RATINGS: IT'S ACCRUAL WORLD.
Insurer financial strength ratings have been studied by academics due to the usefulness of ratings to regulators, consumers, corporations, agents/brokers, and insurers. Ratings largely determine the price insurers can charge for their product, provide an indication of a firm's risk of insolvency, and convey new information to capital markets (Halek and Eckles, 2010). While prior studies have investigated the determinants of financial strength ratings (e.g., Doherty and Phillips, 2002), these studies have not considered accounting quality as a determinant of ratings.
In this article, we examine whether accruals quality is related to a firm's financial strength rating. Eckles, Halek, and Zhang (2014) provide evidence that insurers with higher reserve error volatility have greater information risk. We suggest that insurers with higher reserve error volatility have lower quality, or noisy, accruals. Thus, based on the foregoing, we examine two research questions. First, we investigate whether overall earnings quality, as measured by the noisiness (standard deviation) of an insurer's loss reserve errors, is related to the firm's financial strength rating. Thus, the more volatile an insurer's loss reserve errors, the poorer is the insurer's accruals quality. Second, we decompose accruals quality into two components, discretionary and innate, to determine whether the ratings effect differs between the two types.
This topic is of interest to researchers and several other parties. While prior research has sought to identify the relative importance of various factors in A.M. Best's (hereafter, "Best") ratings process, no prior study has examined the possible role of the quality of accounting information in assigning ratings. In addition, prior studies find evidence that insurers can use discretionary accruals in an attempt to mask solvency issues (e.g., Petroni, 1992; Gaver and Paterson, 2004; Grace and Leverty, 2012). Therefore, if regulatory solvency monitoring is based on regulatory ratios that are sensitive to accounting manipulation, Best could have an advantage in detecting insolvency if Best incorporates accruals quality into its ratings. (1)
Using an ordered probit model, we model financial strength ratings as a function of accruals quality and variables to control for various factors, including proxies for firm risk of insolvency, for a large sample of property--liability insurers from 1993 to 2006. As mentioned above, we measure accruals quality in terms of the volatility (the standard deviation) of loss reserve errors. We also decompose accruals quality into innate and discretionary accruals (Francis et al., 2005; Eckles, Halek, and Zhang, 2014) in an effort to examine differential ratings effects.
We offer two main results. First, we find evidence that financial strength ratings are positively associated with accruals quality. That is, as measured by loss reserve error volatility, insurers with higher quality (less noisy) accruals receive higher Best ratings. This result suggests that Best is able to detect poor accruals quality and assigns a lower financial strength rating to these firms having higher information risk Second, we find that reduction in financial strength ratings tends to be stronger for poor innate accruals quality relative to discretionary accruals quality. This finding suggests that Best perceives greater insolvency risk arising from poor innate accruals (uncontrollable by the firm) relative to discretionary accruals (controlled by managers).
We make several contributions to the literature. First, we contribute to the literature investigating insurer financial strength ratings. We are the first to link accounting quality when examining the determinants of financial strength ratings and find that accruals quality incrementally explains insurer ratings. Second, we contribute to the literature on insurer loss reserve errors. While the majority of extant literature ex-amines reserve manipulation in the context of earnings management (e.g., Petroni, 1992; Anthony and Petroni, 1997; Beaver, McNichols, and Nelson, 2003; Eckles and Halek, 2010; Grace and Leverty, 2012), we utilize the volatility of loss reserve errors as a measure of accruals quality (e.g., Eckles, Halek, and Zhang, 2014). A number of prior studies, such as Gaver and Paterson (2004) and Grace and Leverty (2012), have examined the role of financial performance in reserve management. We extend these studies by first specifically examining the role of accruals quality in financial strength ratings. Second, we examine the role of accruals quality, as measured by reserve error volatility, as opposed to earnings management. To that end, we also extend the evolving literature examining the implications of the volatility of reserve errors, specifically (Eckles, Halek, and Zhang, 2014), and accruals, generally (Francis et al., 2005). Eckles, Halek, and Zhang (2014) and Francis et al. (2005) show that accounting quality affects a firm's ability to raise capital. We show that in addition to higher capital costs, poor accruals quality is also penalized by rating agencies. In addition, our results are of interest to regulators who are responsible for solvency monitoring of insurance firms. Prior literature has found evidence that financial strength ratings perform better in insolvency identification relative to measures used by regulators, such as the risk-based capital ratio (e.g., Cummins, Grace, and Phillips, 1999). The methods used by regulators are generally ratios, which do not account for potential earnings manipulation. One potential explanation for why Best outperforms regulatory ratios is that Best ratings might incorporate the quality of earnings. Our results demonstrate strong and consistent empirical evidence that Best incorporates accounting quality into its ratings, a result that has hitherto been unexamined.
The rest of the article proceeds as follows. The "Background" section provides institutional details and highlights prior literature on both insurer loss reserves and insurer financial strength ratings. The "Hypothesis Development" section develops our hypotheses. The "Research Design" section describes our data and explains our empirical strategy. The "Empirical Results" section presents our results and the "Conclusion" concludes the article.
Insurerloss reserveerrors generally have been used as a measure of managerial discre-tion in the accounting and insurance literature (e.g., Petroni, 1992; Beaver, McNichols, and Nelson, 2003; Grace and Leverty, 2010). Recent work also has applied loss reserve errors to measuring accounting information quality (Eckles, Halek, and Zhang, 2014). Loss reserves are usually the largest liability on a property--liability insurer's balance sheet, representing the estimated cost of settling claims. In general, a firm's actuaries will present a recommended range of acceptable loss reserves, with management choosing the ultimate loss reserve. As claims occur and develop over time, an insurer will revise the original loss reserve estimate. These revisions, called development, indicate whether the insurer initially under or over reserved. An insurer under reserved if the original loss reserve was less than the developed reserve and over reserved if the original loss reserve was greater than the developed reserve. This information, as well as information on the settlement of claims, is reported to the National Association of Insurance Commissioners (NAIC) in annual statutory filings on Schedule P.
An excerpt from Schedule P can be found in Table 1. These data are used to construct the loss reserve error as follows:
[Error.sub.i,t] = Incurred [Losses.sub.i,t] -Incurred [Losses.sub.i,t+n]. (1)
This error is calculated as the initial loss reserve estimate in year t minus the total incurred losses in year t + n. The sum of the boxed values under column 6 in Table 1 are the incurred losses in year t and the sum of the boxed values under column 11 are the incurred losses in year t + n. The error, used in previous studies (e.g., Beaver, McNichols, and Nelson, 2003; Gaver and Paterson, 2004; Grace and Leverty, 2010), will be positive if the initial loss reserve estimate is overestimated and negative if the initial loss reserve is understated. (2) Consistent with the majority of prior literature (e.g., Petroni, 1992; Beaver, McNichols, and Nelson, 2003; Grace and Leverty, 2010), we calculate 5-year reserve errors, so n is 5. To control for insurer size and to express the loss reserve error as a percentage, this difference is scaled by total assets. (3)
McNichols (2000) suggests there are several advantages to using loss reserve errors as a measure of earnings management compared to other accruals-based measures. For one, it is a material accrual, as the loss reserve generally is the largest liability on an insurer's balance sheet. Also, due to reporting requirements, the development of loss estimates over time is observable, allowing for the comparison of initial estimates to the original accounting estimate. The discretionary manipulation of loss reserves has been studied frequently in the literature as a result of its strength as a measure of earnings management. Loss reserve errors have been linked to various incentives such as earnings smoothing (Weiss, 1985; Grace, 1990; Beaver, McNichols, and Nelson, 2003), financial weakness (Petroni, 1992; Gaver and Paterson, 2004; Grace and Leverty, 2012), and executive compensation (Eckles and Halek, 2010; Eckles et al., 2011; Eastman et al., 2014).
Weiss (1985), Grace (1990), and Beaver, McNichols, and Nelson (2003) provide empirical evidence that insurers manipulate loss reserves in order to smooth income. Notably, Beaver, McNichols, and Nelson (2003) find that insurers with small positive profits tend to significantly understate the loss reserve compared to firms with small negative profits. This is consistent with firms managing earnings in an attempt to avoid reporting losses. While they find that public and mutual insurers engage in earnings smoothing, they do not find evidence of private firms smoothing income.
Prior studies have also documented that financially weak insurers undertake income-increasing loss reserve management. Petroni (1992) finds that financially weak firms, as measured by Insurance Regulatory Information System (IRIS) ratios in the "unusual" range, tend to under reserve compared to financially strong firms. Gaver and Paterson (2004) find that firms manage reserves in order to avoid triggering a fourth IRIS ratio violation, as this trigger would require regulatory action. Firms can improve certain ratios by under or over reserving, so firms with three ratios are found to manage reserves to a greater extent than firms with fewer than three violations. Grace and Leverty (2012) estimate a predicted probability of insolvency using a hazard model and find that insurers with a higher probability of insolvency tend to under reserve.
Studies have linked loss reserve manipulation to executive compensation. Eckles and Halek (2010) find evidence that managers of publicly traded firms manipulate reserves in a way consistent with increasing their bonus pay and stock option compensation Eckles et al. (2011) find that strong corporate governance--measured by board independence, board size, and CEO/chairman duality--can limit the ability of managers to extract additional compensation through reserve management. Eastman et al. (2014) find evidence consistent with Eckles and Halek (2010), suggesting that managers of stock firms manage reserves as they receive more incentive-based bonus compensation. However, Eastman et al. (2014) also find that managers of firms organized as mutuals do not manage reserves in a manner consistent with maximizing their overall bonus compensation.
The auditing literature has also used loss reserve errors as a measure of managerial discretion. Gaver and Paterson (2001) and Grace and Leverty (2013) examine whether managers can use discretionary accounting practices in the presence of high-quality external auditing, as measured by Big N auditors and the presence of an external actuary. They find that the ability of external monitoring to decrease loss reserve errors requires both high quality, in the form of a Big N audit firm, and expertise, in the form of a Big N actuary. Since loss reserving is specific to the insurance industry, the expertise of an actuary is an important component of effective auditing. Gaver and Paterson (2007) find that financially weak insurers under reserve less when the insurer is economically important to its audit firm.
Accounting Information Quality
While the majority of prior studies investigating loss reserve errors use them as a measure of earnings management(e.g., Petroni,1992;Gaver andPaterson,2001;Eckles and Halek, 2010; Grace and Leverty, 2012), Eckles, Halek, and Zhang (2014) use loss reserve errors to measure accounting quality. (4) Specifically, they examine whether poorer accounting quality, as measured by the standard deviation of loss reserve errors over the past 5 years, is associated with a higher cost of debt and equity capital, measured by the price of insurance and beta, respectively. They find evidence that lower accruals quality (higher reserve error volatility) is associated with a higher cost of debt, but find no evidence that accruals quality is priced into equity capital.
The accounting literature examining the consequences of accruals quality is expansive (see Dechow, Ge, and Schrand, 2010, for an excellent survey of the literature). One study of note is Francis et al. (2005) who examine whether accruals quality--as measured by the 5-year standard deviation of residuals from a regression relating cash flows to current accruals--is priced into debt and equity capital. One of their proxies for cost of debt capital is S&P's debt rating. (5) They find evidence that lower accruals quality is associated with a lower S&P debt rating. While this is similar in spirit to our study, we improve upon and extend the literature in at least three important ways. First, Francis et al. (2005) examine only total accruals quality and do not decompose accruals quality into innate and discretionary components in their analysis of debt ratings. Second, we use insurer loss reserve errors as our measure of accruals quality as opposed to residual-based models of accruals (McNichols, 2000). Third, as discussed in the next section, we examine financial strength ratings (of firms) instead of debt ratings (of individual securities).
Financial Strength Ratings
Best has provided financial strength ratings of insurers since its incorporation in 1899. These ratings represent Best's opinion on an insurer's ability to continue to pay claims to policyholders in the future. Unlike debt ratings, these ratings are comprehensive and represent an overall assessment of each firm instead of a single security. (6) Prior studies have found empirical evidence that lower ratings are associated with a higher probability of insolvency (Doherty, Kartasheva, and Phillips, 2012) and that financial strength ratings are superior to regulatory measures of risk (e.g., RBC ratios) in predicting insolvency (Pottier and Sommer, 2002).
Ratings are important to insurance firms and also to a variety of stakeholders, as discussed earlier. Doherty and Phillips (2002) find evidence that firms increased their capital during the 1990s in order to maintain their ratings. Epermanis and Harrington (2006) find evidence that firms experiencing a ratings downgrade see a decline in net premiums written. Halek and Eckles (2010)find evidence that ratings downgrades are associated with a significantly negative stock market reaction. They find that this decline following a ratings downgrade is larger in magnitude than the positive reaction following an upgrade. They also find that abnormal market reactions are worse if a firm loses a rating of A-, consistent with certain corporate customers or brokers being unwilling to purchase coverage from insurers with a rating lower than A- (Pottier and Sommer, 1999). Wade, Liebenberg, and Blau (2016) find evidence that short selling for insurers who are about to experience ratings downgrades increases prior to the announcement of their downgrade. They find that this effect is particularly strong for firms with more transparent balance sheets.
While Best provides some guidance as to what it considers in ratings, it does not provide the exact formula it uses in its ratings. Since this is the case, certain studies have examined the determinants of financial strength ratings (e.g., Pottier and Sommer, 1999; Park and Xie, 2014). While these studies find evidence that firm characteristics are significant determinants of Best financial strength ratings, we provide the first study to examine whether accruals quality appears to be a determinant of financial strength ratings.
Accounting quality has been largely ignored in research examining insurer ratings. As noted above, Grace and Leverty (2012) link reserve manipulation with a higher probability of insolvency. It follows that if insurers that engaged in more active reserve manipulation are at greater risk for insolvency, then we might observe a relationship between reserve manipulation and insurer ratings. Of course, there are multiple components to reserve manipulation giving rise to differing solvency-related outcomes. On the intensive margin, firms with aggressive under reserving may well be putting themselves in a more precarious financial situation that could lead to a higher likelihood of insolvency (Grace and Leverty, 2012). Conversely, insurers that over reserve may well be more conservative, and less likely to become insolvent. The broad use of reserving practices, that is, on the extensive margin, also provides information. That is, if firms always under/over reserve, then stakeholders (in particular, rating agencies) can either explicitly or implicitly "correct" the reserves to obtain a more accurate view of the firm. However, if a firm is inconsistent with its reserving practices, then stakeholders will find it more difficult to have an accurate picture of the financial health of the firm. We consider higher variability of accounting results to be noisy and an indication of "low-quality" accounting. (7) Thus, our first hypothesis predicts that firms with more variability in their reserve errors (poorer accounting quality) will have lower ratings. Formally, our first hypothesis is as follows:
H1: As the accounting quality of an insurer decreases (improves), the financial strength rating of the insurer also decreases (improves).
Further, as discussed above, we can also decompose the accounting quality measure into an innate component and a discretionary component. As noted in Eckles, Halek, and Zhang (2014), the innate component measures the degree to which the quality of the accounting results are uncontrollable (given the lines of business the insurer writes). That is, the reserves may well be very difficult to set correctly. For example, firms entering new lines, firms entering new geographic regions, and firms writing more long-tailed lines may find it inherently more difficult to set reserves. The innate measure of accounting quality would, thus, be more indicative of the permanent uncertainty of the firm. Conversely, the discretionary component measures the degree to which managerial discretion is used in the accounting process. Guay, Kothari, and Watts (1996) note that managers have incentives to both improve and reduce the quality of the accounting results. (8) Regardless of managerial incentive to improve or reduce the quality of the accounting results, we hypothesize that the discretionary component is more transitory in nature and will have less of an impact on a firm's rating (relative to the permanence of the innate component). Formally, our second hypothesis is as follows:
H2: The effect of the innate (i.e., permanent) component of accounting quality on the financial strength rating will be stronger than the effect of the discretionary (i.e., transitory) component of accounting quality.
Ultimately, we expect to find a relationship between accounting quality and a firm's financial strength rating. At a base level, a firm with poorer accounting quality should obtain a lower financial strength rating (H1). Then, we further may observe a difference in the degree to which the ratings penalty for accounting quality (or lack thereof) is associated with either an innate difficulty associated with setting reserves or a more discretionary component (H2).
Our initial sample consists of property--liability insurers operating in the United States with data available in the statutory reports filed annually with the National Association of Insurance Commissioners (NAIC) from 1991 to 2011. We require that a firm have an Best financial strength rating to be included in the analysis. We require 5 lead years of data to construct 5-year loss reserve errors. (9) Additionally, we require various lags to construct our accruals quality measures, as described in the next section. Thus, our analysis sample consists of firms from 1993 to 2006. In our broadest sample, this consists of 10,217 firm-year observations and 1,682 unique firms.
We first define accruals quality as the standard deviation of past loss reserve errors as in Eckles, Halek, and Zhang (2014) who use the standard deviation of the past 5 years of loss reserve errors. For robustness, we also use 3- and 4-year standard deviations. We additionally consider a longer horizon with the standard deviation of up to 10 years (with a minimum of 5 years). The measure of accruals quality is given as follows:
[mathematical expression not reproducible]
Equation (2) provides our measures of overall accruals quality, A[Q.sub.j,i, t] (j [member of] [3,4,5,10]) for each firm i in year t. As described earlier, a higher standard deviation indicates noisier information and thus lower accruals quality.
We next decompose accruals quality into innate and discretionary components. We use the methodology of Eckles, Halek, and Zhang (2014) who adapt the methodology of Francis et al. (2005) to insurer loss reserve errors. Specifically, we estimate the following regression (going forward, we will suppress the j subscript on AQ, though we estimate the models below for each of our measures of accruals quality (i.e., AQ3, A[Q.sub.4], A[Q.sub.5], and A[Q.sub.10]):
A[Q.sub.i t] = [gamma]0 + [Ysub.1][Size.sub.i,t ]+ [Y.sub.2[sigma] (CF)i t + [Ysub.3[sigma]] (Total Premium)i, t
+ [y.sub.4]Neg[Earn.sub.i,t] + [[member of]-i,t,] (3)
where [AQ.sub.i t] is the standard deviation of firm i's loss reserve error scaled by assets over the past 5 years. [Size.subi,t] is the natural log of firm i's total assets in year t. a [(CF).sub.i, t] is the standard deviation of firm i's cash flows from operations (CFs) over the past 10 years. [sigma] [(Total Premium).sub.i, t] is the standard deviation of firm i's net premiums written over the past 10 years. [NegEarn.sub.i, t] is the total number of times firm i had negative earnings over the past 10 years. For [sigma] [(CF).sub.i t] and [sigma] [(Total Premium).subi, t,] we use 10 years when possible but also include firms with at least 5 consecutive years of data prior to year t in order to minimize loss of observations. (10)
To obtain a measure of innate accruals, we calculate the fitted values from Equation (3) as follows:
Innate[AQ.sub.i,t] = [gamma]0 + [Y.sub.1] Sizei,t + [Y.sub.2][sigma] [(CF).sub.i, t] +[??][sigma] [(Total Premium).sub.i, t]
+ [[??].sub.NegEarn.sub.i, t] (4)
where all variables are defined above. The residuals from Equation (3) are used as our measure of discretionary accruals quality:
Disc[AQ.sub.it] = [[member of].sub.i,t]. (5)
Innate[AQ.sub.i,t] provides a measure of the quality of accruals that are "innate" to the firm, meaning they exist simply by the nature of a firm's operations. As an example, firms operating in long-tailed lines of business are likely to have lower innate accruals quality as estimating losses for these lines is more difficult (i.e., due to the long period of time between the claim and the payment, it can be difficult to estimate the amount of the actual payment). This will result in larger reserve errors (in absolute value) simply due to the difficulty of accurately reserving. Disc[AQ.sub.i,t] provides a measure of discretionary accruals quality. This measures the quality of accruals subject to managerial discretion. If managers of insurance firms are manipulating reserves in response to various incentives--such as maximizing their own compensation (e.g., Eckles and Halek, 2010)--Disc[AQ.sub.i,t] will be higher, reflecting poorer accruals quality. We note that the interpretation of all three measures of accruals quality (AQ, InnateAQ, and DiscAQ) have the same interpretation in that higher values of these variables indicates lower accruals quality (for overall, innate, and discretionary accruals, respectively).
Table 2 provides a summary of the Best financial strength ratings in our sample. To create our Rating variable, we categorize ratings into five groups. (12) Rating takes a value of 4 for ratings of A++ and A+, 3 for a rating of A, 2 for a rating of A-, 1 for a rating of B++ and B+, and 0 for ratings that are B or lower. Ratings are generally evenly distributed between the top three rating categories with slightly more firms (approximately 30 percent) rated A. Less than 6 percent of ratings in our sample are B or less, which Best classifies as "vulnerable."
Table 3 provides descriptive statistics for our sample. The average Rating for firms in our sample is 2.5436, which translates to a rating between A- and A. The median firm has a rating of A. The average firm's loss reserve errors have a 5-year standard deviation of 0.0371 ([AQ.sub.5]). The mean and median of Innate[AQ.sub.5] are 0.0371 and 0.0363, respectively. On average, the values of Innate AQ are larger than the values of Dis-cAQ .The average value of RE is 0.0130, which indicates that the average firm in the sample over reserved. A t-test indicates that RE is significantly different from zero (p-value<0.0001), which is consistent with summary statistics on reserve errors reported in prior studies (e.g., Grace and Leverty, 2010, 2012). (13) This result could indicate that firms build a "safety loading" into their original loss reserve estimate or that the incentives to over reserve are greater compared to the incentives to under reserve (e.g., taxes, certain IRIS ratios, rate regulation).
Table 4 provides unconditional correlations between our Rating variable and our various measures of accruals quality. The bottom triangle provides Pearson correlations while the upper triangle provides Spearman correlations. Bolded values are statistically significant at the 1 percent level. For both Pearson and Spearman correlations, as accruals quality improves (i.e., lower values for [AQ.sub.3], [AQ.sub.4], [AQ.sub.5], [AQ.sub.10], Innate[AQ.sub.3], Innate[AQ.sub.4], Innate[AQ.sub.5], Innate[AQ.sub.10], Disc[AQ.sub.3], Disc[AQ.sub.4], Disc[AQ.sub.5], and Disc[AQ.sub.10]) financial strength ratings tend to increase. This is consistent with our hypothesis that lower accruals quality will be associated with lower financial strength ratings.
Table 5 provides univariate descriptive statistics of our accruals quality variables sorted by Rating. The accruals quality variables all tend to improve (decrease) as Rating increases, in many cases monotonically. This result again suggests that accruals quality is associated with a higher financial strength rating. In the last column of Table 5, we test the difference in the accruals quality variables between the highest-rated firms (Rating=4) and the lowest-rated firms (Rating= 0). For all 12 of the accruals quality variables, the highest-rated firms have higher accruals quality compared to the lowest-rated firms at the 1 percent level. This relation provides preliminary evidence for our hypothesis that poor accounting quality is associated with lower financial strength ratings. In the next section, we perform multivariate tests of our hypotheses.
Determinants of Ratings
To test our hypothesis (H1) of whether accruals quality impacts an insurer's financial strength rating, we employ the following model: (14,15)
[mathematical expression not reproducible](6)
where i,t = Firmiinyear t; [Rating.sub.i,t] = Firm i's Best financial strength rating in year t, where 4 corresponds to ratings A++ and A+, 3 corresponds to rating A, 2 corresponds to rating A-, 1 corresponds to ratings B++ and B+, and 0 corresponds to all lower ratings; [AQ.sub.j,i,t] = The standard deviation of firm i's 5 -year loss reserve error over the past (j = 3, 4, 5, 10) years relative to year t; (16) Surplus-to-[Assets.sub.i,t] = The ratio of firm i's policyholder surplus to total assets in year t; [Mutual.sub.i,t] = A binary variable equal to 1 if firm i is organized as a mutual in year t and 0 otherwise; Kenny [Ratio.sub.i,t] = Firm i's net premiums written divided by policyholder surplus in year t; [ROI.sub.i,t] = Firm i's net investment income divided by total assets in year t; [Group.sub.i,t] = A binary variable equal to 1 if firm i is a member of a group and 0 otherwise; [Size.sub.i,t] = The natural log of firm i's total assets in year t; [ROA.sub.i,t] = Firm i's net income divided by total assets in year t; [Growth.sub.i,t] = The percent change in firm i's net premiums written from t - 1 to t; [Reinsurance.sub.i,t] = Firm i's reinsurance premiums ceded divided by the sum of direct premiums written and reinsurance assumed in year t; Product [Diverse.sub.i,t] = 1 minus a Herfindahl index based on firm i's net premiums written across 24 lines of business in year t; (17) [Earthquake.sub.i,t] = The percentage of firm i's net premiums written in earthquake insurance in year t; Geo [Herf.sub.i,t] = A geographic Herfindahl index based on direct premiums written in the 50 U.S. states and Washington, D.C. in year t; [Longtail.sub.i,t] = The percentage of firm i's net premiums written in long-tailed lines of business. (18)
Our primary variable of interest, [AQ.sub.j], is our empirical proxy for accounting quality. A negative estimated coefficient for [AQ.sub.j] is consistent with our hypothesis that ratings are an increasing function of accounting quality ([beta]1 < 0). We first test whether overall accruals quality is related to financial strength ratings by running separate models for [AQ.sub.3], [AQ.sub.4], [AQ.sub.5], and [AQ.SUB.10]. To test our second hypothesis (H2), we then run regressions with accruals quality decomposed into innate and discretionary components by including Innate[AQ.sub.3], Innate[AQ.sub.4], Innate[AQ.sub.5], Innate[AQ.sub.10], Disc[AQ.sub.3], Disc[AQ.sub.4], Disc[AQ.sub.5], and Disc[AQ.sub.10].
The remaining variables are consistent with prior studies to examine the determinants of insurer financial strength ratings (e.g., Pottier and Sommer, 1999; Doherty and Phillips, 2002). These variables are intended to measure factors that could increase or decrease insurer financial strength. These variables include measures of capital structure (Surplus-to-Assets and Kenny Ratio), profitability (ROA and ROI), the use of reinsurance (Reinsurance), and exposure to catastrophic risk (Earthquake).
Table 6 provides the results from our ordered probit model on the determinants of financial strength ratings. The four columns provide results when using the past 3,4, 5, and 10 years to construct standard deviations of loss reserve errors. Each coefficient is presented with cluster-robust standard errors in parentheses below. Positive coefficient estimates indicate a higher probability of achieving a higher financial strength rating, while negative values indicate a higher probability of having a lower financial strength rating.
Overall, the results in Table 6 are consistent with our hypotheses. The estimated coefficients on [AQ.sub.3], [AQ.sub.4], [AQ.sub.5], and [AQ.SUB.10] are all negative and statistically significant at the 1 percent level. This result provides evidence that higher reserve error volatility--lower accruals quality--is associated with lower financial strength ratings. Again, this is the result expected if ratings agencies incorporate the quality of a firm's accounting results into their ratings. (20)
Coefficient estimates on the control variables included in the regressions are generally consistent with expectations. Higher surplus, higher ROI, group membership, larger Size, higher ROA, higher Growth, and higher Product Diverse are associated with higher financial strength ratings. Higher net premiums written relative to surplus (Kenny Ratio) is associated with lower financial strength ratings, consistent with a growth penalty or excessive leverage. Results on the control variables also are generally consistent with results in prior studies (e.g., Pottier and Sommer, 1999; Doherty and Phillips, 2002; Park and Xie, 2014).
Table 7 provides results from our ordered probit model on the determinants of financial strength ratings, where we decompose accruals quality into innate (InnateAQ) and discretionary (DiscAQ) components. (21) To account for the inclusion of predicted values in these models, we estimate bootstrap standard errors with 1,000 replications (Pagan, 1984). These standard errors are included in parentheses beneath each coefficient estimate.
Again, consistent with our hypotheses, we find that lower accruals quality, whether it be innate or discretionary, is associated with lower financial strength ratings. (22) Notably, we find that the coefficient estimate and the statistical power of the coefficient estimate for InnateAQ are larger compared to the coefficient estimate and statistical power for the coefficient of DiscAQ. In addition, Wald tests of the hypothesis In-nateAQ= DiscAQ reject the null in all four models (p-values<0.0001). This provides evidence consistent with Best weighing the quality of innate accruals quality relatively more compared to that of discretionary accruals quality. We suggest that this is due to innate accruals being more permanent compared to transitory discretionary accruals. (23)
We find strong empirical evidence that overall accruals quality is significantly related to ratings. In particular, the results here indicate that higher accruals quality (lower reserve error volatility) is associated with higher financial strength ratings for property--casualty insurers. Moreover, decomposing accruals quality into innate and discretionary components indicates that both measures are significantly related to insurer ratings with the innate component appearing relatively more important.
This latter result suggests that more permanent, innate, accounting quality issues are penalized more severely than transitory, discretionary, accounting quality issues.
Thus,whilepriorresearch (Eckles, Halek,and Zhang, 2014) has linked accrualsquality with the cost of debt, we provide evidence of a link between accruals quality and insurer ratings. Arguably, our finding of the link between accruals quality and ratings is more fundamental since ratings largely determine the cost of debt for insurers (Epermanis and Harrington, 2006). Because ratings are an important component in effective market discipline (e.g., Epermanis and Harrington, 2006; Eling and Schmit, 2012), our findings provide insight into how ratings agencies such as Best appear to incorporate accruals quality into their rating process. We additionally contribute to the evolving literature examining the implications of the volatility of reserve errors, specifically (Eckles, Halek, and Zhang, 2014), and accruals, generally (Francis et al., 2005).Inessence,wefind empiricalevidencethat theratingsprocessimposesapenalty on insurers who transmit less credible earnings information (accruals quality) to the market. This research is important beyond the link between accruals quality, ratings, and insolvency since ratings are inherently important to firms. Future research may benefit from an examination of whether a similar relation holds for a broader sample of firms beyond the sample of property--casualty insurers here.
Anthony, J. H., and K. R. Petroni, 1997, Accounting Estimation Disclosures and Firm Valuation in the Property-Casualty Insurance Industry, Journal of Accounting, Auditing & Finance, 12: 257-281.
Beaver, W. H., M. F. McNichols, and K. K. Nelson, 2003, Management of the Loss Reserve Accrual and the Distribution of Earnings in the Property-Casualty Insurance Industry, Journal of Accounting and Economics, 35: 347-376.
Berry-Stolzle, T. R., A. P. Liebenberg, J. S. Ruhland, and D. W. Sommer, 2012, Determi-nants of Corporate Diversification: Evidence From the Property--Liability Insurance Industry, Journal of Risk and Insurance, 79: 381-413.
Carson, J., E. Eastman, and D. Eckles, 2016, Supplementary Material to 'Ratings: It's Accrual World', Journal of Risk and Insurance, http://dx.doi.org/10.1111/jori.12179.
Cummins, J. D., M. F. Grace, and R. D. Phillips, 1999, Regulatory Solvency Prediction in Property-Liability Insurance: Risk-Based Capital, Audit Ratios, and Cash Flow Simulation, Journal of Risk and Insurance, 66: 417-458.
Dechow, P., W. Ge, and C. Schrand, 2010, Understanding Earnings Quality: A Review of the Proxies, Their Determinants and Their Consequences, Journal of Accounting and Economics, 50: 344-401.
Doherty, N. A., A. V. Kartasheva, and R. D. Phillips, 2012, Information Effect of Entry Into Credit Ratings Market: The Case of Insurers' Ratings, Journal of Financial Economics, 106: 723-750.
Doherty,N.A., andR.D.Phillips, 2002, KeepingupWith the Joneses:Changing Rating Standards and the Buildup of Capital by U.S. Property-Liability Insurers, Journal of Financial Services Research, 21: 55-78.
Eastman, E. M., D. L. Eckles, M. Halek, and L. S. Powell, 2014, Earnings Management, Executive Compensation, and Ownership Structure, Working Paper, University of Georgia.
Eckles, D. L., and M. Halek, 2010, Insurer Reserve Error and Executive Compensation, Journal of Risk and Insurance, 77: 329-346.
Eckles,D.L.,M.Halek,E.He,D.W.Sommer, andR.Zhang, 2011, Earnings Smoothing, Executive Compensation, and Corporate Governance: Evidence From the Property-Liability Insurance Industry, Journal of Risk and Insurance, 78: 761-790.
Eckles, D. L., M. Halek, and R. Zhang, 2014, Information Risk and the Cost of Capital, Journal of Risk and Insurance, 81: 861-882.
Eling, M., and J. T.Schmit, 2012, Is There Market Discipline in the European Insurance Industry? An Analysis of the German Insurance Market, Geneva Risk and Insurance Review, 37: 180-207.
Epermanis, K., and S. E. Harrington, 2006, Market Discipline in Property/Casualty Insurance: Evidence From Premium Growth Surrounding Changes in Financial Strength Ratings, Journal of Money, Credit and Banking, 38: 1515-1544.
Francis, J., R.LaFond,P.Olsson, andK.Schipper, 2005, The Market Pricing of Accruals Quality, Journal of Accounting and Economics, 39: 295-327.
Gaver, J. J., and J. S. Paterson, 2001, The Association Between External Monitoring and Earnings Management in the Property-Casualty Insurance Industry, Journal of Accounting Research, 39: 269-282.
Gaver, J. J., and J. S. Paterson, 2004, Do Insurers Manipulate Loss Reserves to Mask Solvency Problems? Journal of Accounting and Economics, 37: 393-416.
Gaver, J. J., and J. S. Paterson, 2007, The Influence of Large Clients on Office-Level Auditor Oversight: Evidence From the Property-Casualty Insurance Industry, Journal of Accounting and Economics, 43: 299-320.
Grace, E. V., 1990, Property-Liability Insurer Reserve Errors: A Theoretical and Empirical Analysis, Journal of Risk and Insurance, 57: 28-46.
Grace, M. F., and J. T. Leverty, 2010, Political Cost Incentives for Managing the Property-Liability Insurer Loss Reserve, Journal of Accounting Research, 48: 21-49.
Grace, M. F., and J. T. Leverty, 2012, Property-Liability Insurer Reserve Error: Motive, Manipulation, or Mistake, Journal of Risk and Insurance, 79: 351-380.
Grace, M. F., and J. T. Leverty, 2013, External Monitor Quality and Managerial Discretion, Working Paper, University of Iowa.
Greene,W.,2004, The Behaviourofthe MaximumLikelihoodEstimatorofLimitedDe-pendent Variable Models in the Presence of Fixed Effects, The Econometrics Journal, 7: 98-119.
Guay, W. R., S. Kothari, and R. L. Watts, 1996, A Market-Based Evaluation of Discretionary Accrual Models, Journal of Accounting Research, 34: 83-105.
Halek, M., and D. L. Eckles, 2010, EffectsofAnalysts' Ratings onInsurer Stock Returns: Evidence of Asymmetric Responses, Journal of Risk and Insurance, 77: 801-827.
McNichols, M. F., 2000, Research Design Issues in Earnings Management Studies, Journal of Accounting and Public Policy, 19: 313-345.
Pagan, A., 1984, Econometric Issues in the Analysis of Regressions With Generated Regressors, International Economic Review, 25: 221-247.
Park, S. C., and X. Xie, 2014, Reinsurance and Systemic Risk: The Impact of Reinsurer Downgrading on Property--Casualty Insurers, Journal of Risk and Insurance, 81: 587-622.
Petroni,K. R., 1992, Optimistic Reportinginthe Property-Casualty Insurance Industry, Journal of Accounting and Economics, 15: 485-508.
Pottier, S. W., and D. W. Sommer, 1999, Property-Liability Insurer Financial Strength Ratings: Differences Across Rating Agencies, Journal of Risk and Insurance, 66: 621-642.
Pottier, S. W., and D. W. Sommer, 2002, The Effectiveness of Public and Private Sector Summary Risk Measures in Predicting Insurer Insolvencies, Journal of Financial Services Research, 21: 101-116.
Wade, C., A. Liebenberg, and B. M. Blau, 2016, Information and Insurer Financial Strength Ratings: Do Short Sellers Anticipate Ratings Changes? Journal of Risk and Insurance, 83: 475-500.
Weiss, M., 1985, A Multivariate Analysis of Loss Reserving Estimates in Property-Liability Insurers, Journal of Risk and Insurance, 52: 199-221.
Additional Supporting Information may be found in the online version of this article:
Appendix Table 1: Ordered Probit Regression Results--Results by Year ([AQ.sub.3])
Appendix Table 2: Ordered Probit Regression Results--Results by Year ([AQ.sub.4])
Appendix Table 3: Ordered Probit Regression Results--Results by Year ([AQ.sub.5)
Appendix Table 4: Ordered Probit Regression Results--Results by Year ([AQ.sub.10])
Appendix Table 5: Ordered Probit Regression Results--Results by Year (Innate[AQ.sub.3] & Disc[AQ.sub.3])
Appendix Table 6: Ordered Probit Regression Results--Results by Year (Innate[AQ.sub.4] & Disc[AQ.sub.4])
Appendix Table 7: Ordered Probit Regression Results--Results by Year (Innate[AQ.sub.5] & Disc[AQ.sub.5])
Appendix Table 8: Ordered Probit Regression Results--Results by Year (Innate[AQ.sub.10] & Disc[AQ.sub.10])
James M. Carson is the Daniel P. Amos Distinguished Professor of Insurance, Department of Insurance, Legal Studies, and Real Estate, Terry College of Business, University of Georgia, 206 Brooks Hall, Athens, GA 30602. Carson can be contacted via e-mail: email@example.com. Evan M. Eastman is a Doctoral Candidate, Department of Insurance, Legal Studies, and Real Estate, Terry College of Business, University of Georgia, 206 Brooks Hall, Athens, GA 30602. Eastman can be contacted via e-mail: firstname.lastname@example.org. David L. Eckles is an Associate Professor of Risk Management and Insurance, Department of Insurance, Legal Studies, and Real Estate, Terry College of Business, University of Georgia, 206 Brooks Hall, Athens, GA 30602. Eckles can be contacted via e-mail: email@example.com. The authors would like to thank the editor, Keith Crocker, two anonymous referees, and Lorilee Medders for helpful comments. The authors are grateful to participants at the 2015 FSU/UGA Research Symposium, the 2015 World Risk and Insurance Economics Congress, and the 2015 Southern Risk and Insurance Association Annual Meeting for their feedback. The authors would also like to thank A.M. Best and Robert E. Hoyt for assistance in obtaining ratings data.
(1) Pottier and Sommer (2002), for example, find evidence that private sector measures of insolvency, including Best's financial strength ratings, are better predictors of insolvency compared to measures used by the public sector. The sensitivity of public sector measures to accounting manipulation could be one potential explanation for this finding.
(2) There are other measures of reserve error that also have been used in the literature. Petroni (1992), Eckles and Halek (2010), and Eastman et al. (2014) use total incurred losses after 5 years minus the initial estimate. This produces the negative of the measure we use here. Grace and Leverty (2012) use the initial estimate minus losses paid after 5 years. Grace and Leverty (2013) use a measure based on stochastic loss reserving models as used in the actuarial science literature, which they call the full information reserve error.
(3) Prior studies report that results are generally robust to different scaling variables. Beaver, McNichols, and Nelson (2003), Gaver and Paterson (2004), and Eckles and Halek (2010) report that their results are robust to scaling choice. At the suggestion of a referee, we also scaled by net premiums written and net premiums earned and obtain results in line with those presented within.
(4) Eckles, Halek, and Zhang (2014) use the term "information risk."
(5) We also conduct our analysis on a subsample of our data for which we have cost of debt data for insurers. As in Eckles, Halek, and Zhang (2014), we use the inverse loss ratio as a measure of an insurer's cost of debt. Our results from this analysis are consistent with those presented here in this article. However, the relation between the cost of debt and accruals quality likely flows from the relation between insurer ratings and accruals quality,asopposedtothe opposite direction.
(6) Best also began to offer "issuer credit ratings" (ICRs) in 2003. We perform similar analysis to that reported in this article on insurers who received an ICR. We find mixed evidence that accruals quality impacts an insurer's ICR. However, we suspect that this is due to a small sample size in this unreported analysis. We also find that ICRs are highly correlated with financial strength ratings.
(7) While we note that the accounting procedures may well be correct and proper, we refer to "low-quality" to reference the amount of information available from the accounting results.
(8) Guay, Kothari, and Watts (1996) note the performance component should improve quality, while the opportunism and noise component should reduce quality. (9) For example, the 2006 reserve error is calculated using data from a firm's 2011 statutory filing.
(10) The [R.sub.2]s from each of these models are 0.0074, 0.0105, 0.0125, and 0.0192 for [AQ.sub.3], [AQ.sub.4], [AQ.sub.5], and [AQ.sub.10], respectively. While these may seem relatively low, we note that this model is simply attempting to decompose accruals quality into its innate and discretionary components. These low [R.sub.2]s likely indicate an omitted variable bias, which is common in the accounting quality literature. The overall impact of this is that it leads to a downward bias (toward zero) in our discretionary accrual measure (Francis et al., 2005). Finally, we also note that our overall accruals quality measures (AQ) are not affected by the [R.sub.2] values.
(11) In particular, this does not change for the interpretation of DiscAQ, which can take negative values due to the nature of the decomposition using ordinary least squares. Negative values of DiscAQ indicate that the discretionary component of accruals quality improves overall accruals quality. For the purpose of our empirical tests, however, lower (higher) values of DiscAQ indicate higher (lower) accruals quality, as with the other two accruals quality variables (Francis et al., 2005).
(12) Our rating classification is consistent with Doherty and Phillips (2002), and our results are robust to alternative ratings classifications.
(13) Based on concern from a referee regarding potential skewness, we also estimate our models separately for firms in the lower 90th percentile and upper 10th percentile of the accruals quality distributions. The results presented here are consistent for those in the lower 90th percentile, suggesting that it is not the potentially skewed upper tail dominating the results.
(14) While we do not include year or firm fixed effects in our model due to econometric issues related to estimating ordered probit models with fixed effects (Greene, 2004), our results are robust to their inclusion. Our main results are also robust to the inclusion of year fixed effects only, as well as year and firm fixed effects for ordinary least squares models. Finally, our results are robust to estimating a random effects ordered logit model.
(15) We additionally perform annual cross-sectional regressions to examine whether Best's consideration of accruals quality has changed over time. Based on these (unreported) tests, we find evidence that accruals quality has been a consideration in financial strength ratings since at least 1993 with no discernible change or trend since 1993. These results can be found in an online appendix (Carson, Eastman and Eckles 2016).
(16[) AQ.sub.10] is the standard deviation of the 5-year loss reserve error for at least 5 and as many as 10 years.
(17) Using net premiums written data from the Underwriting and Investment Exhibit (Part 1B-Premiums Written) in the annual statutory filings, we make the following adjustments as de-scribedin Berry-Stlzle et al. (2012). Fire and Allied Lines isdefined asthe sum of "Fire" and "Allied Lines." Accident and Healthisdefinedasthe sumof"Group Accident and Health," "Credit Accident and Health," and "Other Accident and Health." Medical Malpractice is defined as the sum of "Medical Malpractice--Occurrence" and "Medical Malpractice--Claims Made." Products Liability is defined as the sum of "Products Liability--Occurrence" and "Products Liability--Claims Made." Auto is defined as the sum of "Private Passenger Auto Liability," "Commercial Auto Liability," and "Auto Physical Damage." Reinsurance is defined as the sum of "Nonproportional Assumed Property," "Nonproportional Assumed Liability," and "Non-proportional Assumed Financial Lines." After these combinations, we are left with 24 lines of business from which we construct the Herfindahl index: Accident and Health, Aircraft, Auto, Boiler and Machinery, Burglary and Theft, Commercial Multi Peril, Credit, Earthquake, Far-mowners, Financial Guaranty, Fidelity, Fire and Allied lines, Homeowners, Inland Marine, International, Medical Malpractice, Mortgage Guaranty, Ocean Marine, Other, Other Liability, Products Liability, Reinsurance, Surety, and Workers' Compensation.
(18) We define the following lines as long-tailed lines of business: Farmowners, Homeowners, Commercial MultiPeril, Medical Malpractice,Workers' Compensation,Products Liability,Auto Liability, and Other Liability.
(19) While the magnitude and significance of the coefficient estimates from the ordered probit model indicate that lower accruals quality (i.e., higher values of AQ) is associated with lower ratings, there is no straightforward interpretation of the marginal effect. In untabulated ordinary least squares models, the estimated coefficient of [AQ.sub.3], [AQ.sub.4], [AQ.sub.5], and [AQ.sub.10] is approximately -3.39, -3.92, -4.18, and -4.53, respectively. They are all significant at the 1 percent level. This indicates that for a one-unit increase in any of the AQ measures, all else equal, there is approximately a four-unit decrease in ratings. While this may seem relatively large given our ratings categorization, note that the magnitudes of our AQ variables are quite small (see Table 3).
(20) In untabulated results, we included two measures of earnings volatility in the spirit of Doherty and Phillips (2002) to alleviate concerns that our measure was measuring overall earnings volatility. In these specifications, our measures of accruals quality were always significant. The two overall earnings volatility measures, however, were not always significant.
(21) Recall that we require at least 5 and up to 10 lags to calculate the decomposition into innate and discretionary accruals quality. This restriction is why all three models including the decomposed AQ variables have the same number of observations.
(22) As we did for the results reported in Table 6, we perform unreported tests using ordinary least squares to gain some insight into the marginal effect of accruals quality on ratings. Estimated coefficients of Innate[AQ.SUB.3], Innate[AQ.SUB.4], Innate[AQ.SUB.5], and Innate[AQ.SUB.10] are approximately -104.47, -84.33, -71.45, and -55.11, respectively. These coefficient estimates are significant at the 1 percent level. Estimated coefficients of Disc[AQ.sub.3], Disc[AQ.sub.4], Disc[AQ.sub.5], and Disc[AQ.sub.10] are approximately -2.84, -3.32, -3.69, and -3.82, respectively. These coefficient estimates are significant at the 1 percent level. These estimates are interpretable as the effect of a one-unit increase in either InnateAQ or DiscAQ on Rating.
(23) At the suggestion of a referee, we also consider the degree to which poorer accruals quality is penalized differently for those firms that underreserve relative to thosefirms that overreserve. When estimating our models separately for overreserving and underreserving firms, our results remain consistent with a ratings penalty apparent for both types of firms.
DOI: 10.1111/jori 12179
|Printer friendly Cite/link Email Feedback|
|Author:||Carson, James M.; Eastman, Evan M.; Eckles, David L.|
|Publication:||Journal of Risk and Insurance|
|Date:||Sep 1, 2018|
|Previous Article:||RISK MISPERCEPTIONS AND SELECTION IN INSURANCE MARKETS: AN APPLICATION TO DEMAND FOR CANCER INSURANCE.|
|Next Article:||RATING CHANGES AND COMPETING INFORMATION: EVIDENCE ON PUBLICLY TRADED INSURANCE FIRMS.|