The Sarbanes-Oxley (SOX) Act, implemented in 2002, came in response to a wave of corporate scandals, such as Enron and WorldCom, which raised concerns about the integrity of the accounting information provided to the public.

Distrust of Corporate America was at an all-time high—and the insurance industry was most certainly not immune from the public’s ire.  According to Edelman’s 2011 Trust Barometer, among 13 major industries, the insurance sector consistently ranks near the bottom in terms of consumer trust; almost two-out-of three Americans said they don’t trust insurers “to do what is right.”

(Editor’s Note: The just-published 2013 Trust Barometer puts financial services at the bottom again, with only 50 percent of the informed public trusting the banking and financial services industries on a global basis.)

Additionally, in a Bloomberg national poll released in 2010, three-in-five Americans said they have a “very” or “mostly” unfavorable attitude toward insurers. (Bloomberg, Mar. 24, 2010, article titled, “Wall Street Despised in Poll Showing Most Want Regulation.”) The perceived trustworthiness of the insurance industry was even lower than Wall Street.

A similar story is evident at the agency level where annual Gallup polls assessing the trustworthiness of specific occupations consistently rate insurance agents near the bottom of the list—narrowly edging out used car salespersons and elected representatives.

Suffice to say, the insurance industry has some work to do in burnishing its image with the public.

With more than 10 years having passed since the first arrival of SOX, its impact (or lack thereof) on Corporate America is now being assessed, both formally and informally, by the public media, industry pundits, academics, and other researchers. In this article, we provide a high-level overview of our empirical research that examines the impact of SOX on property/casualty insurer use of earnings management practices with regard to the setting of loss reserves. (See related textbox, “Earnings Management Defined.”)

At the aggregate level, the errors associated with loss reserve estimates present a distinct and relatively predictable pattern, as shown in accompanying Figure 1. In theory, errors in such estimates should be reduced to random effects. The fact that any pattern exists at the aggregate level suggests that something is amiss when it comes to the establishing of loss reserves.

Research Explained

Given the unique nature of the business of insurance, claim obligations that have yet to be paid—loss reserves—are the single largest liability entry on the balance sheets of most insurers. Variation in the magnitude of that entry in an insurer’s financial reports has the potential for affecting a variety of other accounting entries on which insurer performance is based.

Insurers typically employ actuaries in developing loss reserve estimates. While those actuaries generally employ rigorous models and control for a wide variety of potential variances in their analyses, their final recommendations to senior insurance management are typically presented as a range of potential and defensible estimates. Senior management must choose, and sign off on, a final value for reporting purposes. Given the inability for anyone to predict the future with specific accuracy, discrepancies between the original estimates and the final actual loss payments are to be expected.

We use loss reserve error patterns in the industry as a litmus test in assessing the ability of SOX to impact earnings management behavior within the property/casualty insurance industry. If SOX has had its intended impact and improved the accuracy and reliability of the financial reporting of publicly traded firms, we should expect any evident patterns in loss reserves to dissipate after 2002.

We use a five-year loss development period in our analysis. Positive values indicate initial overstatement of loss reserves while negative values reflect initial understatements. Additionally, to mitigate the potential influence of the reserving behavior of relatively larger insurers in the analysis the reserve errors were scaled using each firm’s total admitted assets.

The ownership structure within the property/casualty insurance industry presents itself as a good candidate for a comparative research model as it is comprised of three distinct forms of ownership, only one of which is subject to SOX—publicly traded insurers. The other two categories—privately owned and mutual insurers—were not subject to SOX during the time period we studied, which extended from 1998 to 2006.

(In 2012, the National Association of Insurance Commissioners introduced the Model Audit Rule, MDL-205, to the insurance marketplace. The regulation essentially extends the financial reporting goals of SOX to all insurers operating in the United States, regardless of ownership form.)

Decomposing the loss reserve error data presented in Figure 1 above into categories based on the form of ownership provides us with additional significant insight, as shown in Figure 2.

While the same general undulating pattern exists across all three categories of ownership, it is less acute among mutual insurers. Additionally, the average loss reserve error for the mutual insurer group is always positive, i.e. mutuals tend to over-reserve. That behavior aligns itself with statutory accounting principles that encourage management teams be conservative when settling on discretionary estimates for financial reporting purposes, such as loss reserves.

With respect to an insurer’s financial reports, over-reserving increases the insurer’s liabilities and correspondingly decreases its policyholder surplus. Under-reserving has a reverse effect. The modulation of an insurer’s liabilities and policyholder surplus in the form of increasing/decreasing discretionary loss reserves (within acceptable bounds) has the ability to materially affect various ratios by which insurer performance is assessed.

Also evident in Figure 2 is that the pattern of reserve errors of publicly traded and privately held insurers are roughly comparable across the 1998-2006 period; both tend to under-reserve during the pre-2002 period and over-reserve during the latter period. Additionally, while clear evidence as to the relationship shared by the reserve error pattern and form of ownership is obvious in the earlier period, it becomes far less predictable during the post-2002 period. For example, while mutual insurers tend to overstate loss reserves across the entire period under investigation, and were singularly different in this respect during the pre-2002 period, during the post-2002 period all three categories of insurers tend to overstate loss reserves with public and private insurers typically overstating by a larger margin when compared to the mutual insurers.

It is tempting to attribute the uniquely differing pattern of mutual insurers to the absence of shareholder expectations—to suggest that they can afford to be conservative because their operations are not subject to the same level of shareholder scrutiny and performance expectations heaped upon publicly traded insurers.

Privately held insurers represent something of a hybrid; while their structures are not comparable to that of mutual, neither are they publicly traded. Within the context of loss reserving behavior, privately held insurers more closely reflect the behavior of publicly traded insurers.

The availability of various empirical analytic techniques allows us to perform a more sensitive assessment as to the embedded relationships and factors driving these behaviors.

Given that focus of this research is on the ability of SOX to affect the reserving behavior of publicly traded insurers, we aggregate the data of the privately held and mutual insurers into a single category, identified as non-publicly traded insurers, and use their behavior as a point of comparison in assessing the behavior of publicly traded insurers. Given that mutual insurers comprise approximately 22 percent of the data points captured in this analysis, the aggregation of the mutual with the privately held insurer data (which comprises about 56 percent of the data) results in a dampening of the patterns evident in the privately held insurer group when presented in isolation, as seen in Figure 3.

The persistence of the obvious pattern in each of the graphics presented thus far is difficult to ignore and we address it here. Using the maximum and minimum average error values among the three forms of ownership categories presented in Figure 2, we present a graphic in Figure 4 that roughly defines the boundaries of the average loss reserve errors across the 1998-2006 period for the industry as a whole.

If we can assume that the quality of actuarial models is roughly comparable across all insurers regardless ownership form, we can deduce that the senior management of mutual insurers may have employed far more conservative strategies than did their non-public counterparts. Subsequently, the entire industry began to invoke more conservative reserve estimates and with far more homogenization than was evident prior to 2002. However, the existence of any pattern at the industry level speaks to the existence of some environmental factor that was not accounted for in the underlying actuarial models; controls for predictable patterns should have been controlled for in their analysis.

If one can assume that the industry’s actuaries are capable of generating accurate models, then the evident pattern is derived from another, unpredictable, source. A significant body of empirical research points the finger at insurers engaging in earnings management in an effort to massage performance assessments.

In our research, we use a five-year development period in identifying our loss reserve errors, resulting in a period of analysis spanning the 1998-2006 period.

This allows us to compare the reserving practices of the four years prior to SOX with the five years immediately following the implementation of the legislation. Our empirical analysis relies on difference-in-difference regression, a quasi-experimental technique that allows us to compare the behavior of one group with that of another while controlling for a number of other factors. This technique uses a comparative approach where we assess changes in the publicly traded group’s reserving behavior relative to the non-public insurers while simultaneously controlling for changes in that relationship across the pre and post-SOX periods.

Factors for which we provide control in our analysis include:

• Whether an insurer was a mutual
• Group affiliation (if any)
• Insurer size
• Historical rate of return on assets
• Taxes paid
• A.M. Best rating
• Percentage of business in long-tail lines of business
• Use of contingent commissions and reinsurance
• Degree to which the insurer held a diversified portfolio with respect to product lines and geography
• The aggregate industry loss ratio

(Full details of the analysis are available in the current working version of our research paper. The paper is currently under review for publication in an academic journal.)

The key result of our analysis is that SOX has had no statistically significant impact on the loss reserving behavior of publicly traded property/casualty insurers since 2002. The loss reserving behavior of publicly traded insurers has not changed substantially with the advent of SOX. Given the trust issues noted earlier, these findings do little for improving the insurance industry’s image with the general public.

Another possible implication of our findings is that, given the already high degree of regulation to which the insurance is subject, it is possible that already-existing regulation at the state and federal level may already address the issue of earnings management as well as can be expected. With that suggestion, we do not mean to diminish the significance of earning management within the insurance industry or to imply that shareholders should not be appropriately concerned as to the possibility of less than accurate financial reports. Rather, our analysis suggests that SOX has had little impact on the issue at the industry level.