The Elusive New Middle Class in Brazil

Against the background of the generalized reduction of poverty in the world, and particularly in Brazil, this article intends to gauge the socio-economic profile of Brazilian households that emerged from poverty and have been identified as integrating a "new middle class". Using indicators of standards of living from the 2008-2009 Survey on Family Budgets (POF/IBGE), we found out that, in contrast to what has been assumed on the basis of average income criteria, this social stratum is markedly heterogeneous, most of it being similar in their consumption patterns to the economically vulnerable or outright poor strata. So, we conclude that, from a sociological perspective that demands additional conditions besides income levels to identify social classes, it is a category mistake to call this social stratum a new middle class. We conjecture that this may be consequential in terms of policy priorities and choices.

(2015) 9 (3) 21 -41 characteristic that strongly marked the world economy in the last 20 yearsat least before the full unfolding of the consequences of the "Great Recession" that started in 2008 -was the intense growth experienced by practically all countries, the developing ones in particular, in the wake of the process of market globalization and the rise of China. In social terms, the main consequence of this phase of growth of developing countries was the dramatic reduction of poverty in the world population, and today one may even envisage its virtual disappearance, at least in its more extreme form 1 .
This process has been accompanied by the growth of intermediate layers between the poor and the rich, which, by having access to new levels of mass consumption, some analysts prefer to consider as members of a presumed "new middle class", people who have access to additional resources (nearly a third of the total) after the satisfaction of basic needs like food and housing. It is estimated that, in the emerging markets as a whole, the so-called middle class increased from nearly a third of the population in 1990 to more than half today 2 .
A point to be considered is that these analyses admit that it is not enough to have escaped poverty in order to enter the threshold of the "Middle Class". For this, it would be required to have enough additional monetary resources to allow the discretionary spending in material and cultural consumption. The hypothesis seems to be that there is an intermediate stratum between the poor and the middle class, a stratum whose position is still precarious and unstable, without full access to the consumption market and constantly besieged by the ghost of social decadence.
A recent study by the World Bank on the emergence of the middle class in Latin-America 3 explicitly considers these criteria. The study defines the middle class in terms of economic security, i.e., of a low probability of falling down (for many, falling back) into income poverty. As may be guessed, their results show that the "vulnerable", who do not make enough to enjoy economic security, is the largest group of Latin American population, including that of Brazil. lead of this study, Birdsall et al. (2014) look into the prospects of this group, which they call "the strugglers", and predict that they may well end up marginalized in the not-too-distant future.
Still, the Latin American case awakens interest inasmuch as the reduction of poverty was accompanied by a sustained contraction of inequality; in any case, a rarity in the history of the region (CORNIA, 2014) 4 . Lustig et al. (2013)  The decline in inequality as the income of the poorer increased in a proportion larger than that of the richer seems to be an established fact (see Table   01 below) --and its contribution to the reduction in poverty is only somewhat lesser than the regional average (LUSTIG et al., 2013). 4 See especially chapters 01-02, pages 03-69. Some works empirically relate such changes with stable democracies and center-left coalition governments in the region (e.g. CORNIA, 2010CORNIA, , 2014HUBER et al., 2008). In fact, in Brazil, if currency stabilization with the Real Plan was crucial to cutting poverty levels from their 1980's 30% peak level down to 20% in 1995, while poverty levels had been pretty much pro-cyclical before that, after 2002 and up to 2011 poverty halved (see Figure 02 below) also on account of policy-induced inequality reduction. It is generally acknowledged that public policies were important in this period (HUBER et al., 2008;HUBER and BOGLIACCINI, 2010;LÓPEZ CALVA and LUSTIG, 2010;LUSTIG et al., 2013;SOARES, 2011). Among them are the expansion of educational attainment, labor market policies, such as minimum wage valorization, cash transfers to the poor and non-contributory pensions to rural workers.
However, the jury is out as to whether this large group of socially betterpositioned people migrated to the middle class, as has been repeatedly suggested. The question is not simple. Beyond the issue of class formation, there are different definitions and ways to measure social positions but, in any case, there is a wide consensus that family income is far from being a sufficient criterion for class belonging and that other sociological criteria are to be adopted. This assumption also underscores the work of Ferreira et al (2012)  The Brazilian debate to a large extent centers on the reception of economist Marcelo Neri's research on this issue. His seminal analyses (NERI, 2008(NERI, , 2010(NERI, , 2011 support the positive view that the income gains have qualified a significant stratum of the former Brazilian poor to the middle class club, "the new middle class", and were met with relative sympathy by the empirical work of the sociologist Bolívar Lamounier and political scientist Amaury de Souza (2010). Recent studies have questioned the "class" interpretation, though. Most of them criticize the exclusive focus on income and lack of reference to social structure, especially the occupational structure (POCHMANN, 2012;QUADROS, 2010;SCALON and SALATA, 2012;SOBRINHO, 2011). According to the ageold social stratification tradition, occupational traits and labor market insertion, secondary aspects in Neri's study, are essential markers of class positions 5 . Other analysts take issue with Neri's analysis but also with critical efforts directed to it (SOUZA, 2010), since these pay too much heed to the economic dimension of stratification and neglect essential cultural dimensions such as class culture (e.g., lifestyles, tastes) and the symbolic production of social distinctions in processes of class formation and reproduction. Otherwise diverging in their starting points, all of these authors converge to the conclusion that the shortening of social distances we have been witnessing is mostly the emergence of a "new working class", or "struggling class" (SOUZA, 2010), still deprived of the social and economic security that distinguishes the full-fledged middle class (FERREIRA et al., 2012).
This paper aims at dialoguing with that small but growing literature by attempting to introduce the perspective of "lifestyle" 6 in the debate, a marker privileged in the field of Sociology of Consumption studies 7 . This is actually done through the gauging of the standards of living of the households that, by the income criterion proposed by Neri (2008Neri ( , 2010Neri ( , 2011, have been identified as members of a "new middle class", in order to examine the adequacy of such a classification. We are aware that the translation of some sociological concepts into statistical indicators is risky, and this is not exactly what we intend to do. We are rather interested in the idea of class differentiation occurring at the level of combinations of capitals and resources (economic, social, cultural), but are obviously limited by the available statistical information. Indirect evidence of differential access or possession of resources is given by indicators of standards of living and, in particular, of consumption patterns.
In contrast to the work of Ferreira et al. (2012) and Birdsall et al. (2014), which exclude from the middle class those non-poor people with a high probability of falling into poverty, we do not depart from a previous poverty line definition and income range. To check the adequacy of Neri's classification, we rather start out from a standard of living that characterizes a middle-class lifestyle, according to common perception.
Also we do not establish an objective cutoff, but rather search for similarity among economic strata. But the mentioned authors' intuition is similar to common perceptions in Brazil, that economic security is central to the middle-class identity. To avoid misunderstandings, a couple of disclaimers are in order. In this paper we do not intend to develop an alternative classification of the middle class but instead check the strength of the classification that has been proposed on the basis of the income criterion against consumption patterns criteria. Our hypothesis is that a considerable proportion of the people that on the median/average income criterion have been classified under the new middle class label are more like a vulnerable social stratum than a well-established middle-class stratum when consumption patterns are considered. Secondly, the approach is essentially descriptive and does not intend to advance any kind of normative analysis on inequalities and forms of reparation thereof.
In the remainder of this paper, we present the rationale for the choice of markers in section 01, the adopted methodology in section 02, and the results for the analysis of the so-called "new middle class" in section 03. In the last section, we briefly mention some of the implications.

How to identify middle class?
In his recent study on what he calls the new Brazilian middle class, Neri (2011) states that this stratum is the one that "gets on average the average (sic) income of society, i.e., constitutes the middle class in statistical terms" (NERI, 2011, p. 83)  identify some of the characteristic markers of this stratum and their incidence in the mean income group as identified in Neri's (2011) study. Our intent is to verify the equivalence between socially-shared perceptions and the statistical classification based on income. To this end, we combine the markers with the information available in the POF database. Given that this database provides information on family budgets, we concentrate on consumption patterns.
It turns out that our path to the analysis of the socioeconomic profile of that income bracket finds a rationalization in many sociological approaches, including some of Bourdieu's seminal analyses (2008,2012). According to these theoretical perspectives, social strata are constructed by the marking of their relative positions, as well as by a general perception and recognition of these positions, and not solely as a function of the property of some specific resource. The element of perceived differential consumption or distinction plays a central role. People's perceptions on the middle class seem to confirm this, as we explain below.
Souza and Lamounier's study (2010), on which we draw, indicates the saliency of a certain set of markers, which are socially acknowledged as distinctive of the middle class 9 . Among these markers figure a stable standard of living, house ownership, high income, university education, and a prestigious profession, access to leisure and amusement, and access to good schools for the children 10 . The qualifications "stable", "ownership", "high", "university", "prestigious", "good", besides access to things not considered as primary needs, like leisure and amusement, indicate differentiation relative to the consumption of lower strata (the poor and median strata), from which the middle class wishes to be kept apart. To this list, we added access to "modern" means of communication and information, we qualified the owned house in terms of "good" housing conditions, and we also added free time to the item access to leisure and amusement. It is worth noting that, in contrast to the study authors, who do not call into question the proposition that the families situated in the income bracket defined by Neri (2011) are members of the middle class, we utilized the markers they identified precisely to examine the proposition's adequacy.
Our variable selection, explained in the next section, resulted from the combination of the markers and information available in the database. Our distinction criteria, restricted to information in the database, were limited to possession and access to specified resources (goods and services).

Methodology
The focus of analysis, following Neri ( As already mentioned, we examined households in the income bracket from R$ 1,052.00 to 4,537.00, to which was attributed the status of "new middle class" (from now on NMC stratum or segment) 11 . Besides, having in view a more detailed analysis of the differentiation within the NMC, we further divided this large segment into three equal-length sub-segments, as shown below in section 03. Once we determined the households to be studied, we examined the many aspects of such units and the heads of households' attributes in such a way as to answer whether these households may plausibly be considered as typically middle class ones. This way, we associated the markers or selected criteria (cf. section 01) to variables and for each of these we identified one question in POF, according to Table 02 below.
The equivalence between criteria and variables is to some extent selfexplanatory, but it is worth reflecting on the decisions here taken. First, the identification of variables that correspond to markers was limited by two parameters: on the one hand, by the belief that there is not a necessary relation between marker and variable and that other variables could equally well indicate the occurrence of these criteria. Besides, variables could be classified in a different way than that presented here. In this sense, our sole claim is plausibility. On the other, we are limited to the information available in the database. As a consequence, and additionally, we want to make explicit our reasons to classify variables the way we did.  (2011) ** up to 2 people per bedroom *** Adequate material = adequate construction material of walls + floor + roof. Adequate wall material: masonry and/or wood. Adequate floor material: carpet, ceramic tiles, wood over cement slab. Adequate roof material: wood shingles, corrugated steel sheets or cement slab. **** Services: water, trash collection, street lighting, storm drains, electricity, rapid transit, education, health, street cleaning and maintenance, sewerage.
Beginning by the "stable standard of living" criterion, we focused on the access to forms of private pension and sources of resources (such as wealth, credit and investment income) that reduce uncertainty in the middle and long run, stabilizing the standard of living while making it less dependent on monthly income. This correlates directly with the Lamounier-Souza (2010) survey in which most people explicitly cited a stable standard of living as a characteristic of the middle class. Economic security, a trait suggested in the Ferreira et al. (2012) study, is also captured here 12 .
The "home ownership with good housing conditions" criterion refers to housing conditions in general and, by not concentrating solely on the property of the dwelling and also taking into account its habitability, especially the access to public services, directs attention to the effective quality of housing. This criterionhome ownershipalso appeared with distinction in the Lamounier- Souza (2010) survey.
In the "services" indicator, we decided to consider all ten selected ones, for they are all considered essential. Our interest in this case was to observe the existence of the services as a way of qualifying the housing situation.
The criteria "university education of the household's head" and "access to good schools", which also appeared prominently in the LS (2010) survey, may give us an approximate notion of the level of cultural capital made possible by different income strata, especially its intergenerational transmission.
In the criterion "access to leisure, diversion and spare time", another element cited in the LS (2010) survey, we decided to include both leisure inside and outside the house and availability of free timeelements that conform to a distinctive consumption pattern, not included in what are commonly considered basic needs. The variables that allowed us to identify the availability of free time, to which we associate a decent living standard, were ownership of washing machine and vacuum cleaner, and spending on domestic help, a peculiarity of the Brazilian middle class.
Finally, the criterion "access to information and communication technology" was included to give some sense of possession of social capital or social connections. 12 The idea here is not so much that these people have access to basic goods but that they enjoy a more permanent sense of economic security, given by their "economic capital" in the terminology of Bourdieu (2008). In this sense, economic security distinguishes them from vulnerable or outright poorer strata.
Having selected the variables and defined the proxies in the database, as shown in Table 01 above, we register the numbers, absolute and relative, of households that answered "yes" to the questions. We break the income distribution into six income brackets and proceed to analyze the similarity among them, as we explain in the next section.
A warning applies here. We decided to include expenditures on health insurance plans (in the stable living standard criterion) and private schools (in the access to good schools criterion) as markers of middle class consumption patterns, in spite of the high market stratification of these services, because, as a matter of fact, they seem to constitute distinctive middle class traits in Brazil. If what new consumers are getting is real quality, which is doubtful, is a different matter and something we are not able to assess with the data at our disposal. In concrete terms, this means that we may be overestimating the amount of middle class households, which runs counter to our own hypothesis.

Results
A first approximation to the data shows that the households in the income bracket that correspond to the so-called "new middle class" amounted, in 2009, to 31.6 million, or 55% of the Brazilian households. The sheer width of the income bracket that has been identified as the "new middle class" is very large, and this suggests that different situations may have been considered uniformly when we look at the income bracket as a whole.
In fact, a preliminary analysis indicates that this segment is characterized by a strong inequality within it, and the predominance is of dwellings in the lower NMC income segment, where apparently the largest growth occurred in recent years, be it in absolute terms or in percentage points. This lower NMC sub-segment alone accounts for almost one-third of all households in Brazil and about 60% of all NMC households So, in view of these observations, we broke the NMC group into three subgroups of equal income range, and, to keep our results comparable with Neri's (2011), we decomposed the household income distribution as a whole into six strata: the first stratum corresponds to Neri's class E, where the poor households are located; the second one corresponds to Neri's class D, of vulnerable people; the third, fourth and fifth ones are Neri's class C or the new middle class; and the sixth one stands for Neri's classes A and B (see Table 03, below). ** million. *** Brazil's currency is the Real (R$). Over the study period of this paper, the exchange rate with the dollar fluctuated in an interval between approximately R$ 1.65 to R$ 3.10/US$, with a rough average of R$2.20/US$. A preliminary general observation is that there are some relatively "universalized" things: various housing conditions, as well as possession of color TV sets, are present in most households in all income brackets examined (the same happened with refrigerators, not included in Table 03). These items do not seem to differentiate much along the income groups, at least on the basis of the data at our disposal. However, an interesting second finding in regard to our exploration of the new middle class segment was to see that the second poorest stratum (class D) and the first income stratum of the NMC group (class"C1") seem indeed to show substantial affinities: most of the households in strata D and C1 do not possess 17 and 16 items, respectively, of a total of 23, the shortest distance considering all pairs of adjacent strata on this criterion. To be sure, only the richest stratum (Neri's (2011) classes A and B) enjoys the situation of most of the households holding most of the items, but, again, the absolute distance is somewhat larger among any other adjacent strata than it is between D and C1. Poor households (class E) stood alone only in the minority of those who possessed DVD equipment. The relations between income and every item listed in Table 03 were statistically significant. We report the results of various tests in Table 05 below. Our initial visual impression of a non-robust NMC group was doublechecked with the use of a common technique of multivariate analysis that aims at grouping objects on the basis of their profiles. Our objects are the six income strata and the characteristics that compose the profiles are the list of 23 variables we considered in Table 03. In this technique, the final result is given in graphical form, in which similarities among the objects are usually computed as distances on a Euclidean scale; the shorter the distances, the more similar the objects are to one another.
So, as a way of summarizing the profile similarities delineated above, we submitted the data in Table 03 to a Hierarchical Cluster Analysis, using the SPSS software. The analysis is based on the quadratic Euclidean distance and its final results are presented in graphical form by the dendogram in Figure 03 below. As we can see, in its first step the clustering algorithm aggregated the two poorer income strata; after that, the two upper NMC strata were clustered, clearly differentiating them from the lower NMC stratum; and in its third step, this lower NMC stratum is clustered with the first group produced, those with lowest incomes. Therefore, stopping the analysis at this point, we would end up with 03 strata clusters: (01) a "poor group", those earning up to R$ 2,214.00, composed of the three poorer strata (i.e. including the first NMC segment); (02) a "middle" group with the two upper NMC strata, those earning more than R$ 2,214.00 but less than R$ 4,537.00; and, finally, (03) a "high" income group, those earning more than R$ 4,537.00. Note that the first segment of the NMC, which is clustered with the poorer strata in the "poor group", is where the median income (R$ 1,586.63) is located.

Final comments
There are many problems with the positive identification of a "new middle class" in Brazil, but various attempts support the suspicion that it is not correct to classify the people that just emerged from poverty in that social stratum characterized by economic safety and by distinctive consumption patterns and lifestyles. Taking these elements as a reference, our study showed the heterogeneity of the NMC group, the similarities between the lower layer of this group (accounting for more than 60% of the households) and the poorer income classes, as well as the weak similitude between the moneyed sectors of the NMC and the families with higher income.
In our search for the new middle class we found a group of relatively welloff families still far away from a stable standard of living and other desired and distinctive attributes. This group, the first stratum of the NMC and the more populous one, thus looks much more like a vulnerable, "non-capitalized" social segment than an economically secure one, distinctively middle-class according to ordinary opinion and sociological canon.
In the wake of the global crisis and in the face of deflated expectations of economic and social stability of this group following economic stagnation and the cooling of the labor market, it seems more than a matter of semantic precision to unveil the condition of this chunk of households. Classificatory misspecification, signaling the false premise that poverty and vulnerability might have been overcome, may have harmful consequences in terms of policy priorities and decisions.