institutional af filiations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

I would like to thank Jacques Silber for his constructive comments I also thankRaphael Franck and Grazia Pittau for their comments on earlier versions of some

of the chapters Finally, special thanks to the “Russia Longitudinal Monitoringsurvey, RLMS-HSE”, conducted by the National Research University HigherSchool of Economics and ZAO “Demoscope” together with Carolina PopulationCenter, University of North Carolina at Chapel Hill and the Institute ofSociology RAS, for making these data available

v

1 What Does the Middle Class Refer To? 1

1.1 Importance, Measurement and Characteristics of the Middle Class 1

1.1.1 Importance of the Middle Class 1

1.1.2 Measurement of the Middle Class 3

1.1.3 Characteristics of the Middle Class 5

1.2 The Russian Middle Class 7

2 On the Transition in Russia 9

2.1 Russia in Transition 9

2.2 Data 11

2.2.1 Data description 11

2.2.2 Panel Dataset 13

2.2.3 Summary Statistics 14

3 Distributional Change and What Happened to the Middle Class in Russia 21

3.1 Review of the Literature 22

3.1.1 Relative Distribution 22

3.1.2 Decomposition 23

3.1.3 Relative Polarization 24

3.2 Empirical Results 25

3.2.1 Relative Polarization 25

3.2.2 Decomposition 29

3.3 Conclusions 35

4 Bipolarization and the Middle Class in Russia 37

4.1 Review of the Literature 37

4.1.1 Bipolarization Curves 38

4.1.2 Measures of Bipolarization 40

4.1.3 Extensions of the FW Measure 41

4.1.4 Decomposition of the Bipolarization Measure 42

4.1.5 Inequality and Bipolarization 43

4.1.6 Bipolarization and Mobility 44

4.2 Empirical Results 45

4.2.1 Bipolarization Measures 45

4.2.2 Decomposition of the Bipolarization Measure (FW) by Income Sources 51

4.3 Conclusions 55

5 On Polarization in Russia 57

5.1 Review of the Literature 57

5.1.1 Polarization: Definition and Properties 57

5.1.2 Measuring Polarization with an Arbitrary Number of Poles 58

5.1.3 Decomposition of Polarization Indices 62

5.2 Empirical Results 62

5.2.1 Measuring Polarization 62

5.2.2 Decomposition of Polarization Measures by Income Sources 65

5.3 Conclusions 68

6 The Socio-Economic Characteristics of the Middle Class 71

6.1 Methodology 71

6.2 Empirical Results 74

6.3 Conclusions 92

7 Income Mobility and the Middle Class 93

7.1 Methodology: Income Mobility 93

7.2 Empirical Results 94

7.2.1 Mobility Between Income Groups 94

7.2.2 Income Growth 98

7.2.3 Mobility Within Income Groups 99

7.3 Conclusions 101

8 Concluding Comments 103

Appendices 107

References 113

Figure 2.1 Average incomes (in June 1992 rubles) 14

Figure 2.2 Equivalent incomes 15

Figure 2.3 Ratio of mean to median 15

Figure 2.4 Share of zero-incomes 16

Figure 2.5 Gini index 16

Figure 2.6 Theil entropy measure 16

Figure 2.7 Between-group inequality 17

Figure 2.8 Within-group inequality 17

Figure 2.9 The share of the middle class (equivalent incomes) 18

Figure 3.1 PDF overlays for 1992–1995 25

Figure 3.2 PDF overlays for 1995–1998 25

Figure 3.3 PDF overlays for 1998–2001 26

Figure 3.4 PDF overlays for 2001–2003 26

Figure 3.5 PDF overlays for 2003–2005 27

Figure 3.6 PDF overlays for 2005–2008 27

Figure 3.7 PDF overlays for 1992–2008 29

Figure 3.8 Decomposition of the relative distribution for 1992–1993 29

Figure 3.9 Decomposition of the relative distribution for 1993–1995 30

Figure 3.10 Decomposition of the relative distribution for 1995–1996 31

Figure 3.11 Decomposition of the relative distribution for 1996–1998 31

Figure 3.12 Decomposition of the relative distribution for 1998–2001 32

Figure 3.13 Decomposition of the relative distribution for 2001–2002 32

Figure 3.14 Decomposition of the relative distribution for 2002–2005 33

Figure 3.15 Decomposition of the relative distribution for 2005–2008 33

Figure 3.16 Decomposition of the relative distribution for 1992–2008 33

Figure 3.17 Decomposition of the relative distribution for 1992–1996 and 1996–2005 34

Figure 4.1 First degree bipolarization curve 39

Figure 4.2 Second degree bipolarization curve 39

Figure 4.3 Foster-Wolfson bipolarization measure based

on Lorenz curve 41

Figure 4.4 FW-index 45

Figure 4.5 WT-index 47

Figure 4.6 Distance between mean and median incomes (divided by median) 48

Figure 4.7 PG index 50

Figure 4.8 FW-index for positive incomes 52

Figure 4.9 Gini index for individual incomes 52

Figure 5.1 DER-index, a = 0.5 63

Figure 5.2 Identification component of the DER-index 64

Figure 5.3 DER-index for positive incomes 66

Figure 6.1 Estimated proportions of the groups 77

Figure 6.2 Estimated normalized incomes of the groups 77

Figure 6.3 Age of the main earner 81

Figure 6.4 Households where the main earner is male 81

Figure 6.5 Households in rural area 81

Figure 6.6 Households in Moscow and St Petersburg 82

Figure 6.7 Households in Northern Caucasus 82

Figure 6.8 University graduates 83

Figure 6.9 Pensioners (retired main earners) 83

Figure 6.10 Self-employed main earners 83

Table 2.1 Macro-indicators for 1992–2008 10

Table 2.2 The details of the samples 12

Table 2.3 Number of observations in the panel datasets 13

Table 2.4 Gini index 17

Table 2.5 The share of the population belonged to the middle class 18

Table 3.1 Relative polarization for 1992–2008 28

Table 3.2 Relative polarization for 1992–1996 and 1996–2005 34

Table 4.1 FW-index for individual and household incomes 46

Table 4.2 The FW-index for equivalent incomes 46

Table 4.3 The main significant changes in the FW-index over time 47

Table 4.4 WT-index (θ = 1, r = 0.5) 48

Table 4.5 PGindex for equivalent incomes 49

Table 4.6 The FW-index for positive incomes 51

Table 4.7 The main changes over time in bipolarization for individual positive incomes 52

Table 4.8 The share of the income sources in total income 53

Table 4.9 The Gini index for the various income sources 53

Table 4.10 The FW-index for the various income sources 54

Table 4.11 Decomposition of the FW-index by income sources using the Shapley decomposition procedure 54

Table 4.12 Decomposition of the changes in the FW-index 54

Table 5.1 The DER-index (α = 0.5) 63

Table 5.2 The main significant changes in the DER-index (α = 0.5) over time 64

Table 5.3 The components of the DER-index (α = 0.5) 64

Table 5.4 The DER-index for positive incomes (α = 0.5) 65

Table 5.5 The main changes over time in polarization for individual positive incomes 66

Table 5.6 The DER-index (α = 0.5) for the various income sources 66

Table 5.7 Decomposition of the DER-index (α = 0.5) by income sources using the Shapley decomposition technique 67

Table 5.8 Decomposition of the DER-index (α = 0.5) by income

sources using the Araar (2008) method 67

Table 5.9 Shapley decomposition of changes in the DER-index (α = 0.5) 68

Table 5.10 Decomposition of changes in the DER-index (α = 0.5) using Araar’s (2008) method 68

Table 6.1 The BIC values of the mixtures up to 5 components 74

Table 6.2 Parameters of the components 76

Table 6.3 Socio-economic characteristics of the poorest group 78

Table 6.4 Socio-economic characteristics of the middle class 79

Table 6.5 Socio-economic characteristics of the richest group 80

Table 6.6 Ordered probit estimation 84

Table 6.7 Marginal effects after ordered probit estimation for 1992 86

Table 6.8 Marginal effects after ordered probit estimation for 1995 87

Table 6.9 Marginal effects after ordered probit estimation for 1998 88

Table 6.10 Marginal effects after ordered probit estimation for 2001 89

Table 6.11 Marginal effects after ordered probit estimation for 2004 90

Table 6.12 Marginal effects after ordered probit estimation for 2007 91

Table 7.1 Mobility between the components (percentage of the households in the base year) 95

Table 7.2 Ordered probit estimation for the change in component (decrease, no change, increase) 96

Table 7.3 Average annual equivalent income growth rates per groups 98

Table 7.4 Mobility within the groups (in percentage of the group in 1995) 100

Table 7.5 Estimated proportions of the components for three types of datasets 100

Table B.1 The components of the FW-index 110

This book attempts to study the evolution of the middle class in Russia during thelast decade of the twentieth century and thefirst one of the twenty-first century.This is a critical and even fascinating period because it follows the dissolution

of the USSR in 1991 and covers what is known as the transition period, duringwhich many fundamental reforms were implemented This study focuses on incomeand is based on the RLMS dataset (Russian Longitudinal Monitoring Survey).This book includes eight chapters, most of them taking a somehow differentview of the middle class and implementing a distinctive measurement technique.Chapter 1 presents a survey of the literature on the middle class Chapter 2

intends to provide a description of the economic situation in Russia during thetransition period Although the analysis in all the other chapters is based on thedistribution of incomes, each chapter takes a specific view of the middle class andadopts generally a different methodology to apprehend this concept Thus, while thefocus of both Chaps.4and5is on polarization, the analysis in Chap.4is based onthe idea that there should be a clear link between the middle class and the notion ofbipolarization while that in Chap.5 is centered on the idea that the potential forsocial conflict is related to the degree of polarization But the measures used in theempirical sections of these two chapters are completely different Similarly, whileChap.3 tries, on the basis of the“relative distribution” approach, to discover themain features, in terms of both location and shape, of the change that took placeover time in Russia in the distribution of income, Chapter6, using the so-called

“mixture model”, aims at finding out how many different groups could be derivedfrom the income distribution in Russia and what the main socio-economic anddemographic characteristics of these groups are The goal of Chap.7 is to checkwhether, during the period examined, many individuals moved in and out of themiddle class, and what the determinants of such mobility are Chapter 8 finallysummarizes the mainfindings of this study

Focusing on equivalent incomes, during the whole 1992–2008 period, it seemsthat while inequality did not change and average income increased, the rich groupmoved closer to the median (in relative terms), so that the gap between the poor andrich and the ratio of the mean to the median decreased; thus, bipolarization, when

measured by indices such as those proposed by Foster and Wolfson (FW) or Wangand Tsui (WT), decreased, indicating an increase in the share of the middle class.

A mixture model estimation confirms this trend since it shows an increase in thetotal percentage of the two middle groups from 56% in 1992 to 68% in 2008.Assuming an arbitrary number of poles, between-group inequality which is a proxyfor alienation and within-group inequality which represents a loss of identificationdid not change; as a result, polarization over the whole period did not change.Despite the stability of inequality and polarization and a decrease in some bipo-larization measures, relative polarization estimation shows that the income distri-bution diverged between 1992 and 2008 and the mass of the middle class in themedian-adjusted distribution decreased, moving mostly to the lower quartile Thus,one can conclude that the share of the middle class increased only if the impact ofchanges in the median income is not isolated

It seems therefore that one should be very careful in drawing conclusions based

on traditional bipolarization and polarization indices because they are not able tomake a distinction between the impact on the relative importance of the middleclass of an increase (or decrease) in the average (and median) income and that of avariation in the shape of the income distribution (changes in higher moments of thedistribution) The relative distribution is on the contrary able to make such a veryuseful distinction

Birdsall et al (2000) claim that policymakers should take into consideration theeffects of economic policies on the middle class not only because of the requirement

of a healthy market economy for the active involvement of middle groups, but alsobecause reforms, which are critical to market-based economies in a developing andtransitional world, cannot be sustained over long periods if the middle income stratadoes not grow and if those in the middle perceive themselves as losing as a result ofreforms Thus, thefluctuations in the size of the middle class may be especiallyimportant in economies that are in transition from one economic system to another,such as post-Soviet Russia A growing middle class during the period ofsocio-economic transformations in those countries could be seen as an importantindicator of the effectiveness of reform The role to be played by a strong middleclass should therefore be a serious policy concern in transitional economies

To determine who belongs to the middle class, however, is a complex task,among other reasons because of the scarcity of data sources that would encompassall the potentially relevant variables In addition, as will be clear from the review

of the literature, there is no agreement, among those who have attempted to definethe middle class, about the most important features of the middle class, although theamount of income available is almost always mentioned But even when the mainfocus is on the income level, there is no consensus concerning the critical thresh-olds, those that distinguish the middle class from the poor and from the rich.There is hence a need to be very careful when attempting to assess the size of themiddle class in a given country, tofind out whether its importance grew over time,

to detect its main characteristics or to determine whether the identity of thosebelonging to the middle class does not change or varies a lot over time

Notwithstanding these caveats, the present book attempts to study the evolution

of the middle class in Russia during the last decade of the twentieth century and thefirst one of the twenty-first century This is a critical and even fascinating periodbecause it follows the dissolution of the USSR in 1991 and covers what is known asthe transition period, during which many fundamental reforms were implemented.This study will focus on income and will be based on the RLMS-HSE dataset(Russian Longitudinal Monitoring Survey—Higher School of Economics)

RLMS-HSE1 is a series of nationally representative surveys conducted by theNational Research University Higher School of Economics and ZAO“Demoscope”together with the Carolina Population Center, University of North Carolina atChapel Hill and the Institute of Sociology RAS This dataset should be useful in thecontext of a research on the transition, as it covers the relevant period and monitorsthe effects of reforms on the health and economic welfare of households andindividuals in the Russian Federation A repeated cross-section as well as a lon-gitudinal analysis will be undertaken.

This book includes eight chapters, most of them taking a somehow differentview of the middle class and implementing a distinctive measurement technique.Chapter1is entitled“What Does the Middle Class Refer to?” This introductorychapter presents a survey of the economic and sociological literature on the middleclass and reviews the various definitions which have been proposed to characterizethe middle class It also discusses the potential link between a strong middle classand sustained economic growth

Chapter2(“On the Transition in Russia”) provides a description of the economicsituation in Russia during the transition period, looking at the mainmacro-indicators The chapter describes also the database on which this study of themiddle class in Russia is based

The title of Chap.3is“Distributional Change and What Happened to the MiddleClass in Russia” It intends to analyze changes over time in the relative importance

of the middle class, using data on the distribution of income in Russia between

1992 and 2008 Another method used here to determine who belongs to the middleclass is based on the concept “relative distribution” (see, Morris2 et al 1994;Bernhardt et al 1995; Handcock and Morris 1998, 1999) This nonparametricmethod allows a comparison of two distributions in terms of their ratio Theadvantage of the non-parametric estimation is that, unlike the parametric approach,which assumes that the data are drawn from a known distribution, this proceduredoes not require that the datafit a parametric distribution In the second part of thechapter, a decomposition technique is implemented that enables one to make adistinction between the impact of changes in the median income and in the shape

of the distribution of income

The purpose of Chap.4(“Bipolarization and the Middle Class in Russia”) is toanalyze the determinants of changes in income bipolarization in Russia between

1992 and 2008 A survey of the literature on bipolarization measurement isfirstpresented and the link between this concept and that of the middle class isexplained The empirical part givesfirst the value of various measures of the degree

of bipolarization in Russia during the 1992–2008 period Then, one of thesemeasures is decomposed, using the so-called Shapley decomposition procedure

1 Available at http://www.cpc.unc.edu/projects/rlms

2 Actually, the aspects of the relative PDF and CDF as a basis for the comparison analysis were examined earlier by, for example, Parzen (1977 and 1992), Eubank et al (1987), Cwik and Mielniczuk (1989 and 1993).

(see Shorrocks 1999), in order to estimate the contribution of different incomesources to the change that took place over time in bipolarization.

In Chap.5(“On Polarization in Russia”) the concept of polarization, rather thanthat of bipolarization, is described This concept is different from that of bipolar-ization,first because it is assumed to capture the formation of any arbitrary number

of local poles in the income distribution (see, Duclos, Esteban and Ray 2004) Butthis alternative approach is also relevant because it is grounded on different theo-retical premises, namely the notions of “identification” and “alienation”(see, Esteban and Ray 1994) The idea is that polarization is positively related to the

“alienation” that individuals or groups feel one for another, but also to the sense ofwithin-group “identification” Since “alienation” and “identification” are assumed

to be critical determinants of the potential for social conflict, this approachemphasizes the possible social implications of a“polarized” income distribution Inaddition it has the ability to estimate the degree of polarization for the total pop-ulation without having to divide the sample into an a priori defined number ofgroups The chapter presents estimates of the DER (Duclos, Esteban and Ray)polarization index in Russia during the 1992–2008 period This index is thenbroken down on the basis of both the so-called Shapley decomposition procedureand the method proposed by Araar (2008) The goal of this decomposition is again

to determine the contribution of the various income sources to the change inpolarization over time

The purpose of Chap 6 (“The Socio-Economic Characteristics of the MiddleClass”) is to find out whether the Russian society is stratified in groups which can

be defined on the basis of household incomes More precisely, the chapter aims atfinding out how many groups may be identified and what their mainsocio-economic and demographic characteristics are The so-called mixture modelmethod is used to identify the groups This is a semi-parametric method whichenables one to model unknown distributional shapes The advantage of thisapproach is that it allows representing sub-populations (income groups in our case)and parameters of their densities without having to define in advance the numberand characteristics of these groups

The RLMS data may be used to conduct a cross-section as well as a longitudinalanalysis Thus, when income groups are identified and all the individuals orhouseholds are allocated to these groups, the panel dataset allows one to examinemobility between and within groups over time This is precisely the goal of Chap.7

(“Income Mobility and the Middle Class”), whose focus is on the relationshipbetween the middle class, individual income growth and mobility Its main purpose

is to examine the impact of some household characteristics on the probability tomove from one component to another More specifically, an attempt is made todetermine the factors (like age, gender, locality, education and work status) whichaffect the probability that a household that belonged to the middle class in thefirstpost-Soviet years would remain in the middle class The chapter examines also thefactors which affect the probability that a poor household manages to improve itseconomic situation and enter the middle class or that a rich one moves down to themiddle class The chapter tries also to shed some light on the relative income

growth of the middle class, as compared with that of people belonging to the highand low classes, and to test which group benefits the most from income growth,using both an anonymous and a non-anonymous approach.

Concluding comments (Chap.8), based on the results presented in the variouschapters, summarize finally the main findings of this study This study shouldcontribute to a better assessment of the main factors affecting the formation of asizable middle class in Russia and of the determinants of the changes in its relativesize that followed the implementation of economic reforms This book may alsohelp better understanding the effect of various socio-demographic variables ondisparities in individual income growth during the transition period and the impactthese characteristics have on the probability to move along the income distribution

What Does the Middle Class Refer To?

This chapter presents a survey of the economic and sociological literature on themiddle class and reviews the various definitions which have been proposed tocharacterize the middle class It also discusses the potential link between a strongmiddle class and sustained economic growth The chapter ends with a brief review

of the literature on the middle class in Russia during the transition period

of the Middle Class

1.1.1 Importance of the Middle Class

There is some evidence that the share of the middle class in a population is tively correlated with political stability, economic growth and the quality of stateinstitutions Thus, a sizable middle class is an important factor in economicdevelopment (Thurow 1984; Foster and Wolfson 1992 and 2010; Landes 1998;Easterly 2001; Birdsall et al 2000; Pressman 2007; Birsdall 2007b and 2010;Boushey and Hersh 2012, among others)

posi-Easterly (2001) found that an increase in the share of the middle class is ciated with a rise in per capita income and that an increase in the share of the middleincome increases the growth rate In addition, a greater income share of the middleclass leads to better health outcomes, higher levels of publicly provided healthservices and higher levels of political rights and civil liberties, raisesfinancial depthand reduces consumer price inflation

asso-© Springer International Publishing AG 2017

Z Nissanov, Economic Growth and the Middle Class in an Economy

in Transition, Economic Studies in Inequality, Social Exclusion

1

Birdsall et al (2000) point out that the political support and economic pation of the middle class are critical to market-driven economic growth and, thus,

partici-to poverty reduction in the long run Furthermore, Birdsall (2007a) links a small andweak middle-income group to weak state institutions which are the main problem ofnon-sustainable growth

Pressman (2007) claims that a large and vibrant middle class contributes toeconomic growth, as well as to social and political stability, since it helpsdemocracy flourish, mitigates class warfare and may be necessary for goodmacroeconomic performance

Amoranto et al (2010) considered that the middle class is “important in itsdemand for better goods and services and keeping governments accountable”.Loayza et al (2012), using a cross-country panel dataset containing information

on 128 countries, found that the increase in the size of the middle class wasaccompanied by more activity in social policy on health and education and byimprovement in the quality of governance regarding democratic participation and

official corruption

Bussolo et al (2014) point out that the expansion in the size of the middle class

in developing countries can lead to a shift in demand towards more globalizationsupportive policies and is likely to promote policy goals such as improved trans-parency, intensified anticorruption efforts, and demand for a more open society andcleaner environment

Boushey and Hersh (2012) claim that a strong middle class is important foreconomic growth in the following ways: (i) it promotes the development of humancapital and a well-educated population; (ii) it creates a stable source of demand forgoods and services; (iii) it incubates the next generation of entrepreneurs, andfinally, (iv) it supports inclusive political and economic institutions, and, thus,underpin economic growth

These empirical regularities evidently raise questions regarding the conditionswhich favor the emergence of the middle class

Berkowitz and Jackson (2005) point out that an equitable income distribution isconductive to the formation a powerful middle class because it enables the exis-tence of institutions that are critical for market economies

Birdsall (2007b and 2010) defined inclusive growth as growth which builds amiddle class and implies an increase in the share of the middle class (implying thatsome people exit poverty) and the proportion of total income they command(implying gains at the“expense” of either the initially poor or the initially rich).According to Pressman (2007), macroeconomic conditions might affect the size

of the middle class More precisely, while economic expansion, which creates newjobs and generate higher wages, enables more households to earn enough money toenter the middle class, recession can throw people out of work, decrease wages and,thus, reduce the size of the middle class Also important is governments’ spending

on transfer payments directed toward low- and middle-income families, includingnegative taxes to workers This leads to greater overall income equality and,therefore, raises the share of the middle class In addition, Pressman (2007) foundthat liberal welfare regimes tend to have the smallest middle classes and the largest

declines in size of the middle class over time Conversely, social democratic statestend to have the largest middle classes; moreover, these countries managed tomaintain the size of their middle classes despite adverse economic circumstances.

1.1.2 Measurement of the Middle Class

As pointed out by Pressman (2007), there are three meanings of the middle class:sociological, economic and self-identification These approaches were described byLevy and Michel (1983) as follows According to thefirst one (sociological), themiddle class is measured by a set of behavioral characteristics and socio-economiccriteria Thus, the middle class include individuals“who have achieved a certaineducational level, whose jobs have a certain level of social status, and who have aparticular set of values and attitudes” The economic meaning implies having anincome level that is somewhere in the middle of the income distribution The lastone (self-identification) is the most problematic since it allows people to identifythemselves as the middle class even if they are not

Foster and Wolfson (1992) suggested the following steps in measuring themiddle class: (i) choosing a“space” (individual/family/household income, salary,expenditure etc in “income-space” or “people-space”); (ii) defining the middle(median or mean income, the individual at the 50th percentile); (iii)fixing the rangearound the middle (identifying the middle class by fixing a percentage intervalabove and below median or mean); (iv) aggregating the data

There is however no agreement on how the middle class should be defined Evenwhen the main focus is on the income level, there is no consensus concerning thecritical thresholds, those that distinguish the middle class from the poor and fromthe rich Various approaches and definitions have been used to measure the size ofthe middle-income group Needless to say, different definitions lead to differentresults

Some researchers proposed relative measures based on the“income space” Inthis case, the thresholds of the middle class are multiples of the median income (orconsumption) or based on certain income quintiles The advantage of these relativemeasures is that they are sensitive to changes in the income distribution across andwithin countries and over time

While Thurow (1984) defined the middle class as those with incomes between

75 and 125% of the median income, Lawrence (1984) set the middle-class brackets

at approximately 66% and 132% of men’s median weekly earnings Blackburn andBloom (1985) broadened the middle income range to 60–225% of the median andDavis and Huston (1992) defined the middle class as including those familieswhose incomes are between 0.5 and 1.5 times the current-year median incomes.Some studies used as a specific share of the income distribution to define themiddle class For instance, Barro (1999) and Easterly (2001) computed the share ofthe middle class as the three middle quintiles

Conversely, other studies rely on an absolute measurement, which defines theincomes corresponding to the upper and lower bound of the middle class, usually interms of purchasing power parity (PPP) Milanovic and Yitzhaki (2002), forexample, defined the world middle class as all those with an annual income higherthan 3470$ and lower than 8000$ in the PPP terms.

Other studies in this line of research include Banerjee and Duflo (2008), wherethe middle class is defined (in the context of developing countries) as the house-holds whose daily per capita expenditures valued at purchasing power parity variedeither between 2$ (the international poverty line) and 4$ or between 6$ and 10$.Ravallion (2009) proposed a definition close to that of Banerjee and Duflo (2008),identifying the developing world’s middle class as those living between 2$ and 13$PPP a day Similarly, Kharas (2010) considered as global middle class all house-holds with daily per capita incomes between 10$ and 100$ in PPP and in the report

of the Asian Development Bank (2010) the middle class is defined as those withdaily per capita consumption expenditures of 2$–20$ PPP

Torche and Lopez-Calva (2012, p 11) also identified the middle class as thehouseholds which are“above the threshold of absolute economic deprivation, as

defined by the poverty line, but below the threshold that ensures virtually no risk offalling into poverty”

Birdsall (2007b) combined relative and absolute measures and identified themiddle class as people at or above the equivalent of $10 day, and at or below the90th (95th in the developing world according to Birdsall 2010) percentile of theincome distribution in their own country

Another possibility to identify the middle group in an income distribution is toapply the semi-parametric“mixture model” approach (Pittau and Zelli 2006; Pittau

et al 2010; García-Fernández et al 2011)

Another approach to defining the middle class is based on the “people-space” asexemplified by Levy (1987) In this case, the size of the middle class is fixed andcan neither shrink nor expand The focus is then on the evolution of the incomeshare, in contrast to the “income-space” where the population size is the mainconcern (Atkinson and Brandolini 2013) Levy (1987) defined the middle as the50th percentile and the range from the 20th to the 80th percentile is identified as themiddle class

When examining the middle class, Atkinson and Brandolini (2013) set limits inboth the people and income space They argued that although it is normal amongeconomists to think of classes as income groupings, the concept of“class” requiresthe examination of some other dimensions beyond income, like property and theposition in the labor market

Some other researchers suggested that the definition of the middle class shouldinclude characteristics other than income For instance, Wheary et al (2007),focusing on American families, devised a set of formulas for measuring five keybenchmarks of economic security in terms of assets, education, ability to meetessential expenses, health care access and housing costs If three or more of thefactors in a family’s profile supported financial security, they considered that family

as part of the middle class

The Pew Research Center (2008, 2016), explored the middle class in Americaand, in addition to the households included in the middle-income group, looked also

at those who self-identify as the middle class It claimed that a class should be

defined not only in terms of incomes but also reflect “a state of mind”, so that “itcould be a matter of self-identification”

1.1.3 Characteristics of the Middle Class

Kenny (2011) writes that it is hard tofind a set of characteristics that can tently and uniquely identify the middle class across countries and time However,based on different definitions, he concluded that the middle class members are thosewho are “confident of not slipping into deprivation”, “secure in their assets” and

consis-“can spend on things other than basic necessities”

Similarly, the Australian Government report (2007, p 4) defines the globalmiddle class as“those who have scope for discretionary expenditure over and abovethe basic necessities of life such as food, clothing and shelter; but who, at the sametime, face some constraint on that expenditure Members of this group may, forexample have access to items such as consumer electronics, cars, the possibility ofowning their own housing, leisure travel– whether domestically or internationally –and, perhaps most importantly, the ability to invest in higher levels of education forthe next generation At the same time, their access to these things is, in general,restricted by some significant budgetary constraint – unlike the ‘rich’”

Following Birdsall et al (2000), information on occupation and education canprovide a basis for defining the middle class “that has been broadly understood andreasonably comparable across countries (even given large differences in averageincomes among countries)” They however note that a high education level thatprovided opportunities in the past is“no longer sufficient for upward occupationalmobility” More precisely, while in the 1960–1970s a secondary education wassufficient to attain the middle-class living standard, by the 1990s “it neither guar-anteed a well-paying job nor protection from falling into poverty”

In developed countries, it is usually considered (see Hayashi 2007) that the maincharacteristics of the middle class are“economic independence, strong professionalorientation and a civic sense of values” In addition, the middle class consists ofsmall entrepreneurs, lawyers, medical doctors, engineers, scholars and urbanwhite-collar workers

Furthermore, Amoranto et al (2010) argued that the degree of political activismdriven by the middle class is greater than that of the upper and lower classes Theyfound that the middle class is related to higher education, more skilled and stablejobs, and the ability to save

Wheary et al (2007), focusing on American families, found that 65% of therespondents belonging to the middle class were married; 76% were “white”(non-Hispanic); and 85% lived in urban areas The average age was 44 years oldand the average family size was 2.8.

In developing countries, the characteristics of the middle class are somewhatdifferent Birdsall (2010) found that the middle class in most developing countrieshas close to or above 10 years of education This average is below that of the OECDcountries where a high school education is a minimum for a middle class status Thesituation is even worse if the middle class is defined as those between $2 and 10$,which is the definition adopted by Banerjee and Duflo (2008) In that case, theeducational levels of the household heads are lower than or equal to the averageeducational level of their countries Concerning the employee status, Birdsall (2010)reported a small distinction between the poor and the middle class, at least in thecases of Turkey and the Dominican Republic, and noted that the households in anincome range between $2 and 10$“enjoy regular if low-wage participation in theformal sector”

Banerjee and Duflo (2008) claimed that “nothing seems more middle class thanthe fact of having a steady well-paying job” and pointed out that members of themiddle-class prefer having “an income coming in every month and not just theincome itself” The authors found many entrepreneurs among the middle class, theyhowever argued that it is mostly because middle-class members are still relativelypoor and“every little bit helps” and “if they could only find the right salaried job,they might be quite content to shut their business down” Banerjee and Duflo (2008)also found that the households belonging to the middle class have smaller familiesand fewer children and they spend much more than the poor on education and health.According to the report by the Asian Development Bank (2010, p 27), themiddle class in Asia, compared to the poor, is“less connected to agriculture, lesslikely to own land and less likely to be wage laborers It is much more likely to holdsalaried jobs, has a greater propensity for migration, a higher propensity to seekmore expensive medical care when ill, and has fewer children and invests more inhealth, nutrition, and schooling The middle class is also better educated, and moregeographically concentrated (in urban areas or along coasts)”

Torche and Lopez-Calva (2012) investigated the Latin American middle class,taking Chile and Mexico as case studies, and found that a higher social class isassociated with a higher education level and a higher occupational status Theypointed out that, unlike the US where there is stratification in marriage rates andfemale-headed households, the proportion of female-headed and married-headedhouseholds in Chile and Mexico hardly varies across classes; and, unlike somedeveloping countries, such as South Africa, the demographic factors do not matter.The report by the Asian Development Bank (2010) concluded that the two mainfactors in the creation and sustenance of a middle class are higher education andstable and secure well-paid jobs

1.2 The Russian Middle Class

Remington (2010) points out that the Russian leadership (as well as the Americanone) recognizes that“the development of a larger middle class would be beneficial:

by reducing the polarization between rich and poor, it would improve the provision

of growth-enhancing public goods, reduce social tension and increase stability, andwould make government more effective and accountable” Thus, an increase of themiddle class is a serious policy concern in modern Russia

The Russian middle class estimations vary from 2% (Shkaratan et al 2003) to80% (according to self-identification criteria) of the country’s population Thereason for such a discrepancy is that researchers investigating this issue use dif-ferent definitions of the middle class Bobkov (2014) writes that the lack of auniform acceptable definition of the middle class does not allow to evaluate the realsize of its group in Russia, and, thus, makes it difficult to formulate a policy aimed

at increasing the share of the middle class Moiseev (2012) points out that due todifferences in approaches, Tichonova and Mareeva (2009) divided the researchersinto different groups; at one end there are those who deny the existence of themiddle class in Russia, at the other those who insist on the presence of the sizablemiddle class

Based on the material position (income/consumption and housing criteria), thesize of the Russian middle class in 2008 was in the range of 15% (Rosgosstrach andLevada Center) to 19% (Bobkov 2014)

Most of the studies conducted by Russian researchers however relied on asociological approach where income is not considered as the main indicator(Trusova 2001; Maleva 2002 and 2008; Carnegie Moscow Center 2003; Avramova2008; Tichonova and Gorunova 2008; Maleva and Ovcharova 2009; Tichonova2009; Beliaeva 2011) They argue that the middle class cannot be defined by onecriterion and mostly use Max Weber’s approach that assumes a multidimensional

definition of the middle class, including all the three “meanings” described by Levyand Michel (1983) mentioned above

The Russian Academy of Sciences’ (RAS), Independent Institute for SocialPolicy which is one of the leading academic institutions in Russia in thefield ofmiddle class analysis, uses the following criteria: the income level (equal to orhigher than the median for a given region), the educational level (at leastsecondary-special), the work characteristics (not physical) and self-identification.According to their report (2014), the Russian middle class increased between 2003and 2008 from 29% to 34% Then, due to the economic crisis, it decreased but since

2010 it increased again and in 2014 reached 42% of the total population

There are a few studies that used relative measures based on the“income space”.For example, Birdsall et al (2000) suggested as a measure of the middle class theshare of households with per capita income in the range of 75%–125% of themedian household per capita income They thus found that the Russian middle class

increased by 6.4% between 1992 and 1995 and decreased by 7% between 1995 and

19971reaching 28.6% in 1997 They highlighted a dramatic shrinking of the middleclass in Russia during the years 1995–1997, particularly as far as their income share

is concerned, despite the fact that in the early 1990s there was already a relativelysmall middle-income

Another study is by Lefranc (2012) who defined the middle class by threedifferent ranges: 85–130%, 75–150% and 50–200% of the median equivalentincome He found that the share of the Russian middle class remained relativelystable between 1995 and 2000 and then declined sharply between 2000 and 2005,whatever the limits of the range selected Then the size of the middle class grewsignificantly in 2008 and then again in 2010 While in 1995 the proportion ofpeople whose income was in the range between 85% and 130% of median incomewas 23.4%, in 2005 it was only 18.6% of the total population During the secondsub-period (2005–2010), however, the opposite occurred: the share of the middlegroup increased from 18.6% to 29.6% These decreasing and increasing trends wereconfirmed by the results obtained using the Foster and Wolfson bi-polarizationmeasure

According to official information, in 2007, the size of the middle class as defined

by the Economic Development and Trade Ministry was equal to 20% (RIA Novosti,13/06/2007) In addition, the results provided by the Ministry of EconomicDevelopment of the Russian Federation in 2012 (RBC, 25/04/2012) show that 22%

of the Russian population belongs to the middle class The middle class was defined

in this case as a population with a monthly household per capita income at least sixtimes higher than the subsistence wage and owning a car, housing and the possi-bility of partiallyfinancing education, health, and travel abroad

Concerning the characteristics, Grigoryev (2012), who divided the Russianmiddle class into three groups, claims that the Russian lower middle class is similar

to the Western one in terms of education but it is much poorer than the other twosections of the Russian middle class and usually does not have property and sav-ings, as does its Western analogue The lower middle class mostly consists ofsecretaries, sales people and desk clerks The middle-middle class consists ofpeople with a higher education who have a job, aflat and a car Finally, the Russianupper middle looks like the Western one It includes those with a high-level income(top-managers, bureaucrats and a small part of the intellectuals) who have aflat, asummer house and at least two cars

Remington (2010) however points out that half or more of the middle classindividuals in Russia depend on the state for their livelihoods

The main characteristics of the Russian middle class can be described, in linewith Ovcharova (2012), as a group of individuals who are younger and bettereducated than the average Russian population and who mostly live in large cities

1 Birdsall et al (2000) wrote that they used the RLMS dataset for 1997/1998 However there is no RLMS data for 1997 and it can only be assumed that they used the RLMS round 8 for 1998 Thus, the changes which are reported for the 1995 –1997 period actually correspond the 1995–1998 period.

Chapter 2

On the Transition in Russia

This chapter provides a description of the economic situation in Russia during thetransition period, looking at the main macro-indicators The chapter describes alsothe database used in this study

Russia differs from most transition economies because of its large geographicalsize, ethnic diversity and formally federalist structure (Solanko 2006) Russia isdivided into 89 sub-national jurisdictions (that vary in size, composition andnomenclature), each of which relates to one of seven federal districts: Central, FarEast, North Caucasus, Northwest, Siberia, Urals, and Volga The jurisdictionsinclude 21 republics, 49 oblasts (provinces), six territories, 10 autonomous regions,one autonomous oblast, and two cities (Moscow and St Petersburg) with separateoblast status (FRD 2006) Its population decreased from 148.3 million in 1991 to

142 million in 2008 (Rosstat).2

The transition period from planned to market economy in Russia started with thedissolution of the USSR in 1991 The period that followed was marked by twoprofound crises: the industrial collapse and hyper-inflation in 1992 and the financialcollapse in 1998 According to the FRD (2006) publication, during PresidentYeltsin’s tenure (1992–2000) there was “a chaotic transformation that ended thedominance of communism and brought irregular reforms in the economic, political,and social realms… Economic reform was undermined by corruption and publicsuspicion as Russia nominally moved toward a free-market system Judicial reformwas piecemeal and ineffective” In addition, in the mid-1990s a government

1 Note that the review covers the research period only (1992 –2008) and does not describe the current situation in Russia.

2 Available at http://www.gks.ru/free_doc/new_site/population/demo/demo11.htm

© Springer International Publishing AG 2017

Z Nissanov, Economic Growth and the Middle Class in an Economy

in Transition, Economic Studies in Inequality, Social Exclusion

9

privatization plan (aimed at encouraging private enterprise) was “undermined bycorruption, which concentrated significant economic resources in the hands of awell-connected elite rather than effecting true redistribution” (FRD 2006, p 8).Table2.1gives the main macro-indicators for the period 1992–2008, published

by the European Bank for Reconstruction and Development (EBRD TransitionReports 2000, 2002, 2009 and 2010).3

It can be seen that in the first post-Soviet years, hyperinflation was a majoreconomic problem: its annual rate reached 2,500% in 1992, after price stabilization

it decreased to 11% in 1997 but in 1998 inflation rate rose to 84.5% due to thefinancial collapse and subsequent currency devaluation (FRD 2006)

Konstantinova-Vernon (2002) points out that in thefirst post-Soviet years, 70%

of the country’s productive capital was privatized She writes “in 1992–1996,per-capita real GDP in Russia declined by 8.5% per year on average.Unemployment, essentially nonexistent under communism, rose to 11.7% The rate

of poverty at least tripled to 30%; the extent of human suffering is reflected in thesimultaneous decline in life expectancy, increase in suicide rates and decrease inbirth rates By the year 2000, some positive trends emerged: per-capita real GDPwas rising at a rate of 8.3% annually, unemployment was declining, poverty rates

Table 2.1 Macro-indicators for 1992–2008

GDP growth

(percentage

change)

Unemployment rate (percentage

of labor force)

In flation rate (end year)

Interest rate

Exchange rate (roubles per $)

Population (in millions)

were stable, inflation had abated, life expectancy was increasing, suicide rates weredeclining, and birth rates were stable, although the standard of living had not yetreturned to the Soviet levels” (Konstantinova-Vernon 2002, p 2) In addition, sheclaims that the transition affected the labor market through an increase in wageinequality and a decline in the relative wages of women and older workers.Most of the studies devoted to this transition period (Lavrovsky 1999; Hansonand Bradshaw 2000; Westlund et al 2000; Guan 2003; Galbraith et al 2003;Solanko 2006 among others) agree that it has been characterized by a rapidlygrowing economic inequality among Russia’s regions This suggests that there is astrong geographic element in the rising stratification that was observed In otherwords, choice of place mattered more than choice of occupation.

This dataset is useful in the context of a research on transition, as it covers therelevant period and monitors the effects of reforms on the health and economicwelfare of households and individuals in the Russian Federation These effects aremeasured by a variety of means: detailed monitoring of individuals’ health status,incomes, income sources, expenditures, state subsidies, service utilization and acollection of relevant community-level data, including region-specific prices andcommunity infrastructure data

The data have been collected annually since 1992 (with two exceptions: 1997and 1999) and the survey has been conducted in two phases, each representing aseparate panel In phase I (1992–1994 [round 4]), data were collected amongapproximately 6000 households; phase II (1994 [round 5]–2008) covered about

4000 households The RLMS dataset includes information at both the individualand household levels and it enables two types of analyses: repeated cross-sectionand longitudinal or “panel” Table2.2 presents information on the original(cross-section) samples and the number of observations that were used in thisresearch, i.e the number of recipients who reported their incomes

4 More information can be found at http://www.cpc.unc.edu/projects/rlms-hse

It should be noted that the results of the current research are not comparable tosome other studies, which used the same dataset, because of the following reasons:

• In 1992 and 1993 the RLMS conducted two rounds of questionnaires: rounds 1and 2 for 1992 and rounds 3 and 4 for 1993 In order to maintain consistencywith the following periods (1994–2008), we use the rounds that were conductedduring the second half of year, i.e round 3 for 1992 and round 5 for 1993 Thismight lead to some discrepancies between ourfindings and those of studies thatuse other rounds than the ones we use

• Some researchers used a regional CPI deflator to compute real incomes In thecurrent research the CPI at the national level was used, because (see the RLMSwebsite5),“the RLMS sample has NOT been designed to be regionally repre-sentative, so that the researcher is cautioned not to interpret the data at theregional level”.6

All the incomes used in the research are weighted total incomes (from all sourcesincluding transfers) net of taxes, expressed in June 1992 rubles using the nationalmonthly CPI deflator published by Rosstat.7These weights are essential to obtain

an unbiased estimation since “in RLMS data, the household characteristics that

Table 2.2 The details of the samples

Year Round Survey schedule Number of householdsa Individual observations

7 Available at http://www.gks.ru/wps/wcm/connect/rosstat_main/rosstat/en/main/

explain the greatest variation in weights are the geographic region and theurban/rural character of the civil division in which the dwelling is located Variation

in individual weights will reflect the geographic effects for households as well asdifferentials due to post-stratification of the sample by major geographic regions,age, and sex”8(see the RMLS website)

The OECD equivalence scale9was used to adjust household incomes for ferent household sizes and to obtain equivalent incomes The analysis will mostlyfocus on equivalent incomes; in some cases, however, results for total householdand individual incomes will also be shown

dif-The following statistical software will be used for the empirical analysis: theR-program (Chaps.3,6and 7) and Stata (Chaps.2,4–6)

2.2.2 Panel Dataset

As mentioned previously, the RLMS-HSE enables a longitudinal (panel) analysis.This type of dataset will be used in Chap 7 Table2.3 gives the number ofobservations used in the between-groups analysis The panel datasets in this casewere obtained from the original cross-section samples by dropping the observations

of the households that did not participate in at least two consecutive surveys.For the within-groups analysis, a combined panel dataset10 was used As wasmentioned previously, the period 1992–1994 corresponds to a different panel andcannot be combined with that for the period 1995–2008; as a consequence, theseyears were not included in the within-groups analysis

Note that the combined panel dataset includes all the households which ticipated in all the rounds included in the period under study: 1995, 1998, 2001,

par-2004 and 2007 Therefore, the observations that were not included at least in one ofthese rounds were eliminated from the dataset The number of observations in thecombined panel dataset is made of 1463 households

2.2.3 Summary Statistics

This section presents general results based on the RLMS dataset and three types ofincomes (individual/household/equivalent) and provides information that will beused in the following empirical chapters

Figure2.1gives the average individual, household and equivalent incomes for

1992–2008 The plot includes confidence intervals (95%) to show the differencesbetween the individual and equivalent incomes (LB is a low bound and UB is anupper bound): both started from the same level in 1992, but since 1993 theequivalent incomes became larger The incomes, regardless of the type, decreasedbetween 1992 and 1998 and then increased

Figure2.2gives the mean and median equivalent incomes, including confidenceintervals (95%), and Fig.2.3plots the ratio of these two indicators for all types ofincomes This information will be used when computing the Foster and Wolfson(2010) FW index of bipolarization

Between 1992 and 1993, when incomes increased, the distance between themean and median increased and, since the relative change in the mean was largerthan that in the median, the ratio of the mean to median increased Between 1993and 1996, when incomes decreased, the change in the median was greater and, thus,the ratio increased again Then, in 1998 this ratio decreased but the change forequivalent incomes was not significant Between 2000 and 2008, the relativeincrease in the median was greater, thus, the distance between the mean and medianand the ratio of the mean to median decreased

Figure2.4indicates that the share of the population with zero incomes increasedbetween 1992 and 1996 and then decreased

individual incomes household incomes equivalent incomes LB equivalent incomes

UB equivalent incomes LB individual incomes UB individual incomes

Fig 2.1 Average incomes (in June 1992 rubles)

Figures2.5and2.6give the value of two inequality indices, the Gini index andTheil entropy measure (the computation includes zero-incomes) Both increasedduring the 1992–1996 period and then decreased.

To decompose inequality (Table2.4) into between and within-group nents, the sample was divided into two groups of equal size, the“poor” and “rich”,each one representing 50% of the population The results are illustrated in Figs.2.7

compo-and2.8 The detailed analysis of the changes in these components will be given inChap.4 which is devoted to income bipolarization

Fig 2.2 Equivalent incomes

individual incomes households equivalent incomes

Fig 2.3 Ratio of mean to median

individual incomes households

Fig 2.4 Share of zero-incomes

individual incomes households equivalent incomes

Fig 2.5 Gini index

individual incomes households equivalent incomes

Fig 2.6 Theil entropy measure

Table 2.4 Gini index

Year Equivalent incomes

[CI 95%]

Households [CI 95%]

Individual incomes [CI 95%]

individual incomes households equivalent incomes

Fig 2.7 Between-group inequality

individual incomes households equivalent incomes

Fig 2.8 Within-group inequality

Although both between and within-group inequality moved in the same tion, the relative increase between 1992 and 1996 was greater for the within groupsGini inequality GW and the relative decrease between 1998 and 2005 was greaterfor the between groups Gini inequality index GB Over the whole period, comparing

direc-1992 to 2008, the two inequality components did not change significantly.Table2.5 and Fig.2.9 give the results for the share of the middle class usingthree different definitions of the middle class, based respectively on the followingincome ranges: 75–125%, 60–225% and 50–150% of the median

Table 2.5 The share of the population belonged to the middle class

Year 75 –125% of the median 50 –150% of the median 60 –225% of the median

75-125 percent of the median 50-150 percent of the median 60-225 percent of the median

Fig 2.9 The share of the middle class (equivalent incomes)

The percentage of the population belonging to the middle class, regardless of therange used decreased between 1992 and 1996 and increased afterwards Over thewhole period, the share of the middle class did not change significantly, whatever

Distributional Change and What

Happened to the Middle Class in Russia

This chapter analyzes changes over time in the relative importance of the middleclass, using data on the distribution of income in Russia between 1992 and 2008.The first section is a survey of the literature on relative distribution, relativepolarization and the decomposition of relative distribution into location and shapeeffects

The empirical section is organized as follows The importance of the middleclass isfirst estimated on the basis of the so-called “relative distribution” approach.This nonparametric method allows a comparison of two distributions in terms oftheir ratio

The advantage of the non-parametric estimation is that, unlike the parametricapproach which assumes that the data are drawn from a known distribution, thisprocedure does not require that the datafit a parametric distribution

A decomposition technique is then implemented that enables us to isolate theimpact of changes in the median income and to see how the relative density wouldhave looked like if there had been no change in the distributional shape so that theanalysis is broken down into location and shape effects

A paper (co-authored with G Pittau) similar to this chapter was published in EmpiricalEconomics: Nissanov Z and M.G Pittau, 2016, “Measuring changes in the Russian middle classbetween 1992 and 2008: a nonparametric distributional analysis”, Empirical Economics, 50,503–530

© Springer International Publishing AG 2017

Z Nissanov, Economic Growth and the Middle Class in an Economy

in Transition, Economic Studies in Inequality, Social Exclusion

and Well-Being, DOI 10.1007/978-3-319-51094-1_3

21

3.1 Review of the Literature

3.1.1 Relative Distribution

Relative distribution is a non-parametric statistical method which was proposed byMorris et al (1994),1Bernhardt et al (1995) and Handcock and Morris (1998 and1999), Massari et al (2009) This approach allows comparing distributions byexpressing values of the comparison group as positions in the distribution of thereference group

Let Y0 and Y be the random income variable in the reference group and parison group respectively, F0 and F their cumulative distribution functions(CDF) and f0 and f their densities Cwik and Mielniczuk (1989) define the “gradetransformation” of Y to Y0 as the random variable R¼ F0ðYÞ

com-The CDF of R can be expressed as GðrÞ ¼ FðF1

0 ðrÞÞ; 0  r  1, where r is theproportion of income value y in the reference group and F01is the quantile function

of F0 Then, the relative distribution g(r) is defined as the ratio of the densities ofthe two groups evaluated at the rth quantile of the reference population:

of the relative distribution are independent from the scale of the distributions (e.g.incomes versus log-incomes) If there are no differences between the two distri-butions, then g(r) is uniform in [0, 1] and the CDF of the relative distribution is a45° line

In the current study, the empirical analysis will be based on the kernel densityestimation which is a non-parametric way of estimating the PDF of a randomvariable For a sample of n observations of incomes fx1; ; xng, the Kernel(density) estimator is defined as:

bfhðxÞ ¼ 1nh

Xn i¼1

1 Actually, the aspects of the relative PDF and CDF as a basis for a comparative analysis were examined earlier by, for example, Parzen (1977, 1992), Eubank et al (1987), Cwik and Mielniczuk (1989, 1993).

Ideally, the width of the kernel should vary in different regions of the distributionaccording to the sparseness of data; therefore, it is preferable to use an adaptivebandwidth, so that

bfðxÞ ¼ 1

hki

Xn i¼1

K x  xi

hki

whereki is the bandwidth weighting factor

This procedure will be used to estimate the relative distribution

3.1.2 Decomposition

The relative distribution can be decomposed into a location effect (differencesbetween the two groups due to a difference in the average/median income) and ashape effect (differences in other moments of the distribution like its spread orskewness) This decomposition, unlike the relative distribution, is notscale-invariant

Let FAðyÞ be the location adjusted density function, i.e the density of the randomincome variable (YA) in the reference group having the same median as the com-parison group The CDF of YA is FAðyÞ ¼ F0ðy þ qÞ where q = median(Y)− median (Y0) and its PDF is fAðyÞ Then, the decomposition into location andshape effects is represented as:

Handcock and Morris (1998)2propose to measure the contributions of the effects

of shape, location and other components to the total difference between the twodistributions using the Kullback-Leibler divergence, defined as: