\documentclass[10pt,a4paper]{scrartcl} % KOMA-Script article \usepackage{xltxtra} %some packages for xelatex \usepackage[english]{babel} \setmainfont[Numbers=OldStyle,Mapping=tex-text]{TeXGyrePagella} \setsansfont[Numbers=OldStyle,Mapping=tex-text]{Fontin Sans} \renewcommand*\sectfont{\sffamily\bfseries} \usepackage{graphicx} \usepackage[bf,sc]{caption} \usepackage{booktabs} \usepackage[usenames,dvipsnames]{color} \usepackage{hyperref} \usepackage{xspace} \usepackage{natbib} \usepackage{setspace} \usepackage{longtable} \usepackage{dcolumn} \usepackage{rotating} \usepackage{subfig} \usepackage[noae]{Sweave} % \usepackage{natbib} \newcommand{\myName}{Chris Hanretty\xspace} \newcommand{\myTitle}{Haves and Have-Nots before the Law Lords\xspace} \hypersetup{% colorlinks=true, linktocpage=true, pdfstartpage=3, pdfstartview=FitV,% breaklinks=true, pdfpagemode=UseNone, pageanchor=true, pdfpagemode=UseOutlines,% plainpages=false, bookmarksnumbered, bookmarksopen=true, bookmarksopenlevel=1,% hypertexnames=true, pdfhighlight=/O, urlcolor=Maroon, linkcolor=RoyalBlue, citecolor=ForestGreen, pdftitle={\myTitle}, pdfauthor={\myName}, } \title{\myTitle} \author{\myName} \setlength\parindent{0mm} \setlength\parskip{4mm} \renewcommand{\tabcolsep}{1mm} \newtheorem{hypothesis}{Hypothesis} \begin{document} %% SETUP <>= library(reporttools) library(memisc) library(car) uk<-read.csv("haves_data.csv",header=T) uk<-uk[!is.na(uk$rel.status),] uk$rel.experience[is.na(uk$rel.experience)]<-0 uk$app.status<-factor(uk$app.status,levels=c("Person","Association","Business","Local govt.","National govt."),ordered=TRUE) uk$resp.status<-factor(uk$resp.status,levels=c("Person","Association","Business","Local govt.","National govt."),ordered=TRUE) uk$source.bak<-uk$sourcect uk$sourcect<-car:::recode(uk$sourcect, "c('administrative agency','national top court','original jurisdiction','specialized court','other (includes certification)')='Direct appeal'; 'major trial court'='High Court';'intermediate appellate court'='Court of Appeal';else=NA") uk$sourcect<-factor(uk$sourcect) uk$sourcect<-relevel(uk$sourcect,"Court of Appeal") uk<-uk[!is.na(uk$sourcect),] uk$area<-factor(uk$area) uk$area<-relevel(uk$area,"Other") uk$decade[uk$decade==60]<-70 uk$decade<-factor(uk$decade,levels=c("70","80","90","10"),ordered=FALSE) uk$privy<-as.numeric(uk$sourprov>4) uk$hra<-0 uk$hra[which(uk$year>1997 & uk$app.status=="Person" & uk$resp.status %in% c("Local govt.","National govt."))]<-1 uk$hra[which(uk$year>1997 & uk$resp.status=="Person" & uk$app.status %in% c("Local govt.","National govt."))]<- -1 uk$orig.liberal<-0 ## If it was a liberal decision, and the appelant won, then the original decision must have been conservative uk$orig.liberal[which(uk$direct1=="Pro" & uk$appwin==1)]<- -1 ## If it was a conservative decision, and the appelant won, then the original decision must have been liberal uk$orig.liberal[which(uk$direct1=="Con" & uk$appwin==1)]<-1 ## If it was a immigration case, or a libel case, the normal order is reversed uk$orig.liberal[which(uk$direct1=="Con" & uk$appwin==1 & uk$issue1 %in% c("immigration, deportation, citizenship","libel or defamation"))]<- -1 ## If it's a dispute between governments, it'sneither uk$orig.liberal[which(uk$direct1=="Con" & uk$appwin==1 & uk$issue1 %in% c("disputes between units of government"))]<- 0 uk$plaintif2<-car:::recode(uk$plaintif, "'plaintiff is appellant'=1;'plaintiff is respondent'=-1;else=0") uk$plaintif2<-as.numeric(as.character(uk$plaintif2)) uk$plaintisappellant<-as.numeric(uk$plaintif2==1) uk$oralbeg<-gsub("00$","01",uk$oralbeg) uk$oralend<-gsub("00$","01",uk$oralend) uk$oralbeg<-gsub("32$","31",uk$oralbeg) uk$oralend<-gsub("32$","31",uk$oralend) uk$complexity<-as.Date(as.character(uk$oralend),format="%Y%m%d")-as.Date(as.character(uk$oralbeg),format="%Y%m%d") uk$complexity[uk$complexity==0]<-1 uk$complexity<-abs(as.numeric(uk$complexity)) uk$complexity<-log(uk$complexity) uk$complexity[is.na(uk$complexity)]<-mean(uk$complexity,na.rm=T) uk$complexity2<-uk$endpg-uk$beginpg uk$complexity2[uk$complexity2==0]<-1 uk$complexity2<-log(uk$complexity2) uk$complexity3<-rowMeans(cbind(as.vector(scale(uk$complexity)),as.vector(scale(uk$complexity2)))) uk$app.win<-uk$wins.appatty1 uk$resp.win<-uk$wins.resatty1 uk$app.win[is.na(uk$app.win)]<-mean(uk$app.win,na.rm=T) uk$resp.win[is.na(uk$resp.win)]<-mean(uk$app.win,na.rm=T) uk$rel.wins<-uk$app.win-uk$resp.win @ \maketitle %#$ \doublespacing \begin{abstract} One important characteristic of justice, and \emph{a fortiori} of our judicial system, is impartiality. % One type of impartiality in judicial practice is impartiality between litigants who command status and material resources -- the `haves' -- and litigants who lack resources, the `have-nots'. % Investigation of relative status advantage in litigation outcomes, which springs from the work of Marc Galanter, has talked past a particularly British tradition emphasizing the conservative bias of the judiciary, and in particular its defence of property owners. % In order to investigate these charges, I investigate the success of appeals to the House of Lords between 1969 and 2003 using logistic regression. % I build both on general theories of relative status advantage and the advantages of `repeat players', and on more particularly British interpretations of the judiciary. % I find partial support for theories of relative status advantage, insofar as governmental actors have significant advantages over all other actors, but businesses and associations have no advantages over individual litigants. % Contrary to expectations of a uniformly conservative judiciary, appellants challenging liberal outcomes were less likely to succeed rather than more. % \end{abstract} \section{Introduction} In 1990 McDonald's issued a writ for libel against Helen Steel and David Morris. % Steel and Morris had distributed a pamphlet accusing McDonald's of economic imperialism, mass pollution, and exploitation. % Steel and Morris were forced to represent themselves in the eventual libel trial, having been denied legal aid. % The trial began in June of 1994 and finished in December of 1996, becoming the longest trial (civil or criminal) in English legal history. % The disparity of resources between the two sides meant that McDonald's victory, when it came, was entirely Pyrrhic. % Although Mr Justice Bell ruled that McDonald's was not `to blame for starvation in the Third World\ldots{} [or] the eviction of small farmers or anyone else from their land',\footnote{~\emph{McDonald's Corporation v Steel \& Morris} [1997] EWHC QB 366, Mr Justice Bell's summary, \S{}\S{}86--89. } McDonald's declined to collect the £40,000 in damages they were awarded. % Nor was McDonald's victory permanent: Steel and Morris eventually appealed the judgement to the European Court of Human Rights (ECHR), which found in their favour, arguing that ``the denial of legal aid to the applicants deprived them of the opportunity to present their case effectively before the court and contributed to an unacceptable inequality of arms with McDonald's''.\footnote{~\emph{Steel and Morris v. the United Kingdom}, no. 68416/01, \S{}72, ECHR 2005.} The ECHR's recognition that justice can be vitiated by inequalities between litigants would not, I suggest, come as a surprise to either the many non-expert observers of the `McLibel' trial,\footnote{~David Pannick QC argued that the case ``has achieved what many lawyers thought impossible: to lower further the reputation of our law of defamation in the minds of all right-thinking people''. `Exposing the Flaws in Britain's Libel Laws', \emph{The Times}, 20th April 1999. } or to social scientists who study the operation of the courts. % Indeed, ever since the publication of Marc \citeauthor{galanter1974haves}'s \citeyear{galanter1974haves} article ``Why the `Haves' Come Out Ahead: Speculations on the Limits of Legal Change'', social scientists in general and political scientists in particular have been interested in testing whether `haves' -- in particular, government and business actors -- are more likely to succeed in litigation than `have-nots', whether through greater advantages afforded to `haves' in the law, or through the selection and retention of better, more experienced counsel. % Scholars of US law and politics generally concur\footnote{~\citet{sheehan1992ideology} is an exception. } on the existence of relative status advantage, and disagree about conditions -- such as judicial ideology \citep{black2010us}, the role of amicus curiae \citep{songer2000haves} and case salience \citep{mcatee2007lawyers} -- which exacerbate or mitigate it. % Scholars outside of the US have greater reservations about relative status advantage, finding either a strongly significant \citep{szmer2007does} or moderately significant \citep{flemming2002repeat} effect once lawyer status is controlled for, or no effect once lawyer status is controlled for \citep{haynie2007experienced}; an overall extremely limited effect \citep{yoav1999haves} or an overall extremely strong effect \citep{mccormick1993party} -- even a reversed effect in the Philippines, where have-nots come out ahead \citep{haynie1994resource}. % This research agenda has, with one exception \citep{atkins1991party}, ignored the various British legal systems. % This most likely stems from comparative dearth of expertise in empirical legal studies amongst those who study the English, Welsh and Scottish systems \citep{genn2006law}. There are, however, principled reasons for ignoring these British cases. % Notional adherence to the cab rank rule and restrictions on entry to the legal profession meant that litigants get, in theory, a random draw from a pool of counsel which is relatively homogeneous with respect to quality. % Add to this the fact that, as \citeauthor{atkins1991party} noted, ``scholars of English politics usually assume that courts do not participate in the process by which resources and values are allocated by the political system'', and that, since judging partly on the basis of political preferences is one of the ways in which relative status disadvantage might occur, English courts should be expected to be relatively free of a great advantage for the haves over the have-nots. % Against this we might set the original observation made by Galanter that the British legal profession, in virtue of its relatively closed and legalistic nature (at least compared to the American bar), furnishes considerable advantages to what Galanter dubbed `repeat-players': litigants or counsel who have already appeared before courts, and so bring a competitive advantage to their next encounter. % In the case cited above, both McDonald's and their counsel were repeat players in the libel litigation industry, and, at least domestically, wielded that repeat player advantage -- which accrued to McDonald's because of its status advantage -- to great effect. % Yet this research agenda -- focused as it is on one particular mechanism by which the legal system recapitulates social status -- has talked past another very British line of attack on the judiciary. % \emph{Contra} Atkins, JAG \citet[199]{Gri85} long argued that the British judiciary systematically and relatively universally favours ``the protection of property rights, and\ldots{} political views normally associated with the Conservative party''. % If `haves' are advantaged by British courts, it is not because of the advantages they deploy, but rather because judges would find in their favour in any case due to their personal convictions -- channelled through a particular interpretation of the law. % Griffith's charge has perhaps been dulled somewhat since the passage of the Human Rights Act, and the accusation that judges are becoming activist proponents of expansive human rights claims. % Nevertheless, Griffith's argument -- if correct -- would offer another argument why higher status parties come out ahead. % Unfortunately, we can say little about whether relative status advantage exists in Britain today. % Although \citet{atkins1991party} employed linear discriminant analysis to model the probability of appellate success in the Court of the Appeal between 1983 and 1985, the linear discriminant function included a number of other variables unrelated to status (panel size, judge status, case source, area of law, and strength of claim), and whilst the linear discriminant function as a whole was statistically significant, we cannot test these variables individually.\footnote{~There are additional questions surrounding Atkins' analysis -- in particular whether the independent variables meet the requirement in linear discriminant analysis of being multivariate normal, and whether the linear discriminant function specifies properly the structure of the interactions between different categories of litigants. } % This article therefore attempts to contribute to our knowledge of relative status advantage and the British legal systems. % Using the data from the High Courts Judicial Database \citep{SLH+07}, I model appellant success in the Appellate Committee of the House of Lords between 1969 and 2003 using logistic regression. % I find that a pure relative status advantage effect -- that is, an advantage which accrues solely in virtue of the category to which the respective parties belong, whether private individuals, associations, businesses or local or central government actors -- is present (and positive) only for government litigants. % This government advantage remains even when controlling for the relative expertise of counsel -- which, unlike relative win-rates for counsel, has a significant positive effect on litigant success. % I proceed as follows. % In section \ref{sec:theory}, I describe the theory behind relative status advantage, and list a number of hypotheses concerning different aspects of relative status disadvantage, as well as other factors which might affect appellant success. % In section \ref{sec:data}, I describe the data I use and my modelling strategy, in particular in relation to the coding of variables pertaining to relative status. % In section \ref{sec:analysis}, I present a number of regression models, and comment upon them. % I close by discussing why the particular forms of relative status advantage I identify appear in the UK -- and why others, which arguably exist in other jurisdictions, do not exist here. % \section{Theory} \label{sec:theory} The central insight of Galanter's remarkably fecund article is that haves enjoy a series of advantages in litigation. % These advantages accrue to `haves' through multiple mechanisms. % Not all of these mechanisms have been addressed in the extant literature. % One important part of Galanter's theory is that haves, since they are often larger organisations, or have more extended interests, are more likely to be involved in more than one dispute. % They are thus `repeat players', who enjoy significant advantages over one-shot players in virtue of their superior knowledge (a trifecta of `knowing how', `knowing that', and knowing who'). % The paradigmatic example of the high status repeat player is central government, which has extensive and considerable interests and constantly and continuously litigates before courts, both in criminal and in public law cases. % Much work has gone in to studying the nature of the advantage enjoyed by the office of the Solicitor General in the United States -- including deflationary accounts which suggest the advantage is entirely due to the experienced accumulated by government lawyers (on which point see below) \citep{mcguire1998explaining}. % Less work has gone in to studying the pure repeat-player advantage of other actors, although studies of criminal repeat players seem to suggest no pure repeat player advantage. % This has licensed some pessimistic conclusions: \citet[1103]{lempert1999classic} \begin{quote} ``We cannot be sure from the empirical work that there is even an interaction effect because we are missing the crucial comparison between powerful parties who are repeat players and those who are not'' \end{quote} Studying a pure repeat player effect is not possible with the present data, because it overlaps so tightly with governmental status. % There are extremely few non-governmental repeat players -- the General Medical Council and the Law Society are two exceptions, but even they have only contested a handful of cases listed in the data. % Even without a repeat player advantage, argues \citet[123]{galanter1974haves}, `haves' would still enjoy advantages because the rules of the game ``tend to favor older, culturally dominant interests''. % This argument is perhaps more pertinent to the United States, where the relatively weak institutionalization of political parties and the greater number of veto points compared to the British system has meant that the political system is more easily permeated by particularistic interests, and corporate interests in particular, which in turn write the rules to their advantage \citep{richter2009lobbying}. % Nevertheless, whilst legal rules may not secure corporations or large business much advantage, there is less doubt that British law affords government particular favour. % For a considerable time, ministers have enjoyed a considerable advantage in public law cases, insofar as their actions must be ``so outrageous in its defiance of logic or accepted moral standards that no sensible person who had applied his mind to the question to be decided could have arrived at it'' if they are to be quashed.\footnote{~\emph{Associated Provincial Picture Houses v Wednesbury Corporation} [1947] 1 KB 223} % Although this Wednesbury test has fallen by the wayside, considerable obstacles still face those arguing against the government. % Because higher status litigants are so often repeat players, the confluence of these two factors -- the pure repeat player effect, and a rules-favouring-higher-status litigants effect -- is captured by the following hypothesis: \begin{hypothesis} Higher status litigants will be more likely to succeed in appeals than lower status litigants \end{hypothesis} One extremely important mechanism through which higher status litigants procure an advantage in litigation outcomes is by hiring different counsel. % More experienced or better counsel may have a stronger professional or personal acquaintance with the judges hearing the appeal. % They may know how to construct arguments that are either better in general, or better for those particular judges who must be convinced. % Strong support for this entirely reasonable supposition comes from the United States Supreme Court, where \citet{johnson2006influence} exploited oral argumentation grades privately noted by Justice Blackmun to demonstrate that more experienced counsel deliver better oral arguments, which in turn affect justices' votes on the case. % Given the very limited time allotted to oral argumentation before the Supreme Court (one hour compared to a soft limit of one full day in the House of Lords), this effect ought to be much stronger in our present data. % At the same time, there may be an independent experience effect, for lawyer experience still has a (positive) effect on outcomes even when the quality of oral argumentation is taken into account \citep{mcatee2007lawyers}. % Consequently, some scholars have taken to using a more direct proxy for quality, such as win-rates in prior cases, and have found that this nullifies the effect of experience \citep{haynie2007experienced}. % Consequently, we may hypothesize both that \begin{hypothesis} Litigants who employ more experienced counsel will be more likely to suceed in appeals than litigants who employ less experienced counsel \end{hypothesis} and \begin{hypothesis} Litigants who employ more successful counsel will be more likely to suceed in appeals than litigants who employ less successful counsel \end{hypothesis} The role of the relative \emph{number} of counsel is an issue which, though not unique to the United Kingdom \citep{szmer2007does}, has particular relevance there. % For a long time\footnote{~The \citet[\S{}39]{MaMC76} could not identify the precise point at which the rule came into force, though the first reference to the rule seems to date back to 1871. } the Bar Council prevented Queen's (or King's) Counsel from appearing in a court of law without junior counsel. % The two-counsel rule imposed considerable cost on litigants, especially those who wished to contest in a case in the highest courts, where Queen's Counsel were, and in the Supreme Court remain, a virtual necessity. % Although the rule was formally abolished in 1977, the practice of `double-manning' remained prevalent,\footnote{~\citet[83]{abel2003english} reports that the Government in 1990 reported that ``a two-year campaign to reduce double manning had persuaded QCs to appear without juniors in only eight of 1,700 suitable cases''. } largely because of a strongly held belief that in complex cases (where it was judged necessary to retain Queen's Counsel), two barristers were required. % As Lord Hailsham put it, it is `impossible' to conduct a case lasting more than one day `without having two hands to the pump' (quoted in \citealp[90]{abel2003english}). % If this rhetoric is more than self-serving cant, we should expect that the relative number of counsel should have a significant impact on litigant success. % That is, a team composed of two barristers (QC and junior, or indeed two juniors) should handily defeat a sole barrister, as should three counsel against two, and so on. % Thus, \begin{hypothesis} Litigants who employ more counsel will be more likely to suceed in appeals than litigants who employ fewer counsel \end{hypothesis} The final hypothesis concerning raw numbers relates to third party intervenors as they are most commonly called in the UK, or \emph{amicus curiae} as they are known in the United States.\footnote{~Whilst ``in the United States [the] adversarial aspect of amici curiae became the basis for third party interventions in general\ldots{} [in] the UK\ldots{} the \emph{amicus curiae} (nowadays referred to as an `advocate to the court') remains a largely non-partisan figure, appointed by the Attorney General at the behest of the court \citep[8]{JM09}. } % \citet{songer2000haves} have argued that third party intervenors constitute a resource for the party on whose behalf they intervene, and that since in several United States State Supreme Courts a number of amicus curiae often appear on the side of have-nots, they mitigate the relative status advantage that such parties normally enjoy. % Of course, the minor premise in this argument -- that third party intervenors often appear on the side of the disadvantaged -- is an empirical one, and may not hold in the UK case. % Indeed, because the role of third party intervenors in UK domestic courts (as opposed to the Appellate Committee's work when its members decide cases referred to the Privy Council) is so recent (public bodies began to intervene in cases heard in the Lords from the late nineteen seventies onwards; NGOs from 1995: \citealp[11]{JM09}), we cannot furnish reliable generalisations about the intentions of third party intervenors -- although a cursory examination of the role of intervenors in recent UK Supreme Court cases would suggest that intervenors intervene overwhelmingly on behalf of have-nots. Nevertheless, we may state our hypothesis in broad terms which allow us to test for a pure intervenor effect rather than an intervenor effect which mediates relative status advantage discussed above: % \begin{hypothesis} Litigants who are joined by more intervenors will be more likely to suceed in appeals than litigants who are joined by fewer intervenors \end{hypothesis} Generally, only a minority of appeals heard by the Court of Appeal or by the Law Lords succeed \citep{atkins1990interventions}. % It is not clear why this should be the case. % In a simple model of appeal \citep{shavell24steven} where appellate courts exist solely to correct errors made by trial courts, appellants are rational and have information concerning the likelihood of trial court error, and where no cases are settled between original trial and the appeal, reversal rates ought to be high, as potential appellants whose cases are \emph{ex ante} unlikely to succeed simply withdraw. % If we expand the model to allow for judicial discretion in hearing appeals, the reversal rate ought to be even higher. % The US Supreme Court, the paradigmatic example of a court with considerable discretion over its docket, has extremely high reversal rates. % The discretion exercised by the Lords, whilst not quite comparable to that exercised by justices of the USSC, was still important. % Indeed, the Lords' discretion led \citet{atkins1990interventions} to predict similar rates of reversal between the Lords and the US Supreme Court, despite findings that rates of reversal in cases granted leave to appeal by the Law Lords themselves are roughly equivalent to rates of reversal in cases granted leave to appeal by the Court of Appeal \citep{blom1972final}. Of course, appellants are not strictly rational, and appeals are not simply about error correction or judicial discretion. % Appellants, even unsuccessful applicants, may derive catharsis from being heard on appeal \citep{shapiro1979appeal}. % Representative or accountable litigators such as government departments or public prosecutors may appeal hopeless cases in order to justify to those to whom they are accountable that they tried their hardest (or perhaps as part of a minimax regret strategy). % Trial or intermediate appellate courts may simply get things right more often than not. % We may therefore hypothesise that \begin{hypothesis} Respondents will be more likely to succeed than appellants \end{hypothesis} A penultimate hypothesis concerns the source of the appeal. % The English and Welsh legal system is somewhat unusual in having an intermediate level of appeal and allowing it to be by-passed in certain cases through a special `leap-frog' procedure, whereby cases are appealed directly to the House of Lords from the High Court. % Appeals may also arrive at the Lords (sitting as the Privy Council) directly in a rag-bag of areas including appeals over professional registration of doctors, dentists and vets, and in some ecclesiastical disputes. % We might expect these cases to involve higher rates of reversal, and thus to favour the appellant. % First, leap-frog appeals are not appeals as of right, but must be granted leave by the Lords Appeal Committee. % Since such cases involve greater discretion, we might expect greater reversal rates in line with the thinking above. % Second, leap-frog appeals also require certification from a High Court judge -- which may perform a valuable function in signalling an important and finely balanced case. % Third, we might expect the quality of judging in the High Court to be less than the quality of judging in the Court of Appeal, and thus to require a firmer hand on the part of the Lords. \begin{hypothesis} Appellants from the High Court or other direct appeals will be more likely to succeed than appellants from the Court of Appeal \end{hypothesis} The final hypothesis is the most overtly `political' hypothesis. % As noted above, JAG Griffith argued strongly that the British judiciary was both small-c and capital-c Conservative. % If judges are uniformly conservative, then this has no obvious consequences for appellate courts -- the same (conservative) outcome will be reached no matter what. % If, however, judges are merely disproportionately conservative, there is the possibility of unlikely congeries of `liberal' judges making a `liberal' decision. % In this case, it becomes more likely that an appeal court, disproportionately likely to be staffed by conservative judges, will overturn the ruling. % Thus: \begin{hypothesis} Appellants seeking to overturn a liberal ruling will be more likely to succeed than appellants seeking overturn other rulings. % \end{hypothesis} This concludes the section discussing hypothesis. % I now turn to discuss the data I use and the measures I construct in order to assess these hypotheses. % \section{Data} \label{sec:data} As mentioned, my data come from the High Courts Judicial Database, which includes every case heard either by the Judicial Committee of the Privy Counsel, or the Appellate Committee of the House of Lords, and which was reported in the All England Law Reports. % I have excluded cases heard by the Privy Counsel which originated outside the UK, but have retained a limited number of cases where the Privy Council heard domestic cases. % 1298 cases argued between 1969 and 2003 are included. % \paragraph{Dependent variable.~} % The dependent variable is a dichotomous variable which tracks whether the appellant succeeded in their appeal ($y=1$) or did not succeed ($y=0$). % References to other courts, or cases where the appeal was allowed in part but dismissed in part, are scored 0. % The frequencies of appellant success, along with the raw counts of cases, are displayed in Table \ref{tab:mytable}. % \begin{table}[tbp] \centering \caption{Success rates by appellant (rows) and respondents (columns)} \label{tab:mytable} \scriptsize <>= my.xtab<-tapply(uk$appwin,list(Appellant=uk$app.status,Respondent=uk$resp.status),function(x)paste(sum(x),"/",length(x)," (",format(sum(x)/length(x)*100,digits=3),"%)",sep="")) a<-tapply(uk$appwin,list(Appellant=uk$app.status,Respondent=uk$resp.status),function(x)paste(sum(x),"/",length(x),sep="")) a.col.grand<-by(uk$appwin,uk$app.status,function(x)paste(sum(x),"/",length(x),sep="")) b<-tapply(uk$appwin,list(Appellant=uk$app.status,Respondent=uk$resp.status),function(x)paste(format(sum(x)/length(x)*100,digits=3,nsmall=1),"%",sep="")) b.col.grand<-by(uk$appwin,uk$app.status,function(x)paste(format(sum(x)/length(x)*100,digits=3,nsmall=1),"%",sep="")) union<-cbind(cbind(a,a.col.grand),cbind(b,b.col.grand)) a.row.grand<-t(by(uk$appwin,uk$resp.status,function(x)paste(sum(x),"/",length(x),sep=""))) b.row.grand<-t(by(uk$appwin,uk$resp.status,function(x)paste(format(sum(x)/length(x)*100,digits=3,nsmall=1),"%",sep=""))) row.grand<-cbind(a.row.grand,b.row.grand,"","") row.grand<-row.grand[,c(1,6,2,7,3,8,4,9,5,10,11,12)] union<-union[,c(1,7,2,8,3,9,4,10,5,11,6,12)] union<-rbind(union,t(as.vector(row.grand))) colnames(union)[c(2,4,6,8,10,12)]<-"pct." colnames(union)[11]<-"Grand total" rownames(union)[6]<-"Grand total" colnames(union)<-gsub("National","Natl.",colnames(union)) rownames(union)<-gsub("National","Natl.",rownames(union)) colnames(union)<-gsub("govt","gov",colnames(union)) rownames(union)<-gsub("govt","gov",rownames(union)) my.xtab<-xtable(union,align=c("l", "D{/}{/}{-1}","D{.}{.}{2}", "D{/}{/}{-1}","D{.}{.}{2}", "D{/}{/}{-1}","D{.}{.}{2}", "D{/}{/}{-1}","D{.}{.}{2}", "D{/}{/}{-1}","D{.}{.}{2}", "D{/}{/}{-1}","D{.}{.}{2}")) print(my.xtab, floating=FALSE, hline.after=NULL, sanitize.colnames.function=function(x){ paste("\\multicolumn{1}{c}{",x,"}",sep="") }, add.to.row=list(pos=list(-1,0, nrow(my.xtab)), command=c('\\toprule ', '\\midrule ', '\\bottomrule '))) @ \end{table} \paragraph{Relative status.~} % There are a number of variables relating to status. % These are all constructed on top of two variables classifying the appellant status and the respondent status. % Five status categories are used: \singlespacing \begin{itemize} \item individual persons \item associations (including trades unions, business associations, religious or charitable organisations, and political parties) \item businesses \item local or regional government \item national government \end{itemize} \doublespacing The presumption is that the status of associations is greater than the status of individuals, that the status of businesses is greater than the status of associations, and so on. % This is a common presumption in the literature, and refinements upon such schema typically involve elaborating on one of the categories here rather than offering completely different categorisations. %% REFERENCE? When constructing a variable to measure relative status, it is common to use these ordered categories as if they were interval variables: that is, to score each category one through five, and use to subtract the respondent's status from the appellant's status in order to get a further interval measure of relative status advantage. % If the measure is positive (negative), the appellant (respondent) has higher status than the respondent (appellant). % This assumes that the ordering of categories is correct, but this assumption is often not warranted, particularly, as in this case, where little prior information exists concerning the nature of status advantage. % Measures of relative status coded in this way may still have significant effects -- but these significant effects may result mask a strong government-against-all-others effect coupled with a negative business-against-all-others effect. % Consequently, I employ a number of variables to tap different aspects of relative status advantage. % There are ten different unique pairings of these five different categories: % \singlespacing \begin{enumerate} \item individual person versus association \item individual person versus business \item individual person versus local government \item individual person versus national government \vskip .5em \item association versus business \item association versus local government \item association versus national government \vskip .5em \item business versus local government \item business versus national government \vskip .5em \item local government versus national government \vskip .5em \end{enumerate} \doublespacing I score each of these as 1 if the first named category appeared as appellant (and the second named category appeared as respondent), -1 if the first named category appeared as respondent (and the second named category appeared as appellant), and zero if the pairing does not apply. % Because the lower status category is listed first, and because this is a model of \emph{appellant} success, we should expect that the coefficients on all of these variables will be negative, and that coefficients will decrease within each group (that is, the coefficient on individual persons versus national governments will be negative and less than than the coefficient on individual persons against associations, implying that individuals facing governments face a harder task than individuals against associations). % These variables thus allow us to tap different types of relative status, and see whether they confer particular advantages. % \paragraph{Counsel characteristics.~} % There are three variables pertaining to characteristics of counsel. % The \emph{relative number of counsel} is simply the number of counsel for the appellants less the number of counsel for the respondents. % The \emph{relative experience of counsel} is calculated as follows. % Let the experience of any individual barrister in any given equal the number of cases that barrister has argued in the Lords up to and including that case. % Then let the experience of the appellant (respondent) counsel equal the sum of the logged experiences of all the barristers acting for the appellant (respondent). % The relative experience is simply the experience of appellant counsel less the experience of respondent counsel. % This way of calculating experience -- which is equivalent to the geometric mean where two or more barristers are aggregated -- is useful here because of the large disparities in experience between senior counsel and juniors. % The \emph{relative success of counsel} is simply a running percentage of cases won by the lead barrister in the appellant team, minus the same figure for the lead barrister in the respondent team. % In principle it would also be possible to take account of the win rates of junior counsel, but averaging out wins rates across the entire team seems to undervalue the contribution of senior counsel, and there are no other obvious suggestions for how to weight the marginal contribution of another counsel's win rate when these rates derive from different numbers of cases argued. % I also include the simple \emph{relative number of intervenors}, though this is the relative number of intervening parties, not the relative number of counsel retained by intervenors. % \paragraph{Case characteristics.~} % In order to test whether judges are disproportionately likely to reverse `liberal' rulings, I use data on the Lords' own decision and the identity of the victorious party. % That is, there is a \emph{liberal status quo ex ante} (coded 1) if the outcome in the Lords is `conservative' and the appellant wins; there is a \emph{conservative status quo ex ante} (coded -1) if the outcome in the Lords is liberal and the appellant wins. % All other cases are coded zero. % This assumes that litigants do not appeal cases which they have `won' in the sense of securing their preferred outcome, and that reversing a conservative decision means taking a liberal decisions, and vice versa. % In part, this assumption derives from the structure of the coding of liberal and conservative outcomes, which tend to be dichotomous. % In the HCJD, liberal outcomes are victories for defendants in criminal trials, for persons alleging infringement of human rights in civil liberties cases, for economic underdogs in private law, for injured parties in torts, and for the government in public law (except in the case of benefits recipients or government employment, where this is reversed). % Conservative outcomes are the reverse of these. % This follows standard North American usage, and I have only amended to the effect that victories for the government in immigration and citizenship cases are coded as conservative rather than liberal, as are victories for putative victims of libel (which would normally be coded as liberal, since they have suffered a tortious injury). The \emph{source court} is a categorical variable with three values -- direct appeal (most often from Immigration or other Tribunals), appeal from the High Court, and appeal from the Court of Appeal (the reference category). I also include a number of dummy variables relating to the \emph{broad area of law}. % Dummies for civil liberties cases, criminal cases, public law cases, tort cases, and family and bar law are included, with all other cases as the reference category. % \section{Analysis} \label{sec:analysis} \renewcommand{\tabcolsep}{0mm} \begin{table}[tbp] \centering \caption{Regression on appellant success} \label{tab:myregression} \small <>= my.mod0<-glm(appwin~PersonvAssociation+PersonvBusiness+PersonvLocalGov+PersonvNatGov+AssociationvBusiness+AssociationvLocalGov+AssociationvNatGov+BusinessvLocalGov+BusinessvNatGov+LocalGovvNatGov+rel.counsel+rel.fidu+rel.experience+rel.wins+sourcect+orig.liberal,data=subset(uk,!is.na(rel.status) & privy==0),family=binomial(link="logit")) my.mod1<-update(my.mod0,.~.+decade+area) PCP<-c(mean((predict(my.mod0,type="response")>0.5)==my.mod0$y), mean((predict(my.mod1,type="response")>0.5)==my.mod1$y)) reg.out<-mtable("Baseline"=my.mod0,"Full"=my.mod1,coef.style="horizontal") reg.out<-relabel(reg.out, "(Intercept)"="(Intercept)", "PersonvAssociation"="Person v. Association", "PersonvBusiness"="Person v Business", "PersonvLocalGov"="Person v Local Govt.", "PersonvNatGov"="Person v Natl. Govt.", "AssociationvBusiness"="Association v Business", "AssociationvLocalGov"="Association v Local Govt.", "AssociationvNatGov"="Association v Natl. Govt.", "BusinessvLocalGov"="Business v Local Govt.", "BusinessvNatGov"="Business v Natl. Govt.", "LocalGovvNatGov"="Local Govt. v Natl. Govt.", "rel.counsel"="Relative no. counsel", "rel.fidu"="Relative no. intervenors", "rel.experience"="Relative experience, counsel", "rel.wins"="Relative success, counsel", "sourcectDirect appeal"="Direct appeal", "sourcectHigh Court"="Appeal from High Court", "orig.liberal"="Liberal status quo ex ante", "decade80"="1980s / 1970s", "decade90"="1990s / 1970s", "decade10"="2000s / 1970s", "areaCivil liberties"="Civil liberties", "areaCrime"="Crime", "areaPrivate"="Private law", "areaPublic law"="Public law", "areaTort"="Tort") write.mtable(reg.out,file="modelholder.tex",forLaTeX=TRUE) @ \include{modelholder} \end{table} Table \ref{tab:myregression} reports the results of two regression models of appellant success: one baseline model which omits dummies for decade and area of law, and a second fuller model which includes these variables as controls. % Both models are very similar in terms of their predictive ability and (with one exception concerning the interpretation of the intercept) are similar in terms of the variables which have statistically significant effects. % Insofar as the first model has greater parsimony with marginally greater predictive power, it should be preferred. % As is common with models of appellate success, the variation explained (in either model) is relatively limited, whilst the number of decisions correctly predicted is beguilingly high: \Sexpr{round(PCP[1]*100,digits=3)}\% of decisions are correctly predicted by the baseline model, and \Sexpr{round(PCP[2]*100,digits=2)}\% by the full model. % \paragraph{Status.~} % The parameters for variables related to status reflect the log odds of success for the first named category. % Thus, according to the full or reduced models, individual persons litigating against central government (whether as appellants or as respondents) are very roughly three-fifths as likely to succeed ($e^{% <>= cat(format(my.mod1$coef[grep("ersonvNa",names(my.mod1$coef))],digits=3),"}=",format(exp(my.mod1$coef[grep("ersonvNa",names(my.mod1$coef))]),digits=2)) @% $) compared to the baseline odds -- or, put equivalently (and more exactly), governments are one and three-quarter times as likely to succeed than litigants in any other case. % The status-related parameters show that the only status differences which are statistically significant are those involving a governmental actor. % Individual litigants face a disadvantage against both local government and central government; this disadvantage is greater when facing central government, as was to be expected. % Businesses face a disadvantage against central government which is less in magnitude than the disadvantage faced by individuals against central government -- but businesses suffer no disadvantage against local government. % Finally, local government suffers from a very considerable disadvantage against central government, with central government more than two and a half times more likely to win. % None of the remaining status-related parameters come close to statistical significance. % Some of the signs on these coefficients are in the predicted direction: thus, individuals face a (statistically insignificant) disadvantage against associations, associations face a (statistically insignificant) disadvantage against national government, and businesses face a (statistically insignificant) disadvantage against local governments. % Yet some of the signs are in the opposite direction to that predicted. % Some of these may result from a limited number of cases -- a number of (complex and drawn-out) appeals involving associations have considerable leverage within the model. % Yet it is interesting to note that individuals face no statistically significant disadvantage against businesses, controlling for other factors, in particular the relative experience of counsel. % To return to the example discussed at the beginning of this article, this suggests that Steel and Morris, had the Lords decided to hear their appeal against the Court of Appeal decision which found against them, and had they been able to retain Keir Starmer QC for that case (as they ultimately did in their appeal to the ECHR), they would have faced no disadvantage in court. % \paragraph{Counsel.~} In terms of characteristics related to counsel, only the relative experience of counsel has any statistically significant effect on the odds of winning. % Although the coefficients on the variables related to the relative success and number of counsel and intervenors all have the correct sign, none reach conventional levels of statistical significance. % Relative experience is however both statistically and substantively significant. % Imagine an appellant has a choice between two barristers, one of whom has argued ten cases before the Lords, the other just one. % By instructing the former barrister, the odds of winning the case improve by 24\%. % The absence of a statistically significant effect of relative success is perhaps unsurprising in light of the other findings of the model. % Some barristers specialize in certain areas of law -- asylum and immigration, criminal defence -- where their clients face significant competitive disadvantages. % They are hardly maximizing their success rate. % Given two barristers with identical win rates, one of whom specialized in criminal defence and the other specialized in company law, we would expect the barrister specialized in criminal defence to be the better lawyer, all things considered. % That is: win rates are no doubt pleasing, but they do not confer magical powers on counsel. % \paragraph{Case characteristics.~} The most obvious characteristic of the case is whether the appellant won or lost, and whether appellants in general are more likely to succeed than not. % Because of the way the model is formulated, the intercept has a substantive interpretation here as the \emph{probability} (rather than the odds-ratio) of appellant success. % In the reduced model (which I have indicated as being preferable on the grounds of greater predictive power and greater parsimony), the intercept is negative, and the probability of appellate success is less than 50\%, at $p=\frac{e^{ <>= intercept<-my.mod0$coefficients[1] cat(format(intercept,digits=3)) @% }}{1+e^{ <>= cat(format(intercept,digits=3)) @% }}= <>= cat(format(exp(intercept)/(1+exp(intercept)),digits=3)) @ $. % Thus, the hypothesis concerning the disadvantages faced by appellants is confirmed. % Although the intercept (and thus the probability of appellant success) ceases to be significant different from zero when the broad area of the law is taken into account, none of these area dummies is itself significant. In terms of the origin of the appeal, whilst appeals direct from the High Court were more likely to be successful (in line with expectations), the effect was not significant, and other direct appeals were less likely to be successful. % The decade in which the decision was reached had no significant effect, nor did most of the dummies for the broad area of law. What did have a significant effect was the ideological direction of the \emph{status quo ex ante}. % Yet contrary to expectations that liberal status quo ex ante would be associated with appellant success, it actually made it less likely that appellants would succeed. % Put differently, liberal rulings in the High Court or Court of Appeal are surprisingly robust to decisions of the Lords. % It is important not to over-interpret this finding: it does not mean that the House of Lords is more `liberal' than the Court of Appeal or High Court; still less does it mean that the House of Lords is liberal in some absolute sense. % Only if judges decide exclusively based on their personal preferences can we interpret it in that fashion. % Indeed, we might even interpret it to give succour to Griffith's interpretation: judges in the Court of Appeal/High Court are so conservative that the only liberal rulings which get through are so well-prepared and based in law that they survive considerable levels of scrutiny on appeal. % \section{Conclusion} \label{sec:conclusion} Thus far I have investigated the factors determining the success of appeals to the House of Lords between 1969 and 2003, and in particular whether the relative status of litigants affects the outcome. % I have suggested that there are no significant effects when considering individuals or associations versus associations or businesses, but that governmental actors do enjoy significant advantages in litigation. % Litigants who employ more experienced counsel also enjoy an advantage. % Contrary to expectations, Law Lords are less likely to allow appeals against liberal decisions of lower courts. % When compared to the body of research on the US Supreme Court, these findings ought to give succour to those involved in the various British legal systems. % Certain kinds of status advantage found in the US are not present in the House of Lords, and thus there are grounds for believing that justice is impartial between, say, individuals and businesses. % Those kinds of status advantage which are seen -- specifically, the advantage enjoyed by governments over all other types of litigants -- are more innocuous, since they likely derive from the rules of the game, rather than the resources litigants bring to play. % At the same time, these findings are limited to a particular court at the apex of the British judiciary. % For litigants to arrive at the House of Lords, they must first litigate at lower levels. % This may be extremely costly. % Impartiality at the apex of the legal system may be of little worth if it is not also found at lower levels. % The progressive curtailment of both the eligibility criteria for legal aid and the overall legal aid budget \citep{HR09} alluded to in the beginning of this article, are factors which might affect the relative status advantage of litigants at lower levels, but which cannot be directly tested with the data available here.\footnote{~Of course, if counsel representing litigants in receipt of legal aid were systematically less experienced than counsel representing all other litigants, then there would be an indirect effect. } % Nevertheless, the findings presented in this article place the onus back on those who would argue that haves enjoy a status advantage in British courts. % \newpage \bibliographystyle{apalike} \bibliography{article} \newpage \appendix \section{Descriptive statistics} \renewcommand{\tabcolsep}{1mm} \footnotesize <>= #tmp<-uk[,c("complexity","rel.counsel","rel.fidu","rel.wins","rel.experience")] #names(tmp)<-c("Complexity","Relative counsel","Relative intervenors","Relative win rate","Relative experience") #tableContinuous(tmp,longtable=FALSE) @ <>= corMatrix <- function(df, digits=2) { ## taken from Felix Schönbrodt: http://www.nicebread.de/blog/index.php dims <- ncol(df) dnames <- colnames(df) ps <- matrix(rep(0,dims ^ 2), nrow=dims, ncol=dims, dimnames = list(dnames, dnames)) # calculate p-values of correlations for (i in 1:ncol(df)) { for (j in 1:ncol(df)) ps[i,j] <- cor.test(df[,i], df[,j], use="p")$p.value } # calculate zero-order-correlations co <- round(cor(df, use = "p"), digits) for (i in 1:ncol(co)) { for (j in 1:ncol(df)) { if (ps[i,j]<0.001) co[i,j] <- paste(co[i,j], "***") else if (ps[i,j]<0.01) co[i,j] <- paste(co[i,j], "**") else if (ps[i,j]<0.05) co[i,j] <- paste(co[i,j], "*") else if (ps[i,j]<0.1) co[i,j] <- paste(co[i,j], "†") if (i==j) co[i,j] = 1 } } ret <- list(cor = co, annotation = "***: p < 0.001; **: p < 0.01; *: p < 0.05; †: p < 0.1") return(ret) } foo<-corMatrix(uk[,grep(":|sourcect|area|decade",attributes(my.mod1$terms)$term.labels,invert=TRUE,value=TRUE)]) foo$cor[upper.tri(foo$cor)]<-"" addtorow <- list() addtorow$pos <- list() addtorow$pos[[1]] <- c(nrow(foo$cor)) addtorow$command <- c(paste("\\hline \n &\\multicolumn{",ncol(foo$cor)-1,"}{l}{",foo$annotation,"} \\\\",sep="")) print(xtable(foo$cor,align=c("l",rep("p{.6cm}",ncol(foo$cor)))), add.to.row=addtorow, hline.after=c(-1,0), size="tiny", #floating.environment="sidewaystable", sanitize.colnames.function=function(x){ paste("\\begin{sideways}",x,"\\end{sideways}",sep="") } ) @ \end{document}