Optimal Consumption in the Age of Family Networks: Beyond the Representative Agent

TL;DR

This research challenges the traditional "representative family" model in economics by proposing a framework that accounts for heterogeneity within family networks. By deriving closed-form solutions for consumption and welfare, the authors demonstrate how family members with varying risk tolerances and discount rates—interconnected through modern social platforms—make rational inter-temporal decisions.

Background: The Death of the Homogeneous Householder

In orthodox neoclassical theory, the "family" is often treated as a single, uniform point of data. However, the rise of social networking services like FamilyWall or Pinterest has highlighted that family units are actually complex networks. Within these networks, individuals may share a general "vibe" (preferences) but differ wildly in their "deep parameters"—basically, how much they crave immediate gratification (subjective discount rate) and how much risk they can stomach.

The Core Insight: Modeling Heterogeneity as a Distribution

The authors assume that while every family member uses a Negative Exponential Felicity Function, their specific parameters are not identical. Instead, they are distributed according to an exponential type.

1. The Mathematical Setup

The model splits families into groups (mothers, fathers, and descendants). The total utility of the couple/family is evaluated by integrating the individual utilities over the probability distributions of their risk and time-preference parameters.

The optimization problem is defined as: $extMaximize{\int }_0^{\infty }(extAggregatedExpectedUtility)dt$ Subject to: $k_0=\sum {\int }_0^{\infty }{c}_i{e}^{-rt}dt$

Methodology: From Couples to $N$ Members

The paper first solves for a two-person household (parents) and then generalizes the solution to a family of $N$ members.

The Optimal Consumption Path for any individual $j$ is derived as: $c_j=\frac{e^{\frac{A}{2}(t-\lambda )}}{\sqrt{t+\lambda }}\left(\frac{A{k}_{j0}+{\mu }_j}{\sqrt{8\pi A}\left(1-\Phi \left(\sqrt{A\lambda }\right)\right)}\right)-{\mu }_j$

Where:

• $A$ is the marginal product of capital.
• $k_{j0}$ is the initial endowment.
• $\mu$ and $\lambda$ represent parameters of the risk and discount distribution.

Experimental Analysis: Sensitivities and Welfare

The authors performed "comparative static" exercises to see how these families react to changes in the economy.

Key Findings:

• Wealth Effect: There is a strictly positive relationship between initial capital and consumption. More starting resources lead directly to higher consumption paths.
• Interest Rate Ambiguity: Unlike simple models, the effect of interest rates ($r=A$) on consumption is ambiguous and depends on the time horizon and subjective discount rates.
• Total Welfare: The welfare formula is obtained in a "closed-form," allowing economists to calculate exactly how "happy" a network is based on its consumption rationalization.

Figure: The visual representation of how the real interest rate influences consumption levels over a two-year period.

Critical Insight & Future Outlook

The primary value of this work is its analytical elegance. Getting a closed-form solution for a network of $N$ heterogeneous agents is mathematically difficult. By using Fubini’s theorem to solve these nested integrals, the authors provide a "plug-and-play" formula for family welfare.

Limitations:

• Autarchy Assumption: The model assumes a closed economy without government intervention or external trade.
• Infinite Horizon: It assumes families live/plan forever, which contradicts biological reality.

Future Directions: The authors suggest moving toward Archimedean copulas to model the dependence between risk aversion and discount rates (e.g., do people who have no self-control also tend to take more risks?). This would bridge the gap between pure economic theory and the complex psychological realities observed on social media.

Conclusion

This paper serves as a bridge between Social Network Analysis and Neoclassical Optimization. It reminds us that "The Average Family" is not a single person, but a complex, statistically distributed network of individuals.