Middle Class, Government Effectiveness and Poverty in the EU: A Dynamic Multilevel Analysis

• Published date: 01 March 2020
• Author: Bruno Bosco,Ambra Poggi
• Date: 01 March 2020
• DOI: http://doi.org/10.1111/roiw.12403

© 2019 Internation al Association for Re search in Inco me and Wealth

94

MIDDLE CLASS, GOVERNMENT EFFECTIVENESS AN D POVERTY IN

THE EU: A DYNAMIC MULTILEVEL ANALYSIS

by bruno bosco* and AmbrA Poggi

Departme nt of Economics,Un iversity of Milan-Bic occa

Using the 2008-2011 EU-Statistics on Income and Living Conditions data, we implement a dynamic three-

level model to analyze poverty persistence in 26 EU countries. We isolate true state dependence phenom-

ena by disentangling the effects of observed and unobserved heterogeneity at country level and employ

cofactors not previously considered by the literature. Estimates show that unobserved heterogeneity across

individuals remains large, even after explicitly controlling for the observable components of individual

characteristics. The initial value of poverty has large effects on current poverty status but this effect is not

uniform across countries. The risk of poverty is negatively related to the size of the structural middle class

and to the level of structural social expenditure but it increases when lagged total public expenditure

increases (with respect to the structural value). There is strong evidence of true state dependence.

JEL Codes: C23, C25, I30

Keywords: cross-country poverty dynamics, government effectiveness, middle class, multilevel model

1. introduction

The 2008 economic crisis and the protracted period of instability and stag-

nation that almost immediately followed the outbreak of the crisis came with an

increase in poverty across the EU (Dueilla and Turrini, 2014; Stockhammer, 2015,

Bosco, 2016a). In particular, in the member states most severely hit by the crisis, the

prospects for the most vulnerable parts of the population became a serious source

of concern. In the 2010, the EU member states endorsed a new EU strategy (called

Europe 2020) designed to promote, among other objectives, “social inclusion in par-

ticular through the reduction of poverty” by aiming to lift at least 20 million people

out of the risk of extreme poverty and social exclusion. Logically, the fulfillment

of the above ambitious 20-million-people program requires a thoughtful analysis

of how future income growth, educational levels, technical development and any

other economic improvements as well as political accountability and institutional

quality can generate social inclusion and poverty risk reduction or persistence. Yet,

Note: Financia l support was provi ded by University of Mi lan-Bicocc a (2015-ATE-0425; 2015-

ATE-0425). Thi s paper was also su pported by CEf ES-DEMS. It has been pre sented at the 2016

HDCA Conferen ce on Capability a nd Diversity in a Globa l Society, Tokyo, at the 2016 Conference

on Desar rollo Sostenibl e: PerspectvaL ocal y Global, Unive rsity of Havana, and at t he 2016

Conferenc e Inequality a nd Fairness of Politi cal Reforms, Univers ity of Mannheim. W ith the usual

disclai mers, we than k two anonymous refe rees as well a s conference s’ particip ants, and part icularly

S. Juhl, T. Swalve, C. Saroji ni Hart, V. Tselios, R. Gonz alez for helpful com ments and sugge stions.

Correspondence to: Bruno Bosco, DEMS, University of Milan-Bicocca. 1, Piazza Ateneo Nuovo,

20126 Milan, Italy (bruno.bosco@unimib.it).

Review of Inc ome and Wealth

Series 66 , Number 1, March 2020

DOI : 10.1111 /roi w.124 03

bs_bs_banner

Review of Income and Wealth, Series 66, Number 1, March 2020

95

© 2019 Internation al Association for Re search in Inco me and Wealth

the causes behind the dynamic process that leads an individual to be poor are both

macro and micro and the persistence of poverty may arise from individual hetero-

geneity as well as countries’ economic and political factors. Hence, if individuals

experience poverty in a specific time because of adverse idiosyncratic personal

characteristics, they will be likely to experience poverty in other subsequent peri-

ods. This may happen notwithstanding the improvements in the country’s general

economic conditions from one period to the next. These adverse characteristics can

be either observable or unobservable. Empirical literature has widely inquired the

impact on poverty of individual observable characteristics (i.e. sex, level of educa-

tion, household status, occupation status) and country-level (mainly) observable

characteristics (i.e. per capita income, different qualitative and quantitative mea-

sures of the welfare state effort and country specific effects). We contribute to this

literature by analyzing in a dynamic longitudinal multilevel framework the rela-

tionship between poverty dynamics and observable as well as unobservable country

factors. We feel that a proper understanding of the above dynamic relationship can

be a useful support to any anti-poverty program in the EU as elsewhere.

The specific aim of this paper is twofold. We want to estimate the probability

of poverty persistence in the EU when observable and unobservable heterogeneity

at the country level are correctly disentangled and the dynamic process that leads

an individual to be poor has been defined in order to isolate true state dependence

phenomena. Distinguishing between the two processes (true state persistence and

heterogeneity) is crucial since the policy implications of the two are very different.

If persistence of poverty is (at least partly) due to a true state dependence, then it

makes sense to plan measures aiming at pulling the individual out of poverty at time

t in order to reduce her chance of experiencing poverty at time t+1 and to imple-

ment measures that intervene on factors that generate the true state dependence in

order to break the “vicious poverty circle”. On the contrary, if the persistence of

poverty is mainly due to unobserved heterogeneity, any short-term policy aimed to

pull the individual out of poverty at time t would not be effective at time t+1. It is

therefore crucial to disentangle the effects of country-level explanatory variables

and to isolate state dependence. To this end, the paper implements a three level

dynamic multilevel model. Building on the technique proposed by Wooldridge

(2005)—that allows to consistently estimate a two level model with both lagged

dependent and exogenous variables to distinguish between true state dependence

and heterogeneity at the individual level—we propose a method that adds a further

level of the analysis in order to investigate the impact of heterogeneity at country

level on the individual risk of poverty. This three-level approach permits to dis-

entangle the effects of country-level explanatory variables and to appreciate the

effects of country dummies by specifying country membership as a random effect.

Framing the analysis in the above methodology, we contribute to the literature by

expanding the set of hypotheses about the effects of observed heterogeneity and include

for the first time the size of the middle class in each country as one of the fundamental

drivers of individual poverty dynamics. We also control for the quality of public policy

implementation measured by an appropriate internationally recognized index of gov-

ernment effectiveness. Yet, direct reference to the middle class represents the main new

hypothesis of this paper and it is motivated by the observation that the size of the mid-

dle class is beheld to be instrumentally important to economic development. A large

Review of Income and Wealth, Series 66, Number 1, March 2020

96

© 2019 Internation al Association for Re search in Inco me and Wealth

stand of current literature views middle class as a factor that fosters entrepreneurship,

shifts the composition of consumer demand and makes it possible to implement those

political and institutional reforms conducive to growth and, ultimately, to poverty

reduction. Consequently, middle-class is a key factor in the implementation of some

(theoretically common) anti-poverty EU policy measures. After controlling for the

dissimilarities among national welfare systems and the organization and operation of

each public national administration as possible sources of cross-country heterogeneity,

we specifically conduct a test of the role of the middle class size as a cofactor explaining

cross-country differentiated levels of people exposition to the risk of poverty.

Hence, this paper proposes a dynamic three-level estimation strategy and expands

the hypotheses about the countries’ sources of observable heterogeneity by including,

alongside welfare related factors (e.g. social expenditure), the size of the middle class

and a measure of the quality of public policy. In doing so we intend to offer fresh

empirical evidence on the determinants of poverty in the EU during a period charac-

terized by a deep economic crisis. Results show that the new hypotheses are consistent

with the data generation process. The coefficient on the lagged poverty is highly statis-

tically significant and the initial value of poverty shows, with significant cross-country

differences, that there is substantial correlation between the initial poverty condition

and the unobserved heterogeneity. This will affect future realizations of poverty: indi-

viduals experiencing poverty at a certain point in time have a higher probability to

experience poverty in the future than non-poor individuals do (i.e. there is evidence

of true state dependence). Yet, structural high levels of both middle class and social

expenditure reduce the above probability of experiencing future poverty, which is also

reduced by positive deviations of lagged values of social expenditure from their struc-

tural mean values. On the contrary, “good” public policy reduces the risk of poverty

only in a subgroup of countries or when the poverty line is held constant over the

period covered by the analysis. Moreover, the estimate of the true state dependence

under the hypothesis of a structurally limited middle class is much higher than the one

observed assuming a structural extended middle class. Thus, estimates allow inferring

that an extended middle class combined with high social expenditure and good public

policy may help reducing the adverse future impact of experiencing poverty in a spe-

cific period. However, the effectiveness of similar measures implemented to contrast

initial poverty (e.g. a one percent increase of social expenditure) can be hampered/

amplified by unobserved country characteristics according to the initial state of pov-

erty. Our estimates show how serious this problem can be.

The rest of the paper proceeds as follows. In Section 2, the literature on poverty

dynamics is shortly reviewed. Section 3 contains our discussion of the hypotheses to

test about the role of the middle class and the quality of institutions. In Section 4,

we present the data set employed in the estimation strategy. Section 5 contains a pre-

sentation and a discussion of the estimation results and tests. Section 6 concludes.

2. LongitudinAL studies of Poverty

2.1. L ongitudinal Approach and Micro -Drivers of Poverty

Longitudinal research on poverty has covered two main themes: the duration

of poverty spells (and its persistence) and the determinants of poverty states