Visible Consumption, Income Inequality and Social Comparisons. Evidence from Four Latin American Countries

1 We are very grateful for the comments and suggestions provided by Guillermo Alves, Veronica Amarante, Ana Fascioli, Robert Frank, Ignacio Munyo, Ori Heffetz, Juan Olano, Ambra Poggi, Gustavo Pereira, Agustín Reyes, Analía Rivero, Andrés Rius, Gonzalo Salas two anonymous referees, and attendants to Seminario del Instituto de Economia attendants to the meetings of the Human Development and Capabilities Association in Athens and Asociacion Latinoamericana y del Caribe para el Estudio de las Capacidades Humanas in Lima. All errors remain our own.

Abstract

Increased conspicuous consumption motivated by status-seeking behavior can undermine the gains in well-being derived from economic growth. Using expenditure surveys from Argentina, Brazil, Mexico and Uruguay, we assess visible consumption motivated by status seeking behaviour, analyzing different hypotheses about the role of externalities driven by social comparisons and income inequality. Specifically, we provide evidence for the relevance of different reference groups, assessing comparisons with richer (Veblen effects), poorer and similar profile groups (Duesenberry effects). We show that individuals are affected by multiple reference groups. Regarding between-group inequality, we find that individuals mainly seek to differentiate themselves from poorer groups, rather than imitating richer ones. In all cases, average reference group income of similar profile individuals is negatively related to visible consumption. Meanwhile, within-reference group responses are relevant in Argentina, Brazil and Mexico: increased visible goods consumption is associated with larger relative income gaps. In Argentina and Brazil, the sensitivity of visible goods expenditures to income is higher among households located above the reference threshold, which is consistent, again, with more affluent households trying to signal status by differentiating themselves from the most deprived group members. Meanwhile, in the case of Mexico, there is a symmetric response.

JEL codes:

• D12
• D31

Keywords:

• positional consumption
• visible goods
• reference groups
• latin america
• inequality

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 We are very grateful for the comments and suggestions provided by Guillermo Alves, Veronica Amarante, Ana Fascioli, Robert Frank, Ignacio Munyo, Ori Heffetz, Juan Olano, Ambra Poggi, Gustavo Pereira, Agustín Reyes, Analía Rivero, Andrés Rius, Gonzalo Salas two anonymous referees, and attendants to Seminario del Instituto de Economia attendants to the meetings of the Human Development and Capabilities Association in Athens and Asociacion Latinoamericana y del Caribe para el Estudio de las Capacidades Humanas in Lima. All errors remain our own.

2 Veblen also refers to the leisure enjoyed by the rich, which he calls “conspicuous leisure” and which can also be emulated.

3 The commodity boom exports led by the Chinese demand shock that mainly took place in 2003-2013 turned into high growth rates for Argentina, Brazil and Uruguay, but did not affect Mexico (Alvaredo and Gasparini, 2015).

4 The rationale for this is that if the middle increases, poorer individuals will be rare, and the middle strata will want to differentiate from them to a larger extent. Furthermore, the middle sectors will notice that with small changes in consumption they can differentiate from others.

5 Since the authors expect these effects to be heterogeneous within the group, they discard the use of synthetic inequality measures, such as the Gini index. To define the local peer group, in their empirical exercise the authors use a +-5% income interval.

6 Roychowdhury (2016) and Roychowdhury (2019) use instrumental variables to identify peer effects in conspicuous consumption in the former place and consumption in the latter case. Our work is in the spirit of Charles et al (2009), Kaus (2013); Chai et al. (2019), who use income as the reference group outcome variable and not consumption.

7 If this relation is convex, a fall in inequality (that leaves unchanged the group´s minimum) will reduce positional consumption, whereas if it is concave, it will increase it. This point is presented in Glazer and Konrad (1996), Property 3, page 1022.

8 Hirschman used the “tunnel effect” metaphor to uncover how social interactions could affect the long-term relationship between growth and inequality throughout the development process. In the short run, if individuals perceive improvements in others’ wellbeing, they will understand it as a forecast of their future situation and they might tolerate increasing inequality. However, if in the long run, these prospects are not achieved, frustration might arise, yielding discontent and political instability. This effect holds in societies in which different social groups interact or in contexts of generalized economic change.

9 Although our focus is on comparisons in income, we carry out a robustness check that fixes three of the four characteristics that define our reference groups and lets the remaining one vary. We comment on these results in Section IV.

10 We also tested other inequality measures such as the standard deviation, the variation coefficient and the ninth to first decile ratio.

11 The issue of endogeneity between total expenditure and its components can result from the simultaneity of decisions and from measurement errors. Besides carrying out an instrumental-variable estimation using per capita household income, additional tests were conducted using income decile, a durable goods index, current income and a third-order polynomial in current income.

12 Since race was available only for Brazil, we were not able to include race in our main specifications. However, we run additional regressions in the case of Brazil, including race in the reference group definition and all the remaining coefficients were similar. Results are available from the authors upon request.

13 If specifications including the average income from randomly chosen reference groups yielded statistically significant estimates, this might also be interpreted as individuals considering a representative sample of the population to be their reference point.

14 The results we obtained were similar to the ones reported in the main text and are available on request to the authors.

15 The corresponding coefficients for this set of variables, including reference group variables (IV.B), are similar (Table SM.4).

16 As a robustness check, expenditure was substituted by income in the OLS model estimation, and results were similar. Regression outputs are not included in this article due to space constraints, but they are available upon request to the authors.

17 Our estimations we include average reference group income, whereas Charles et al. (2009) use the logarithmic transformation. Thus, to compare the two sets of coefficients, we divide by 10 the ones obtained for Argentina, Brazil and Uruguay.

18 To check whether the bottom percentiles have the same effect when controlling for the average reference group income, we run additional regressions that are available on request to the authors. As expected, when both summary statistics of the reference group are included, their statistical significance and magnitude fall. In the cases of Uruguay, Mexico, and to a lesser extent Brazil, the average income effect dominates, confirming a negative and significant coefficient. In the case of Argentina, average income of the reference group loses statistical significance in both cases and the same result holds for average income of the bottom 10%. Conversely, the lower 5% dominates.

19 Results are available upon request to the authors.