Dividend decisions in the property and liability insurance industry: mutual versus stock companies

Lee and Forbes (1980) examine some 30 US P–L stock insurers and find that dividend payout ratios and share prices are closely related. Lee and Forbes (1982) show that the dividend payout ratios of 61 US P–L stock insurers are affected by the previous payouts and ownership structure characterized by widely held versus group-affiliated insurers. Akhigbe et al. (1993) employ an event study methodology in testing the signaling effect of dividend changes. They observe a more significant market reaction to the dividend changes in P–L insurers than in life insurers. They ascribe such a difference in the market reaction to the volatile nature of property and liability business as opposed to the relatively stable life insurance business.

Akhigbe et al. (1993) argue that because the financial statements of insurers are not a reliable signaling mechanism, investors may have to rely heavily on other signals (e.g., changes in dividend policy). This forces many firms to minimize unnecessary changes in the dividend policy in order to avoid undesirable market consequences. Akhigbe et al. (1993) therefore argue that adjusting insurers’ dividends may create an effective signal and they document a positive and significant market reaction to a sample of listed insurers’ announcements of dividend increases.

Insurers may also increase liabilities by taking on more businesses. However, regulation requires that the amount of insurance premiums that an insurer can write is up to certain times of its capital. For example, the famous Kenney rule argues that the ratio of premiums to capital for a P–L insurer should not exceed two. In recent years, there is a declining trend in Kenney ratio in the US P–L industry probably reflecting that insurers leave more financial slack for catastrophes (Cummins and Doherty 2002). Indeed, Smith (2001) reports that over the period 1994–1999, the industry average for US P–L insurers was in the range of 0.84–1.3.

When capital is inadequate relative to liabilities, P–L insurers can also choose to scale down their liability side (i.e., taking on less business) in addition to raising capital externally. However, this is an inflexible and potentially costly solution given its adverse impacts on customers (e.g., see Lee and Forbes 1982). Harrington and Niehaus (2002) also contend that there is no reason to expect the costs associated with adjusting the liability side to vary between mutual and stock P–L insurers.

However, insurance regulation may help alleviate policyholders’ concern about the potential opportunistic behaviors of stock insurers’ managers and stockholders.

See Wells et al. (1995) for a review of the relevant evidence.

Other proxies like dividends per share are not available for mutuals.

The beginning-of-period total assets figure is used to mitigate the possible endogeneity between the dividend payouts and contemporaneous total assets.

Because of this, free cash flow is at most a noisy proxy for testing the owner-manager incentive conflict argument in relation to dividend payouts and managerial perquisite consumption.

Stock insurers may face another type of agency costs in relation to excessive free cash flow—the costs arising from the incentive conflicts between stockholders and policyholders. Krishnaswami and Pottier (2001) argue that stock life insurers can mitigate the stockholder-policyholder conflicts by issuing more participating policies, where policyholders have the right to share insurers’ residual claims. The implication is that stock P–L insurers can also increase policyholder dividends (e.g., by issuing more new participating policies and/or increasing the dividend payouts of existing participating policies) to reduce such agency costs in relation to free cash flow. Therefore, for both mutual and stock insurers, a large aggregate dividend payout can be used to reduce agency costs associated with excessive free cash flow.

In our analysis, only one variable needs figures on total assets and net earnings going back to 1989 (i.e., the volatility of earnings for 1994); however, we believe this should not be a serious concern as insurers’ total assets and earnings are unlikely to be greatly affected by the introduction of the RBC system.

In an early version of the current paper, we use 9,142 firms/years that include both unaffiliated and group-affiliated P–L insurers as our sample and we include a group status as a control variable in our model, we obtained similar results to the results from using unaffiliated insurers.

The effect of organizational form conversions on dividend payouts is an interesting topic for future study.

A Tobit model with fixed-effects cannot be used because of the multicollinearity between the key variable of interest—MUTUAL and fixed-effects. We also tried a Tobit model with random-effects; however it could not converge due to the insufficient intra-company variations in firm-specific error terms. This suggests that a Tobit model with random-effects is not appropriate for our data as the assumption of treating firm-specific error terms as random does not seem to be met (see Greene 1999).

Including these observations generally produced similar results except that in some years, the coefficients on MUTUAL × LAGDIV in Table 4 can be over one (see Sect. 5.2.2).

Myers and Pritchett (1983) argue that in addition to being a way of refunding premium overcharges, dividends issued by mutual insurers also include a profit sharing element.

One may argue that if the current dividend payout ratio is over its long-run target level, a mutual P–L insurer may have incentives to quickly lower its dividend payout ratio to the target level. If this happens, mutual P–L insurers may at least have an equally quick speed of adjustment to their long-run target levels as stock P–L insurers. While this is possible, it is unlikely for mutual insurers to frequently pay dividends higher than their long-run target levels due to their greater capital constraints and higher financing costs. Overall, the effect of the slow adjustment that arises when mutual insurers have a lower-than-target dividend payout ratio is likely to dominate.