CALCULATING ‘REASONABLE’ SUCCESS FEES FOR EMPLOYERS’ LIABILITY ACCIDENT CLAIMS A Report to the CJC, May 2004 Paul Fenn, University of Nottingham Business School & Neil Rickman, Department of Economics, University of Surrey
Objective The objective of this study is to assist in the determination of „reasonable‟ success fees for Employers‟ Liability (EL) accident claims1. We suggest that one measure of reasonableness is to calculate success fees such that a CFA case yields the same revenue to the solicitor as an hourly fee counterpart (on average) – that is, a success fee which would make the choice between CFA cases and hourly fee cases revenueneutral over a sufficiently large number of EL accident cases2. In turn, this requires detailed knowledge of the stages through which a case may proceed, the probabilities of transition between these stages and the costs associated with the transition. A decision tree was constructed to facilitate the calculation of revenue-neutral success fees, and the data we required to populate the tree with costs and probabilities were collected through spreadsheet templates. These templates were the basis of our negotiations with representatives of defendants and claimants. We met or corresponded over the summer with the ABI, APIL and Thompsons. The discussions helped clarify what was the essential core of data required to deliver the objectives set out above, and various datasets were delivered to us through August and September 2003. Defendant data On the defendant side, we received data from the following insurance companies/panel solicitors: NU Insurance Allianz Cornhill Zurich Commercial RSA ITIC AXA Churchill Groupama Beechcroft Wansborough
1
This is part of a wider study including EL disease claims as well as PL claims; only results for EL accident claims are reported here. 2 It should be emphasised that this is only one criterion that could be used to determine the level of a success fee. We propose that these calculations should be used as initial guidance for subsequent negotiations.
1
We also obtained data from the Compensation Recovery Unit (CRU). As anticipated, none of these by themselves have produced a complete set of data sufficient to populate the decision tree with parameters. In particular, it was clearly impossible for defendants to know what costs were incurred prior to the claim becoming known to them. Moreover, it became equally clear that the following limitations applied: 1. claims that were dropped or repudiated before any payment was made by insurers were not always recorded on insurer databases in a way that allowed an estimate of the success rates by type of claim; claims with costs were typically sent to cost negotiators, but other claims were not easy to recover. 2. not all insurers keep a record of the means of funding used for each claim 3. not all insurers keep a record of whether the case was litigated (proceedings issued) or not 4. not all insurers record whether a claim has gone to trial For these and other reasons, it was necessary to draw data from several sources in order to produce a complete set of parameters for the decision tree; there is no single pooled dataset equivalent to that used in our earlier CJC study. Nevertheless, we have produced estimates for all parameters relevant to the process after a claim has been made (some with more statistical confidence than others). Claimant data On the claimant side, we received data from Several (anonymous) members of APIL, including one large BTE insurer Thompsons As with the defendant data, it is clear that we have sufficient material to populate parts of the decision tree, but that we need to rely on a variety of sources for completing the process. Issues arising with the claimant data are as follows: 1. A significant portion of the data only records the inter partes costs that are agreed/paid. 2. Where data are restricted to recovered inter partes costs, there are no cost data for cases which have been dropped or successfully defended. The recording of these cases is subject to some uncertainty because of the absence of revenue consequences. 3. Not all of the datasets provided to us record the fee scheme, the damages or the stage reached in litigation. 4. APIL have commented that an important element of risk in CFA cases has been introduced by Part 36 offers. We have no means of estimating the impact of this factor
2
Evidence Our first task was to collate together the various pieces of information provided to us by both claimants and defendants in a way that allowed us to determine some baseline values for the decision tree parameters. We structured this process into three main parts: information on transition probabilities (the likelihood that a claim will move through successive stages of litigation); information on costs (the mean cost of cases settled or closed during each stage of litigation); and failure rates (the likelihood that claims are withdrawn or repudiated during each stage of litigation). We have summarised our analysis correspondingly in three separate tables below. The columns in these tables represent the various sources of data we used. Not all of the datasets provided to us were usable; we have only included those which satisfied us with respect to their representativeness, and the quality of the data. Also, as pointed out above, within each dataset there are some variables that could be estimated, and some that could not. Some of the sources provided large numbers of cases, and we have noted the relevant sample sizes in the tables. The penultimate column in each table presents our choice of baseline assumption for each parameter – in most cases this is the approximate midpoint of the range of estimates provided to us, but we have taken into account other factors, particularly the statistical confidence in the estimates based on relative sample sizes. Nevertheless, it is important to stress that the judgements we have made are to some extent subjective, and should be treated as starting points for further discussion and mediation. One of our sources, reported in column 2 of the tables below, can be identified as the Compensation Recovery Unit (CRU). This has obvious advantages as a source of information about claim outcomes, in that all personal injury claims for compensation are notionally reported to the CRU and their outcomes recorded. However, some reservations have been expressed about the accuracy of the CRU data on “failure rates”, and for this reason a review of a large database of claim level data from CRU was undertaken, and is the subject of a separate report previously circulated. The queries raised in relation to the CRU data have on the whole been addressed, and we are confident that the CRU estimates of the failure rate of EL accident claims can be relied upon to be representative of UK settlements in this category of claim. However, CRU do not record whether a claim was litigated or not, and for this reason we cannot rely solely on this source for failure rates in our model. When deciding what values to use in the model, we have therefore attempted to combine estimates from other sources in relation to failure rates by stage of litigation, in such a way that is nevertheless consistent with the overall failure rates reported by CRU.
3
Table 1: Evidence on transition probabilities; EL claims
Source Claim to issue Mean N Issue to trial Mean N
1
2
3
4 0.131 9208
5 0.115 165
6
7
8 0.148 19226 0.045 2847
9
Base case notes 0.150 source 8 has high numbers 0.045 only one source
4
Table 2: Evidence on the costs of settling EL claims (base costs, net of VAT)
Pre-issue Post-issue Trial
Source Mean N Mean N Mean N
1 2140.79 1992 5239.28 1818
2
3
5413.82 2410
4 1420.46 8001 3851.56 1207
5 1516.89 94 4059.18 16
6 2464.60 12 5195.24 8
7
8 1561.41 6233 4157.81 2466 6020.31 78
9
Base case notes 1500.00 midway between source 4 and 8
3818.89 4800.00 midway between source 3 and 8 77 6000.00 only one source
5
Table 3: Evidence on failure rates, EL accident claims
Source Pre-issue Mean N Post-issue Mean N Trial Mean N
1
2 0.259* 111731 0.259* 111731
3
4 0.303 4833 0.158 627
5
6
7
8 0.584 16379 0.074 2719 0.375 128
9
Base case notes 0.300 Source 8 includes pre-claim drops 0.160 0.375 only one source
*Note: source 2 (CRU) does not distinguish between pre-and post-issue failures; this estimate should be seen as a weighted average of the two.
6
As explained above, it has proved particularly difficult to obtain reliable cost data on sufficient numbers of dropped claims (i.e. those that were withdrawn or repudiated) mainly because there is very little incentive for either side to collect these data. Defendants do not know these costs and claimants cannot normally recover them on a case-specific basis. Only one claimant dataset provided to us (source 6) presented clear information on numbers and base costs for claims that failed. This dataset consisted of 100 personal injury cases that had been closed since January 2000. 27 of these had been closed without payment, and, of these, 12 had been closed prior to a letter of claim being sent, 12 were closed after a letter of claim but before proceedings, and 3 were closed after proceedings. The mean base costs were £418, £1216, and £3725 respectively. Given that these costs had not been subject to cost negotiation, for consistency we applied the average reduction through negotiation observed in our pooled defendant database (20%) to these figures, resulting in mean base costs of £334 (pre-claim), £973 (pre-issue), and £2980 (post-issue) respectively. This compares with the mean base costs for settlements pre- and post-issue of £2227 and £3951 respectively (which we assume had been subject to costs negotiation). While the numbers here are too small to permit a breakdown by type of claim, they do give an indication of the relative cost of failed cases. We suggest a base case assumption that claims dropped pre-issue cost 44% of those settled pre-issue, and claims dropped post-issue cost 75% of those settled post-issue. Moreover, although this dataset showed 12% of client contacts failing to produce a letter of claim after review, indirect information from a much larger claimant database (source 8, Table 3) implied that this figure was probably an underestimate. We felt therefore that a plausible assumption for the proportion of EL accident cases dropped pre-claim was 0.2, at a mean cost of £334 per case. These assumptions, based mainly on a single, relatively small dataset, and not broken down by case type, are amongst the least firm of all our estimates, and we suggest therefore that they in particular are subject to sensitivity analysis in relation to their impact on revenue-neutral success fees.
Success fee estimates Using the base case assumptions described above, we ran a decision tree model in order to compare the expected revenue from undertaking a large number of EL accident cases as CFAs with the expected revenue from running the same cases on an hourly fee basis. The outcome of this exercise is shown in Figure 1 below. At each stage of litigation, it is possible to compare the expected revenues for the two funding options, and the difference can be interpreted as the success fee that would be necessary at that stage to make the expected revenue streams equal. Given that the CFA success fee would normally be agreed at the point of first contact with the client, in what follows we calculate the revenue-neutral success fee at that stage of the process.
8
Figure 1: Base case EL tree for EL accident claims
drop
0.300
0.00 1,500.00 drop
0.160
settle claim made after review
0.800 0.550
1,422.71
0.00 4,800.00 lose 0.00 6,000.00
issue proceedings review
1.000 0.150
settle 3,984.75 0.795 trial
0.045
1,138.17
3,750.00 win
0.375
CFA application
1,138.17
0.625
reject after review
0.200
0.00
reject after phone call
0
0.00 drop
0.300
660.00; P = 0.240 1,500.00; P = 0.440 drop
0.160
settle claimant contact
c 0= 0 c 1= 334 c 2d= r 1*c 2s c 2s= 1500 c 3d= r 2*c 3s c 3s= 4800 c 4l= 6000 c 4w =6000 p0= 0 p1= 0.2 p2= 0.15 p3= 0.3 p4= 0.045 p5= 0.16 p6= 0.625 r 1= 0.44 r 2= 0.75
Hourly fee application : 1,444.64
claim made after review
0.800
0.550
1,722.30
3,600.00; P = 0.019 4,800.00; P = 0.095 lose 6,000.00; P = 0.002 6,000.00; P = 0.003
issue proceedings review
1.000 0.150
settle
0.795 4,662.00
1,444.64 trial
0.045
6,000.00 win
0.375
Hourly fee application
1,444.64
0.625
reject after review
0.200
334.00; P = 0.200
reject after phone call
0
0.00
9
To summarise the base case results in percentage form, we estimated revenue neutral success fee percentages. Tables 4 and 5 provide the results: in table 4 it is assumed that the success fee is applied to all costs incurred throughout the claim (as in Figure 1). In table 5, it is assumed that any claims which go to trial will attract a 100% markup on all costs incurred on that claim (and consequently the revenue neutral success fee applied to the remaining costs are necessarily lower than in table 4).
Table 4: Base case success fees for EL accident cases Expected revenue per case (£) CFA application Hourly fee application 1444.64 26.93% Revenue neutral success fee (%)
1138.17
Table 5: Base case success fees for EL accident cases (with 100% uplift at trial) Expected revenue per case (£) CFA application Hourly fee application 1444.64 24.71% Revenue neutral success fee (%)
1158.42
These base case estimates are of course valid only to the extent that the parameter values used are plausible. As we have noted in several places, some of the data we have used to arrive at the base case estimates are based on relatively small numbers of cases. Moreover, even where the sample sizes are large, issues arise as to the representativeness of the datasets provided to us. For both these reasons, there is a significant degree of uncertainty surrounding the base case estimates, and this should be borne in mind when interpreting them. In our previous reports, we found that the results were most sensitive to the assumptions used in relation to the pre-issue failure rates, and least sensitive to assumptions about mean costs.
10
Conclusion As in our previous reports, we would like to emphasise the following notes of caution when interpreting the outcome of this exercise: 1. As agreed at the beginning of this project, we have defined a “reasonable” success fee as one which would yield just sufficient revenue to offset the foregone revenue resulting from running a large number of EL accident cases as CFAs rather than hourly fees. What is meant by “large” here is open to question: clearly, a firm with a small number of cases will be subject to some year-on-year variation in revenue when using CFAs with the success fees as defined in this report. In some years they will gain and in others they will lose, relative to the use of hourly fees. By contrast, a very large firm with many cases will be expected to observe virtually no difference whether they use CFAs or hourly fees: the choice will be risk-free. It should be possible to estimate the relationship between firm size and CFA risk, but that is beyond the scope of the present study. 2. We should also make clear that our data has not been able to address some of the concerns raised in our communications with the various parties – for example, with respect to the role of Part 36 offers as they affect the balance of risk between claimant and defendant. Moreover, it was put to us by defendants that the hourly rate charged currently by claimant solicitors could have an element in it already which reflects the expected cost of non-chargeable work. 3. An important assumption behind the method used here to arrive at revenueneutral success fees is one of behavioural neutrality. We have assumed that a case run as a CFA will be run in the same way, with the same costs and litigation stages, as one run on an hourly fee basis. We doubt that this assumption is fully justified, but again, this issue is outside the scope of the current study.
11