accident year, we heavily rely on the loss assumptions that are used in pricing our business. Historically, and at present, we utilize loss ratios that are approximately 8 to 10 percentage points above the ratios we believe we have achieved in our pricing for the current accident year. We believe this considers inherent risks associated with our rate development process and the historic volatility of professional liability losses (the industry has experienced accident year loss ratios as high as 163% and as low as 57% over the past 30 years) and produces a reasonable best estimate of the reserves required to cover actual ultimate unpaid losses. In the current environment this equates to an initial loss ratio of approximately 85% as compared to an average loss ratio of approximately 74% targeted in our pricing. Although we remain uncertain regarding the ultimate severity trend, as discussed in following paragraphs, the consideration we have given to observed lower loss costs has resulted in rate reductions in recent years. On average, excluding our podiatry business acquired in 2009, we have gradually reduced the premium rates we charge by 16% from 2006 to 2012.
Reestimation of Prior Years' Loss Reserves Process
The foundation of our reserve re-estimation process is an actuarial analysis that is performed by both our internal and consulting actuaries. The very detailed analysis projects ultimate losses on a line of business, geographic, coverage layer and accident year basis. The procedure is intended to balance the use of the most representative data for each partition, capturing its unique patterns of development and trends. In all there are approximately 140 different partitions of our business for purposes of this analysis. We believe that the use of consulting actuaries provides an independent view of our loss data as well as a broader perspective on industry loss trends.
The analysis performed by the consulting actuaries analyzes each partition of our business in a variety of ways and uses multiple actuarial methodologies in performing these analyses, including:
• Bornhuetter-Ferguson (Paid and Reported) Method

• Paid Development Method

• Reported Development Method

• Average Paid Value Method

• Average Reported Value Method

• Backward Recursive Development Method

A brief description of each method follows.
Bornhuetter-Ferguson Method. We use both the Paid and the Reported Bornhuetter-Ferguson methods. The Paid method assigns partial weight to initial expected losses for each accident year (initial expected losses being the first established case and IBNR reserves for a specific accident year) and partial weight to paid to-date losses. The Reported method assigns partial weight to the initial expected losses and partial weight to current expected losses. The weights assigned to the initial expected losses decrease as the accident year matures.
Paid Development and Reported Development Method. These methods use historical, cumulative losses (paid losses for the Paid Development Method, reported losses for the Reported Development Method) by accident year and develop those actual losses to estimated ultimate losses based upon the assumption that each accident year will develop to estimated ultimate cost in a manner that is analogous to prior years, adjusted as deemed appropriate for the expected effects of known changes in the claim payment environment (and case reserving environment for the Reported Development Method), and, to the extent necessary, supplemented by analyses of the development of broader industry data.
Average Paid Value and Average Reported Value Methods. In these methods, average claim cost data (paid claim cost for the Average Paid Value Method and reported claim cost for the Reported Value Method) is developed to an ultimate average cost level by report year based on historical data. Claim counts are similarly developed to an ultimate count level. The average claim cost (after rounding and adjustment, if necessary, to accommodate report year data that is not considered to be predictive) is then multiplied by the ultimate claim counts by report year to derive ultimate loss and ALAE.
Backward Recursive Development Method. This method is an extrapolation of the movements in case reserve adequacy in order to estimate unpaid loss costs. Historical data showing incremental changes to case reserves over progressive time periods is used to derive factors that represent the ratio of case reserve values at successive maturities. Historical claims payment data showing the additional payments in progressive time periods is used to derive factors that represent the portion of a case reserve paid in the following period. Starting from the most mature period, after which all of the case reserve is paid and the case reserve is exhausted, the next prior ultimate development factor for the prior case reserve can be calculated as the case factor times the established ultimate development factor plus the paid factor. For each successive prior maturity, the ultimate

development factor is calculated similarly. The result of multiplying the ultimate development factor times the case reserve is the total indicated unpaid amount.
Generally, methods such as the Bornhuetter-Ferguson method are used on more recent accident years where we have less data on which to base our analysis. As time progresses and we have an increased amount of data for a given accident year, we begin to give more confidence to the development and average methods, as these methods typically rely more heavily on our own historical data. These methods emphasize different aspects of loss reserve estimation and provide a variety of perspectives for our decisions.
Certain of the methodologies utilized to estimate the ultimate losses for each partition of our reserves consider the actual amounts paid. Paid data is particularly influential when a large portion of known claims have been closed, as is the case for older accident years. In selecting a point estimate for each partition, management considers the extent to which trends are emerging consistently for all partitions and known industry trends. Thus, actual, rather than estimated severity trends are given consideration. When actual severity trends are lower than those estimated at the time that reserves were previously established, the recognition of favorable development is indicated. This is particularly true for older accident years where our actuarial methodologies give more weight to actual loss costs (severity).
The various actuarial methods discussed above are applied in a consistent manner from period to period. In addition, we perform statistical reviews of claims data such as claim counts, average settlement costs and severity trends when establishing our reserves.
We utilize the selected point estimates of ultimate losses to develop estimates of ultimate losses recoverable from reinsurers, based on the terms and conditions of our reinsurance agreements. An overall estimate of the amount receivable from reinsurers is determined by combining the individual estimates. Our net reserve estimate is the gross reserve point estimate less the estimated reinsurance recovery.
Use of Judgment
Even though the actuarial process is highly technical, it is also highly judgmental, both as to the selection of the data used in the various actuarial methodologies (e.g., initial expected loss ratios and loss development factors) and in the interpretation of the output of the various methods used. Each actuarial method generally returns a different value and for the more recent accident years the variations among the various methodologies can be significant. For each partition of our reserves, the results of the various methods, along with the supplementary statistical data regarding such factors as closed with and without indemnity ratios, claim severity trends, the expected duration of such trends, changes in the legal and legislative environment and the current economic environment, are used to develop a point estimate based upon management's judgment and past experience. The process of selecting the point estimate is based upon the judgment of management taking into consideration the actuarial methods and other environmental factors discussed previously. For each partition of our reserves, we select a point estimate with due regard for the age, characteristics and volatility of the subset of the business, the volume of data available for review and past experience with respect to the accuracy of estimates. The series of selected point estimates is then combined to produce an overall point estimate for ultimate losses. Given the historical volatility of the MPL line of business we are cautious in giving full credibility to emerging trends that, when more fully mature, may lead to the recognition of either favorable or adverse development of our losses. There may be trends, both positive and negative, reflected in the numerical data both within our own information and in the broader MPL marketplace that mitigate or reverse as time progresses and additional data becomes available.
Over the past several years the most influential factor affecting our analysis of reserves has been the changes, or lack thereof, in the severity of claims. Severity is defined as the average cost of resolving claims. The severity trend assumption is a key assumption for both pricing models and actuarial estimation of reserves. The severity trend is an explicit component of our pricing models, whereas in our reserving process the severity trend's impact is implicit. Our estimate of this trend and our expectations about changes in this trend impact a variety of factors, from the selection of expected loss ratios to the ultimate point estimates established by management.
Because of the implicit and wide-ranging nature of severity trend assumptions on the loss reserving process it is not practical to specifically isolate the impact of changing severity trends. However, because severity is an explicit component of our pricing process we can better isolate the impact that changing severity can have on our loss costs and loss ratios as regards our pricing models. Our current pricing model assumes a severity trend of 2% to 3% in most states and lines of business. If the severity trend were to be higher by 1 percentage point, the impact would be an increase in our expected loss ratio of 3.2 percentage points. An increase in the severity trend of 3 percentage points would result in a 10.1 percentage point increase in our expected loss ratio. Due to the long tailed nature of MPL claims and the previously discussed historical volatility of loss costs, selection of a severity trend assumption is a subjective process that is inherently likely to prove inaccurate over time. Given this long-tail and the previously discussed historical volatility of loss costs, we are generally cautious in making changes

to the severity assumptions within our pricing model. Also of note is that all open claims and accident years are generally impacted by a change in the severity trend, which compounds the effect of such a change.
From 2004 to 2009 both our internal and consulting actuaries have observed an unprecedented reduction in the frequency of claims (or number of claims per exposure unit) that cannot be attributed to any single factor, which has complicated the selection of an appropriate severity trend for our pricing models. It has also made it more challenging to factor severity into the various actuarial methodologies discussed above. We believe that much of the reduction in claim frequency is the result of a decline in the filing of frivolous lawsuits that have historically been dismissed or otherwise result in no payment of indemnity on the part of our insureds. With fewer frivolous claims being filed we expect that the claims that are filed have the potential for greater average losses, or greater severity. As a result, we cannot be certain as to the impact this decline will ultimately have on the average cost of claims. Based on a weighted average of payments, only 85% are resolved after eight years for a given accident year. Due to this long tail, it can take several years before we are able to determine what impact, if any, has resulted from the decline in frequency and whether there is a related increase in severity. Loss Development
We recognized net favorable development related to prior accident years of $272.0 million for the year ended December 31, 2012, $325.9 million for the year ended December 31, 2011, and $234.0 million for the year ended December 31, 2010. In support of our concern that the decline in frequency will result in a higher severity trend, we have seen our closed-with-indemnity-payment ratio (i.e., the number of claims closed with an indemnity or loss payment as compared to the total number of closed claims) increase from 7% in 2005 to 14% in 2012. While this trend has been in keeping with our expectations, the anticipated increase in severity incorporated into our loss assumptions has not occurred. Rather, we have experienced lower than expected severity which has been the primary driver of the favorable development recognized in recent years. The following tables present additional information about our loss development:
Estimated Ultimate Losses, Net of Reinsurance ("established ultimates"):

        (In thousands)                    2012*                      2011                  2010
2012                           $                  451,951                   N/A                   N/A
2011                                              483,264     $         488,152                   N/A
2010                                              476,449               490,061     $         493,354
2009                                              451,298               475,676               497,766
2008                                              403,641               459,299               510,861
2007                                              385,530               436,577               498,240
2006                                              340,983               379,692               438,487
2005                                              362,319               387,280               423,380
2004                                              348,738               367,656               406,944
2003                                              492,576               507,673               529,722
Prior to 2003                                   5,065,129             5,089,899             5,120,925

Reserve Development, (favorable) unfavorable:

(In thousands)    2012         2011         2010
2011           $ (4,888 )        N/A          N/A
2010            (13,612 )   $ (3,293 )        N/A
2009            (24,378 )    (22,090 )   $  2,004
2008            (55,658 )    (51,562 )     (5,035 )
2007            (51,047 )    (61,663 )    (36,858 )
2006            (38,709 )    (58,795 )    (67,418 )
2005            (24,961 )    (36,100 )    (42,259 )
2004            (18,918 )    (39,288 )    (38,182 )
2003            (15,097 )    (22,049 )    (17,362 )
Prior to 2003   (24,770 )    (31,025 )    (28,880 )

An extended period of time is required to get a clear estimate of the loss cost for a given accident year. As an example, looking at the 2008 accident year, we had resolved 71.2% of the known claims by the end of 2010, 83.1% of the known claims by the end of 2011, and 90.3% of the known claims by the end of 2012. These statistics are based on the number of reported claims; since many non-meritorious claims are resolved early, percentages of ultimate loss payments known at the same points in time are considerably lower. A similar pattern can be seen in each open accident year as demonstrated in the table below. Historically we have resolved more than 85% of our physician and hospital professional liability claims with no indemnity payment and generally these claims are the first to be resolved. As an accident year matures, the number of claims resolved with indemnity payments progressively increases. In a similar fashion, we typically expend more in loss adjustment expenses (legal fees) as claims mature.
The following table represents the percentage of known MPL claims closed:

Accident Year  2012     2011     2010
2012          13.5 %    N/A      N/A
2011          45.0 %   13.2 %    N/A
2010          68.8 %   45.3 %   16.4 %
2009          80.6 %   67.7 %   46.3 %
2008          90.3 %   83.1 %   71.2 %
2007          93.9 %   89.5 %   82.5 %
2006          97.1 %   94.7 %   90.0 %
2005          98.4 %   96.9 %   94.4 %
2004          99.1 %   98.3 %   96.8 %
2003          98.7 %   98.0 %   96.8 %

Based upon the additional claims closed during 2012, 2011 and 2010, as shown above, and the continuation of better than expected severity trends, management reduced its expected ultimate losses in each of these years resulting in the recognition of corresponding amounts of favorable development in the income statements of those periods. At December 31, 2010 management reserve estimates for the three most recent prior accident years (which have closed claim percentages below 85%) were more heavily influenced by the initial reserves estimates set for these years, while estimates for older accident years with higher percentages of closed claims were more heavily influenced by the more moderate severity trend observed during the period, particularly with regard to claims closed during the period. At December 31, 2012 and 2011 the continued lack of increase in the severity trend, in spite of the increase in the ratio of claims closed with indemnity, influenced management's judgment both in regard to the older accident years where this trend has more fully played out and in regard to the more recent accident years. The increased favorable development recognized in 2011 as compared to 2010 is largely attributable to the initial effect of giving greater consideration to the moderated severity trend. Management continued to reflect approximately the same level of consideration of the moderated trend in its estimate of the reserve required at December 31, 2012; however, the effect was not as pronounced in 2012 as that initially recorded in 2011.
This can be seen in looking at both the absolute amount of favorable reserve development recognized for the less developed accident years as well as the size of such development when compared to established ultimates for those same accident years at the end of the preceding calendar year. The following table provides this information for years ended December 31, 2012, 2011 and 2010 with respect to the three then most recent prior accident years:

         ($ in millions)               2012             2011             2010
Prior accident years                2009-2011        2008-2010        2007-2009
Net favorable development
recognized for the specified
years                                 $42.9            $76.9            $39.9
Development as a % of established
ultimates, prior calendar year
end                                    2.9%             5.1%             2.6%

Variability of Loss Reserves
As previously noted, the number of data points and variables considered and the subjective process followed in establishing our loss reserve makes it impractical to isolate individual variables and demonstrate their impact on our estimate of loss reserves. However, to provide a better understanding of the potential variability in our reserves, we have modeled implied reserve ranges around our single point net reserve estimates for our professional liability business assuming different confidence levels. The ranges have been developed by aggregating the expected volatility of losses across partitions of our business to obtain a consolidated distribution of potential reserve outcomes. The aggregation of this data takes into consideration the correlation among our geographic and specialty mix of business. The result of the correlation approach to aggregation is that the ranges are narrower than the sum of the ranges determined for each partition.
We have used this modeled statistical distribution to calculate an 80% and 60% confidence interval for the potential outcome of our net reserve for losses. The high and low end points of the distributions are as follows:

                     Low End Point    Carried Net Reserve   High End Point
80% Confidence Level $1.418 billion     $1.863 billion      $2.374 billion
60% Confidence Level $1.535 billion     $1.863 billion      $2.161 billion

Any change in our estimate of net ultimate losses for prior years is reflected in net income in the period in which such changes are made. Over the past several years such changes have been to reduce our estimate of net ultimate losses, resulting in a reduction of reported losses for the period and a corresponding increase in income.
Due to the size of our reserve for losses and the large number of claims outstanding at any point in time, even a small percentage adjustment to our total reserve estimate could have a material effect on our results of operations for the period in which the adjustment is made. Reinsurance
We use insurance and reinsurance (collectively, "reinsurance") to provide capacity to write larger limits of liability, to provide protection against losses in excess of policy limits and to stabilize underwriting results in years in which higher losses occur. The purchase of reinsurance does not relieve us from the ultimate risk on our policies, but it does provide reimbursement for certain losses we pay.
We make a determination of the amount of insurance risk we choose to retain . . .