Professor Jiangmin Xu of Guanghua School of Management collaborated with Ioannis Branikas and Harrison Hong to find out the role of proximity in determining households’ stock holdings of firms headquartered near their city of residence. This article was published in the Journal of Financial Economics in 2020.

A long-standing puzzle in financial economics, the local bias puzzle, is that households hold undiversified stock portfolios tilted toward firms headquartered near where they reside. Studies typically assign a causal role for proximity, yet this paper points out that extant empirical work on local bias crucially assumes that households locate randomly, which is likely to be counterfactual. Households may locate based on unobservable factors such as optimism about a city’s economic prospects, which can be correlated with their latent local-stock demand. Thus, this study uses location-choice models to account for this endogenous selection issue.

Background

Existing literature has typically assigned a causal role for proximity, be it a familiarity heuristic whereby investors favor local stocks because they feel competent in evaluating them, or “Keeping up with the Joneses”preferences, which leads to demand for local stocks as a form of hedging. However, the literature neglects that proximity is endogenous. Households naturally prefer to locate in areas which they view as having a bright economic future, not only for themselves but for their family and next generations. These latent expectations driving their location choices are naturally correlated with optimism about local versus distant stocks, to the extent stocks are sensitive to economic conditions of the region of their firm headquarters. This optimism can generate speculative purchases of local stocks or hedging demand. Even though existing empirical studies control for detailed observable outcomes such as demographics or occupational status of households, they cannot account for latent subjective expectations about the economic prospects of different cities. More importantly, households’ latent expectations about cities that they did not locate to play an important role in their stock portfolio choices as well.

In other words, to what extent does proximity play a causal role in local bias and to what extent does it simply reflect selection bias? As a thought experiment, if we were to randomly locate households in different cities, would they still exhibit the same degree of local bias?

Location-choice model

To answer the above questions, the team developed a methodology to account for the effect of endogenous location decisions on household portfolio choice. Household investment data were drawn from the database of a national discount brokerage firm, including 57 metropolitical statistical areas (MSA) and 8,688 households from January 1991 to November 1996. A widely-used portfolio specification was utilized: a non-linear Tobit model where the dependent variable is portfolio weights and the distance between a household’s location and the location of the firm’s headquarters is the independent variable of interest.

Distance is captured using a location choice model from urban economics, where agents optimally locate in cities that provide them with the highest utility. This sorting or self-selection depends on both pecuniary (i.e., productivity) and non-pecuniary (i.e., life, leisure or recreational) household motives. In the context of this study, one of the most important pecuniary latent factors is subjective expectations about the economic prospects of a city, which the team demonstrates to be important for understanding local bias.

Main Results

The research group used the first-stage regression data of 8,688 households as the sample of the optimal location choice model and obtained results like those in the existing literature. That is, urban environment and household characteristics play a vital role in explaining the location choice and preferences. For example, white-collar families with high income or managerial positions are more likely to live in densely populated central areas. Meanwhile, older people and blue-collar families are more likely to live in areas with high unemployment rate.

Accounting for household location preferences, the research team then studied the portfolio choices of households. After correcting for selection bias generated by endogeneity of location choice, the team found a large reduction in the local bias estimates from standard models, which is approximately 30 to 40 percent. Therefore, the findings of the team point to a significantly smaller role for proximity in determining portfolio choices of households compared to those in the existing literature.

Contributions to existing literature

The findings have a number of implications particularly for the household finance literature.

-This method can be used to adjust household portfolio choice regressions involving proximity in other datasets and settings since location choice driven by retirement is common in many countries.

-The household finance literature has focused on how observable household or asset characteristics might influence or bias portfolio decisions. The study explores the less researched latent expectations that are embedded in locational choice decisions.

-While the team focused on stocks, this analysis naturally applies to general portfolio constructions including purchases of homes.

Future Research Venues

One venue for future work is to relate or extend the team’s analysis to an international setting. While international equity home bias is not doubt driven by other factors such as trading costs, the selection bias mechanism analyzed in this study might nonetheless naturally have a role. Some households, particularly the wealthy ones or those working in certain industries, have a choice of living abroad in cities such as London or Tokyo. To the extent one has data on households living abroad and their stock portfolios, research teams can integrate the location-portfolio choice framework into their international settings and examine this issue.

Conclusion

The local bias puzzle is generally explained by theories that assign a causal role to proximity. But households in practice optimally locate in a city depending on latent subjective expectations about the economic prospects of that city, which are correlated with demand for local stocks. Through this study, the team proposed a correction for this selection bias in standard portfolio regressions using location choice models. In short, the team has shown how latent expectations regarding location choice are important for understanding the local bias puzzle in financial economics.

About Author

Jiangmin Xu is an Associate Professor of Finance at Guanghua School of Management, Peking University. He joined Guanghua in 2014. Prior to this, he obtained his Bachelor of Arts degree from the University of Cambridge with the highest honors in the United Kingdom, and earned his Ph.D. in Economics from Princeton University in the United States. At Princeton University, he studied under Professor Yacine Ait-Sahalia, founder of the Bendheim Center of Finance at Princeton; Professor Harrison Hong, winner of the 2009 Fischer Black Prize; and Christopher Sims, winner of the 2011 Nobel Prize in Economics. His main research areas are climate finance, climate economics, behavioral finance, and financial econometrics. His research results have been published in top international economics and finance academic journals such as theJournal of Financial Economicsand theJournal of Econometrics, and he has won the Best Paper Award of the China International Conference in Finance. He has been invited many times to present academic papers at the Annual Meeting of the American Finance Association, the Annual Meeting of the Western Finance Association, the Annual Meeting of the European Finance Association, and China International Conference in Finance.