Winners and Losers in the Court of Appeal: An Empirical Study of Personal Injury Cases (2002-16)

This article reports findings from an empirical study of 458 personal injury cases decided by the Court of Appeal spanning fifteen years. The study used conventional statistical software, SPSS Statistics, and two machine learning platforms, Data Robot and IBM’s Watson, to analyse the dataset. The analysis reveals a general pro-Defendant bias within the Court of Appeal. Although neither Claimants nor Defendants reverse first instance decisions more than 50% of the time, Defendant Appellants reverse more often (47.3%) than their Claimant counterparts (39.5%); Defendants also successfully resist more appeals and are 20% more likely than Claimants to obtain a favourable outcome in appeals overall. These findings are broadly consistent with findings from other studies. However, within a subset of cases involving judges with greater experience of deciding personal injury appeals, there is a shift – albeit slight – in favour of Claimants. The study tested a variety of factors which could potentially explain favourable outcomes in general and this pro-Claimant shift in particular. Those factors included the identity of the Appellant, the type of case, the type of advocate, the legal issues at stake, and the identities of the appeal judges. Controlling for the various factors, the study found that at least one appeal judge within the subset delivered pro-Claimant decisions at statistically significant levels. None of the other factors contributed to favourable outcomes at statistically significant levels. Since a number of potentially pro-Claimant judges retired over the period of the study, it might be anticipated that the pro-Defendant bias will intensify. Such a trend is evident over the last four years of the study: favourable outcomes for Claimants fell from an average of 48% (2002-11) to 37.9% (2012-16) with an absolute low of 26.3% in 2016. Although Claimants win less than Defendants, this dramatic fall in the success rate for Claimants is only partly explained by the increasing proportion of Claimant-initiated appeals over the last four years. Although that proportion rose from an average of 47.1% to 50% over the two intervals, the number of appeals dwindled from 34.2 to 23.2 per annum over the same timeframe. On the other hand, pro-Defendant intensification may help explain why the number of decided personal injury appeals is dwindling despite the increased proportion of such work in the High Court. Whatever the reasons behind the pro-Defendant bias, if Claimants’ legal advisors have the (accurate) impression of an intensification of that bias, that impression may well serve as a powerful disincentive on them to initiate or resist appeals. And if the average reversal rate for any appellant remains or falls significantly below 50%, it is possible that the availability of litigation funding for appeals will be compromised. The combined effect could be further reductions in the number of appeals alongside an entrenchment of pro-Defendant rights within tort law. In the alternative, as the doors to litigation and in particular appeals close, litigants might consider using machine learning technology to obtain more accurate information about the prospects of success and litigation risk in order to compromise appeals on a probabilistic rather than all-or-nothing basis. One possible consequence of this approach, however, is that novel claims would never be litigated unless litigation funders reassess their requirements of Appellants that the prospects of success should exceed 50%. It could potentially leave more individual Claimants under- rather than over-compensated