There is a new paper from Ricardo Perez Truglia, from Harvard, on this topic. It strikes me as quite speculative, but nonetheless a step forward in addressing a very difficult question and adding some structure to the analysis of a very difficult problem. Here are his conclusions:
The goal of the paper is to provide a quantitative idea of the practical importance of conspicuous consumption. We estimated a signaling model using nationally representative data on consumption in the US, which we then use to estimate welfare implications and perform counterfactual analysis. We found that the market value of NMGs [TC: non market goods, as result from the signal] is non-negligible: for each dollar spent in clothing and cars the average household gets around 35 cents of net benets from NMGs. However, the large value of NMGs does not imply that the losses from the positional externality are also large. The results suggest that richer household would still consume relatively more of the NMG even in absence of NMGs, so the cost of the signal that they send is not very high. As a result, the signaling equilibrium attains almost 90% of the full potential benets from the NMGs, which is very efficient. The unattained benefits, $32 per household per month, can serve as an intuitive upper bound to the benefits that can be gained through economic policy, such as a tax on observable goods aimed at correcting the positional externality.
A related point is that if utility functions evolve so that people enjoy the very act of sending the signal, that too will lower the associated deadweight loss from signaling.
Here is the author’s home page.