research

Cancer and portfolio choice: evidence from Norwegian register data

A TFI research project by Trond Doeskeland & Jens Kvaerner
Posted on October 26, 2020

"Cancer and portfolio choice" is one of the research projects supported by the Think Forward Initiative.

In this research, Trond Doeskeland and Jens Kvaerner investigate how negative health shocks affect portfolio choices. More specifically, they look at investment decisions after a cancer diagnosis or after the loss of a partner. How do these events affect households' risk taking behavior and stock market participation?

Download the report or read the summary below to find out.

SUMMARY

As we live longer in a world that is increasingly globalized, we'll all experience a certain health risk at some point in our lives. Longer life expectancy thanks to advanced biotechnical research and treatment options means that more people will fully recover from previously fatal health shocks. International trade and traveling will continue to raise the pace at which new infectious diseases (like COVID-19) spread. Given the prevalence of health shocks, it is important to understand how, when, and for whom such negative life events affect personal investment decisions. And that's exactly what we focus on in our research.

The negative life events we picked for our study are cancer diagnosis and widowhood. Both events affect people of all genders, both young and old, rich and poor, and can have a major impact on earnings, mortality, family dynamics, roles, priorities and mental health, which are variables typically tied to individuals’ and households’ investment styles.


Research objective

The literature on the importance of health in financial decision-making - such as how much to invest in the stock market - has struggled with two challenges: the unavailability of large-scale household-level data on health, family links and financial outcomes, and the difficulty of isolating the causal effects of health shocks in the presence of complex dynamics and systematic measurement error. Systematic measurement error may reflect a justification bias: respondents may report a worse subjective level of health to justify their current economic status. As a result, it isn't enough to rely solely on subjective measures of health status and self-reported measures of wealth.

In our study, we tackle these challenges. We are particularly pleased with our dataset, which is the result of merging anonymized individual medical records with administrative data on financial and demographics for the Norwegian population from 2005 to 2013. Thanks to this individual-level data, we are able to investigate how financial decision-making depends on the severity of cancer, cancer type, family income, and socio-demographic variables such as gender and having children.

Furthermore, we developed a simple identification strategy to deal with confounding. Smoking, for instance, can affect the likelihood of developing cancer and correlate with personal investment decisions. However, such confounding is cancelled out in our research design because we identify the effect of cancer on investment decisions by comparing individuals and households who were all diagnosed with cancer - but at a different point in time. Finally, because all cancer treatment costs are covered in Norway, the investment behaviour we identify reflects individuals' responses to the negative life event itself, including mental/somatic aspects and income effects.


Summary of results

The analysis of about 70,000 households - over a period of 7 years in which one household member experiences cancer or widowhood - allows us to draw three main conclusions:

1. A cancer diagnosis increases the likelihood of exiting the stock market two to three years after the diagnosis, but has little impact on the composition of the financial wealth of those that stay in the stock market.

2. Widowhood increases the probability of a stock market exit by a factor of ten, relative to the effect of the cancer diagnosis itself.

3. The effects are larger overall if the affected individual in the household is also the primary-earner, as measured by the individual’s contribution to family income prior to the event.

4. There is no link between prior experience with cancer and current personal investment strategies.

The first finding is in line with previous studies using all types of health shocks, and mainly based on survey data. The second and third findings are in line with our expectations: losing a spouse has a major negative impact on a household’s current and expected income, which in turn forces the household’s consumption closer towards a minimum consumption level. It also makes the household’s consumption more dependent on other income sources. In order to reduce the increase in sensitivity towards asset price fluctuations, the household takes less risk. These effects are amplified if the affected individual was also the primary-earner. The fourth conclusion is based on our analysis of historical records, which allowed us to investigate whether personal experience with cancer (many up to 10 years ago) is associated with stock market participation and composition of financial wealth.


Conclusion

In order to maximize the welfare benefits of financial markets, it is critical to ensure that households invest effectively. Our results have practical implications for financial advisors and households that will experience adverse health outcomes in the future. For financial advisors, it is important to understand what triggers the surviving spouse to exit the stock market. For households, more information about what other people do in similar vulnerable situations may provide guidance on how to proceed.

Our research can be expanded in several academic and non-academic directions. The first step is to evaluate the external validity of this study, investigating whether our results apply outside Norway and across various diseases. Secondly, as already indicated, it would be interesting to explore not only whether negative health shocks affect financial choices, but also why they do so. We hypothesize four key channels: mortality, wages, medical expenditure, and the ability to work. Thirdly, our ongoing research shows that negative health-related lifetime events can lead to a major reduction in family income and increase the likelihood of liquidity constraints. Building on these insights, it would be interesting to investigate the effect of health crises on personal bankruptcy. One could, for instance, identify the key variables that explain variation in realized bankruptcy for a given health crisis.