Investment risk taking by institutional investors

Abstract

According to theory, institutional investors face both risk-management and risk-shifting incentives. This article assesses the relevance of these conflicting incentives for Dutch pension funds and insurance firms over the period 1995 to 2009. Using a unique and extended data set, we observe a significant positive relationship between capital and asset risk for insurers, indicating that risk-management incentives dominate in the Dutch insurance industry. Risk-shifting incentives, however, also seem relevant, as stock insurers take more investment risk than their mutual peers. For Dutch pension funds, we conclude that overall neither risk-shifting nor risk-management incentives seem to dominate. Interestingly, we find that professional group pension funds take significantly less investment risk than other types of pension funds. This finding is in line with expectations, as in professional group pension funds potential incentive conflicts between pension fund participants and the employer are effectively internalized.

Keywords:

• portfolio choice
• insurance firms
• pension funds
• ownership structure

JEL Classification:

• G11
• G22
• G23
• G32

Acknowledgements

The authors are grateful to Dirk Broeders, Paul Cavelaars, and the participants of the EGRIE 2011 conference in Vienna. We thank Jack Bekooij for his excellent research assistance. The views expressed in this article are personal and do not necessarily reflect those of DNB.

Notes

1. In a study of the regulation of institutional investors in the major OECD countries, Davis (2002) reports that only the United Kingdom and the Netherlands do not impose quantitative restrictions on equity holdings for life insurers.

2. See Badrinath and Wahal (2002) for a relatively recent literature overview.

3. In this article, the capital ratio is an important proxy of an investor’s risk-bearing capacity. The capital ratio is defined as net asset value (i.e. assets minus liabilities) to total assets. While common in banking, the capital ratio is not the typical solvency indicator in the insurance and pensions industry. In the insurance industry, solvency conditions are usually presented in terms of the solvency ratio. The solvency ratio equals the actual solvency margin divided by the required solvency margin. In the context of pension funds, solvency conditions are typically presented in terms of the funding ratio, which is the ratio of total assets to total liabilities. For ease of comparison, however, we use capital ratios for both insurers and pension funds.

4. This will change when the European Union will implement a new, risk-based regulatory approach for insurance companies, called ‘Solvency II’.

5. For life insurers, we use data from the Solvency II QIS5 study to establish a sample correlation of 0.5 between interest-rate risk and equity risk, where both risks are approximated by the respective solvency capital requirements scaled by total investments. For pension funds, we use comparable data, but then from the Dutch financial assessment framework data, to establish a sample correlation of 0.4 between interest-rate risk and equity risk.

6. Note that nonlife insurers are not permitted to write life insurance policies, and, vice versa, life insurers are not permitted to write nonlife insurance policies. That being so, insurance holding companies are allowed to have both life and nonlife subsidiaries. A nonlife insurer is predominantly active in a specific nonlife insurance line-of-business when more than 50% of its premium income comes from this line. When a specific nonlife insurer has multiple lines of business and is not predominantly active in any line of business, the line-of-business dummies are all 0 for this insurer. For life insurers, our data set discerns two lines-of-business: traditional life insurance and unit-linked life insurance. Given the long-term nature of life insurance, we distinguish between both types of life insurers by reserves than by premium income. A life insurer is predominantly active in traditional life when more than 50% of its reserves are traditional life reserves.

7. The data set does not include funeral insurers and tiny mutual nonlife insurers exempted from supervision.

8. The raw data set contains 18 416 institution-year observations. We have excluded defined contribution (DC) pension funds (573 observations), as they would confound our analysis. In DC funds, the investment risk is typically borne by the pension fund participants and not by the pension fund itself. We have also excluded observations that have zero or negative total assets, negative equity allocation, equity allocation >100%, capital ratio >1 or capital ratio <−0.3. Note that our data source does not distinct between zeroes and missing values.

9. The average equity-allocation difference between pension funds and insurance firms is picked up by the insurance dummy. Since the insurance dummy is interacted with other dummy variables (i.e. STOCK, LOB. and G), the equity-allocation difference between insurers and pension funds is a weighted sum of the estimated coefficients of α₃ and coefficients . Note that coefficients α₄, α₅, and α₁₄ do not play a role in this respect, since the corresponding variables are in deviation from their respective sample means.

10. These results are not shown but available on request.