Three Essays in Household Finance

Abstract

This thesis explores the impact of two behavioural finance concepts, social psychology and psychology, on household financial decisions. Under social psychology, I investigate whether the variety and intensity of social engagement enhances stock market participation. With regard to psychology, I examine two behavioural biases. First, I investigate whether mental accounting influences portfolio choice in three asset classes and whether financial advice and housing tenure increase (decrease) the effects of mental accounts on portfolio choice. Second, I examine whether households’ self-reported housing wealth are anchored on published house price indices and whether anchoring bias is mediated by market information, mortgage refinancing decisions and social factors.

The main contributions and findings in the three studies are as follows. First, although there is an elaborate body of research concerning the relationship between social engagement mechanisms and portfolio choice, most studies investigate specific mechanisms in isolation. Using three waves in the British Household Panel Survey (BHPS), I bring together five social engagement measures in one model and show that socially engaged individuals are more likely to participate in the stock market. Consistent with Granovetter’s (1973) theory of social networks I find that a weak tie (measured by social group involvement) has a positive effect on stock market participation whereas a strong tie (measured by talking to neighbours) has no effect. More trusting individuals are more likely to participate in the stock market, as are those who identify with a political party. In contrast, the degree to which religion is important appears to have little impact. These results are robust using different specifications. Overall, the results of this study demonstrate that the likelihood of stock market participation increases with the variety and intensity of social engagement.

Second, despite the established theoretical underpinnings of mental accounting in behavioural portfolio theory (BPT) and recent extensions, not much is known about their implications in real life situations. I use a recent UK household survey, the Wealth and Assets Survey (WAS), which has comprehensive information about financial assets to investigate whether there are differences in the ownership and portfolio share of three asset classes among individuals who exhibit no mental account, a single mental account and multiple mental accounts, and the conditional influences of financial advice, housing, cognitive ability, time preference and risk tolerance. Overall I find that mental accounting together with financial advice and housing tenure explain variations in both the probability of ownership and portfolio share in the three asset classes. Households that exhibit a single mental account have low share of investments in, and are less likely to own, a risky asset when compared to those that exhibit no mental account or exhibit multiple mental accounts. I also find that, when compared to having no mental account, exhibiting a single mental account or multiple mental accounts increases both the probability and investment share in a fairly safe asset but decreases portfolio share in safe assets. In addition, among those that exhibit a single mental or multiple mental accounts, financial advice decreases portfolio share in risky assets and fairly safe assets and increases portfolio share in safe assets. Housing tenure increases both the probability and portfolio share in risky assets, decreases portfolio share in fairly safe assets and increases portfolio share in safe assets. These results are consistent using multi-equation regressions, sub-samples, reparametrised variables and poisson regressions.

Finally, as little is known about how households derive the self-reported house prices estimates that are commonly used to determine housing wealth, the third study examines whether households are anchored on published house price indices. The key conjecture is that, while assessing the values of their homes, homeowners place more weight on house price news at the expense of property characteristics and other market information. I find support for this hypothesis using sixteen waves of the BHPS, multiple methods, and both regional and national house price indices. I conclude that changes in self-reported housing wealth are anchored on changes in published house price indices. Specifically, ownership through a mortgage and greater financial expectations increase anchoring effects while mortgage refinancing decreases the effects. Moreover, use of money raised from refinancing for home investment, as opposed to other consumption purposes, has a positive association with change in self-reported house value and both uses reduce anchoring bias. In addition, I find that computer use increases anchoring bias and, among social engagement mechanisms, religiosity reduces anchoring while other measures have no effect. These results are robust to internal instrumental variables, national aggregate house prices, alternative indices and sub-samples.