The idea that we may be influenced by the consumption behaviour of our peers dates back at least to Thorsten Veblen’s theory of conspicuous consumption: "the competitor with whom [an individual] wishes to institute a comparison is [...] made to serve as a means to the end. He consumes vicariously for his host at the same time that he is a witness of that excess of good things which his host is unable to dispose of singlehanded" (Veblen 1899). Duesenberry (1948) followed up and emphasised the role of social influences on consumption in his relative income hypothesis, stating that "the strength of any individual’s desire to increase his consumption expenditure is a function of the ratio of his expenditure to some weighted average of the expenditures of others with whom he comes into contact".
Recent studies of peer effects in consumption confirm the existence of important consumption externalities for the US and other countries (Ravina 2007, Maurer and Meier 2008, Charles et al. 2009, Kooreman et al. 2011). These studies have been very influential. However, since they rely on aggregate definitions of peers, small samples, or strong identifying assumptions, they also have important limitations.
In new work, we use administrative data for the entire population of Denmark to study consumption externalities (De Giorgi et al. 2016). In particular, once we verify that household consumption is affected by peers’ consumption, we address two sets of questions. We first ask why this is the case, and second, whether those effects are large enough to be important for policymaking.
We focus on the influences that a husband’s and wife’s co-workers have on the household’s consumption. Co-workers (rather than neighbours and relatives, for example) seem a natural reference group, as adults spend most of their waking time at the workplace. Moreover, friendship often causes co-employment due to job search strategies adopted by job seekers. To make the definition of peers even tighter, we focus on co-workers sharing similar characteristics in terms of education, age, and occupation.
Our measure of consumption at the household level is constructed from information on disposable income and assets, available in administrative tax records and register data. We merge this information with data on the individual’s demographics, and to information on the workplace. Finally, we match our administrative dataset with a household consumption survey where we observe household expenditures on various goods, such as food, entertainment, clothing, jewellery, and durables (home appliances, etc.). This latter data source allows us to distinguish between competing hypotheses when we interpret our results on consumption network effects. In particular, we can classify goods according to their conspicuousness or visibility, hence allowing a direct test of Veblen’s theory for the presence of consumption peer effects.
In our analysis, we find sizeable network effects, which translate into a non-negligible social multiplier. We estimate an elasticity of own consumption with respect to peers’ consumption of about 0.3. This effect is similar for the husband’s and wife’s co-workers. While the positive peer effect translates into a non-trivial aggregate effect, its magnitude depends upon the degree of connectedness of peers and households in the population. To illustrate this point, we propose three policy experiments transferring the equivalent of 1% of aggregate consumption equally among (1) households in the top 10% of the consumption distribution, (2) a 10% random sample of households, and (3) households in the bottom 10% of the consumption distribution. These three policies are financed by issuing debt and running a government deficit. As an alternative to a debt-financed policy, we also consider (4) a purely redistributive policy in which the receivers of the transfers are households in the bottom 10%, and the policy is financed by a ‘tax’ to the top 10% of households.
From experiment 1, we learn that consumption policies targeted at the top 10% of the consumption distribution (presumably also the wealthier households) have limited aggregate effects, and in particular the effects do not spread along the distribution of consumption. Experiment 2 reveals that policies that target a random sample of households have larger and more far-reaching consequences. This is because those households are located throughout the consumption distribution and thus tap into larger and denser networks. Even larger aggregate effects are found when the policy targets the bottom 10% of households (experiment 3), where individuals presumably work in very tight networks. An added consequence is that redistribution towards the lower part of the income distribution implies a noticeable reduction in consumption inequality (a 13% decline in the standard deviation of log consumption). In experiment 4, where we consider a balanced budget experiment in which a transfer to poorer households is financed by a tax imposed on the richer households (who hence mechanically reduce their consumption), we obtain an intermediate multiplier effect and a large reduction in consumption inequality.
With positive peer effects influencing aggregate consumption, it becomes important to understand what drives such effects. We hence run a horse race of three models that have enjoyed favour among researchers. The first model revisits Veblen’s idea of conspicuous consumption and suggests that the allocation of consumption among goods may be tilted towards goods that are ‘conspicuous’, such as jewellery, luxury cars, restaurants, and so forth. The second model is the ‘keeping up with the Joneses’ model, in which individual utility depends on the average consumption of peers (and not necessarily the visible one). The third model is one where risks are shared among members of the reference group, which creates correlation among their consumptions with peer effects being, effectively, a spurious manifestation of the risk sharing agreements. We find empirical support for a ‘keeping up with the Joneses’ model – based on our empirical analyses, we can rule out models of conspicuous consumption as well as full or partial risk sharing. Hence, our results point towards an inter-temporal distortion of the spending profile rather than a tilting of consumption towards luxury and conspicuous goods.
In sum, why is the study of consumption network effects important? There are at least two reasons.
- First, measuring and understanding the type of distortions induced by the presence of peer effects is important from a welfare point of view.
Our results point towards an intertemporal distortion of spending profiles, in which case the saving profile differs from the optimal one obtained if agents had acted atomistically (without considering what their peers do). In turn this implies under-saving (or over-borrowing) in the attempt to keep up with the Joneses. Had the underlying mechanism been status-seeking, it would have implied an inflated share of ‘visible’ or conspicuous goods in the household’s budget. Since visible goods are typically luxuries (cars or jewellery being the most notable examples), consumption peer effects would have implied welfare consequences in the form of excess consumption of goods that are typically disfavoured by the tax system. Had risk sharing been the main reason for correlated consumption profiles it would have been associated with important welfare gains.
- Second, consumption network effects are relevant for policymaking.
Our tax simulations show that uninsured idiosyncratic shocks, such as a tax changes targeting particular taxpayers, have dramatically different consequences that depend on the peer effects and the underlying network structure of peers and households. Hence, studies neglecting peer effects would tend to either overestimate or underestimate the effects of tax reforms that target certain groups (the rich) and neglect the percolation of effects to groups not targeted by the reform but potentially connected to them through network effects (even in the absence of general equilibrium effects).
Charles, K, E Hurst, and N Roussanov (2009), "Conspicuous Consumption and Race", Quarterly Journal of Economics, 124 (2), 425-467.
De Giorgi, G, A Frederiksen, and L Pistaferri, (2016), “Consumption Network Effects”, CEPR Discussion Paper 11332.
Duesenberry, J S (1948), “Income-consumption relations and their implications” in Lloyd Metzler et al., Income, Employment and Public Policy, New York, WW Norton & Company, Inc.
Kooreman, P, P Kuhn, A Soetevent, and A Kapteyn (2011) "The Effects of Lottery Prizes on Winners and their Neighbors: Evidence from the Dutch Postcode Lottery", The American Economic Review, 101 (5), 2226-2247.
Maurer, J, and A Meier (2008), "Smooth It Like the `Joneses’? Estimating Peer-Group Effects in Intertemporal Consumption Choice", The Economic Journal, 118 (2), 454-476.
Ravina, E (2007), "Habit Persistence and Keeping Up with the Joneses: Evidence from Micro Data", mimeo., Columbia University.
Veblen, T (1899), The theory of the leisure class: An economic study in the evolution of institutions, Macmillan.