The Costs of Ordinary Litigation

This Article was prepared for the National Conference on the Lawyer's Changing Role in Resolving Disputes, October 1982, at Harvard Law School. Support for this Article and funding for the Conference were provided by the American Bar Association and its Section on Litigation, Charles E. Culpepper Foundation (Harvard Law School Program in Dispute Settlement), National Institute for Dispute Resolution, Control Data Corporation, General Motors Corporation and the Gillette Company. A revised version of this Article will appear in the volume of conference papers and comments to be published by Law and Business, Inc., Harcourt Brace Jovanovich, in 1984. Permission to publish this Article is gratefully acknowledged.

This Article reports some of the principal findings of the Civil Litigation Research Project (CLRP). CLRP was funded under United States Department of Justice Contract JAOIA-79-0040 and National Institute for Justice Contract J- LEAA-003-82 with supplemental support from the University of Wisconsin Law and Graduate Schools.

The results of this study are set forth in D. Trubek, W. Felstiner, J. Grossman, H. Kritzer & A. Sarat, Civil Litigation Research Project: Final Report (Mar. 1983) (University of Wisconsin Law School). The entire report consists of a summary and three volumes. Volume I is entitled Studying the Civil Litigation Process: The CLRP Experience; Volume II is entitled Civil Litigation as the Investment of Lawyer Time, and Volume III is entitled Other Studies of Civil Litigation and Dispute Processing. This Article summarizes aspects of the report, principally those in Volume II. The views contained herein are those of the authors and do not necessarily represent the official views or policy of the United States Department of Justice. The authors acknowledge the contribution of the entire CLRP staff in all phases of the work reported on, but wish especially to thank Elizabeth McNichol for her assistance in data analysis, Jeanette Holz for typing, and Richard Miller, Stephen McDougal, Robert Sikorski, Kristin Bumiller, Laura Guy and James McLaughlin for research assistance.

[CLRP resulted in extensive publications by its various participants; click here for a bibliography of those publications.]

INTRODUCTION

It is widely believed that the costs of litigation are rising and that these costs are an important public problem. In professional and policy discourse, the "costs"' discussion focuses on the amount of money clients must spend to use courts for processing disputes. Rising costs are seen as a barrier to some and a problem for all litigants.⁽¹⁾ The debate over "costs"' merges with a broader range of *75 issues about the role of courts in society and problems created by too much litigation or litigation about matters best handled outside the courts.⁽²⁾

In this Article we seek to contribute to the debate over "costs"' by analyzing direct expenditures of time and money on processing disputes through litigation. We believe this analysis will help clarify the debate on litigation costs in particular and the role of courts in general. We approach this task by reporting on the result of a study conducted in five parts of the United States. In each of five federal judicial districts, we studied randomly sampled civil cases from the federal district court and at least one state court. We also surveyed the general population in these districts. These data, including over 1,600 cases and many thousands of interviews, form part of a larger data base collected by the Civil Litigation Research Project (CLRP).

This Article begins with a description of the central theme of the empirical investigation--litigation as an investment--and discusses the questions that arise as a result of viewing litigation as an investment.⁽³⁾ Next, this Article provides some descriptive information about the sample of cases we have used in the analysis and the lawyers who responded to our survey.⁽⁴⁾ Then we explain the model we used to estimate the amount of resources invested inlitigation.⁽⁵⁾ Next, we describe how much time lawyers spend on the typical civil lawsuit, what they spend their time doing, and explain what influences a lawyer to spend more or less time on a case.⁽⁶⁾ Finally, we assess whether clients recover as much as they spend on lawsuits.⁽⁷⁾

I. THE CENTRAL THEME--LITIGATION AS AN INVESTMENT PROCESS

Litigation can be examined in many ways.⁽⁸⁾ We have chosen to conceptualize the process as the investment of scarce resources to achieve a future result. The resources to be invested include time and money; however, as it is frequently possible to translate the value of time expended on litigation into monetary terms, these may come to the same thing. The results to be achieved include recovering money (plaintiffs) or avoiding paying money (defendants), stopping something from happening or causing some act to be carried out. In theory, all results could probably be given a monetary value, but in practice this proves extremely difficult. In the empirical investigation of litigation as an investment, we have therefore distinguished between expenditures of time and of money, and between monetary and non-monetary results.

We claim no originality in our decision to conceptualize litigation as an investment. This approach has been widely used by economists, from whom we have adapted the conceptual framework that oriented our data collection and analysis.⁽⁹⁾ We have, *77 however, translated this approach into specific hypotheses that can be tested empirically with our data. In this Article we focus on two questions about the litigation investment:

A. Investment Levels

We begin by exploring the level of resources invested in litigating a case. We expected to find significant variation among cases in the amount of resources (time or money) committed. We hypothesized that the value of the expected result will significantly affect investment levels--ceteris paribus, the higher the "stakes"' (positive or negative) the more resources a party will devote to the case. But if the stakes are a major factor influencing any party's investment decisions, they are not the only one. Perhaps the most obvious additional factor is the actual or expected investment by the other side. Litigation is an interactive process and one side's investment is likely to be influenced by the other side's actual or expected expenditure. Further, litigation investments, like most investments, occur under conditions of uncertainty; one cannot be sure of the outcome, the relationship between investment and result, or the other side's expenditures. Risk preferences, the willingness of a party to risk resources for uncertain outcomes, will thus influence litigation investment levels.⁽¹⁰⁾

The schematic diagram in Figure 1 explains the model set forth above. In this diagram, A makes an initial "investment decision"' by considering (i) the expected return in light of A's risk preferences, and (ii) B's initial investment decision, which results from the same factors as A's. The actual investments of both parties influence returns. The model in Figure 1 is highly abstracted and fails to take account of the actual complexity of real cases. And because the model incorporates a simultaneous interaction process, it cannot be directly tested unless very stringent conditions are met.⁽¹¹⁾

FIGURE 1
THE INVESTMENT DECISION MODEL

To examine the litigation data we collected, we developed an empirical model that is both more comprehensive, because it includes many variables not incorporated in the simplified scheme of Figure 1, and more suitable to statistical analysis, because it approaches the problem of interactive investments in an indirect way which we will describe below. We use this model to explain variation in the time lawyers spend on civil lawsuits.⁽¹²⁾

B. Does Litigation Pay? Assessing Costs and Benefits

Viewing litigation as an investment naturally raises a series of questions: Do resources invested in litigation yield acceptable benefits? How large are the gains? Do the benefits from litigation exceed the costs? Does litigation yield more than other forms of dispute processing?

At the theoretical level, these questions are complex. In the first place, from whose viewpoint should we assess the "yield"' from litigation investments? We have to look separately at whether litigation pays for clients, for lawyers, and for society as a whole. As Earl Johnson has suggested, litigation investments that may be highly lucrative for lawyers may not be optimal for clients, and vice versa.⁽¹³⁾ Moreover, even if both lawyers and clients gain from litigation, it does not follow that litigation is a cost-effective process for society. The simple fact that taxpayers rather than litigants *79 pay the cost of operating the courts shows why calculations of social and private costs must diverge.

A second issue is whether one can analyze litigation investments in isolation. If one says "litigation pays"' (or doesn't pay), the question arises: Compared to what? CLRP was designed, in part, to permit comparative analysis of the costs of litigation and alternative approaches to resolution of disputes, and to determine if clients did "better,"' objectively and subjectively, in litigation or in other processes available for dispute resolution. For this reason, we collected data on cases in "alternative institutions,"' such as arbitration, administrative agencies and mediation programs, and on disputes that were "resolved"' by negotiation without third party intervention.⁽¹⁴⁾

A third issue is whether monetary indicators of gains and costs are, by themselves, adequate to assess whether litigation "pays."' A client, for example, might secure a substantial net recovery after paying lawyer's fees and other costs, and still be considered in a worse position if one took into account the non-monetary costs of the litigation experience.⁽¹⁵⁾ In theory, the "investment"' approach does not inherently limit analysis to monetary costs and benefits. But methodological problems, complex as they are when one limits the focus to monetary factors, become formidable when one seeks to go beyond this dimension.

A complete analysis of the costs of litigation would examine private and social costs, study the relative cost of litigation and other dispute processing modes, and in some way incorporate non-monetary costs and benefits. We are not, however, able to deal with all these facets of the problem. Although we have collected data from alternative institutions and "bilateral disputes,"' we have not yet analyzed them, and can only report findings on litigation. Further, for methodological reasons, we have restricted our analysis to monetary (or easily translated to monetary terms) costs and benefits of litigation. Finally, we focus on whether litigation, examined in isolation, "pays off"' in monetary terms for plaintiffs and defendants. We discuss some of the monetary costs of litigation not borne by litigants, using the limited data available, *80 but do not reach any overall conclusions from a social, as opposed to a private, point of view.

Before turning to the analysis, we present some descriptive information on the sample of cases we have used and the lawyers who responded to our survey. These data provide a picture of important dimensions of the world of civil litigation in the United States, a picture which may help correct some distortions in the way we think about litigation in general and its costs in particular.

II. THE WORLD OF LITIGATION--THE CASES AND THE LAWYERS

A. Source of Data

Our analysis of litigation investment is based primarily on a survey of lawyers throughout the country. We asked lawyers participating in selected cases about the case, its costs, and their background. For the principal analysis reported here, we have added information about the cases from court records. We also draw, to a lesser degree, on data from a parallel survey of the clients and a general household survey.

These sources are part of an even larger data base collected by CLRP under a contract from the United States Department of Justice. The primary source of data for CLRP was a sample of civil cases from state and federal courts and from "alternative institutions"' like commercial arbitration. We supplemented this case sample by a set of "bilateral disputes"' which never reached third parties. The case sample was drawn on a random basis in five federal judicial districts: Eastern Wisconsin, Eastern Pennsylvania, Central California, South Carolina and New Mexico. In each district we sampled terminated cases from the records of the federal district court and one or more representative state courts, and from a series of "alternative"' institutions. We coded the data from these institutional records and then sought to interview clients and lawyers in each case. In addition, we conducted surveys of households and private organizations in each district to locate "bilateral disputes."' Data from these surveys were integrated into our overall base, and some are reported here.⁽¹⁶⁾

We designed a sampling scheme to limit our data to what we call the "middle range"' of civil disputes. We excluded disputes in which the initial claim was under $1,000, and dropped a few *81 "megacases"' from the sample.⁽¹⁷⁾ As a result, the sample excludes small claims and slightly underweighs the very large civil lawsuit.

We selected 1,649 civil lawsuits from court records; the court sample is about evenly divided between state and federal cases. We then tried to interview parties and attorneys of record in all these cases. Our most successful survey was of the attorneys. We interviewed 1,812 lawyers from these cases.⁽¹⁸⁾ In the analysis which follows, we rely primarily on a subset of the responses--the 1,387 attorneys who took an hour or more to answer our full questionnaire.

In considering the results we report, the reader must appreciate the strengths and weaknesses of our data. While our original selection of cases from five districts was designed to be as representative of civil cases in the United States as possible, no sample limited to only five of the ninety-five federal judicial districts can be fully representative. We selected the cases at random from all cases that had terminated in 1978. We excluded certain types of civil cases, such as those which did not involve "disputes,"' as we defined this term.⁽¹⁹⁾ For the types of cases included, however, our sample is basically representative of civil litigation in each district.⁽²⁰⁾ By including an equal number of federal and state cases in the sample, we substantially oversampled federal cases, since the latter probably comprise less than 3% of all civil cases filed in courts of general jurisdiction in the United States.⁽²¹⁾ For this reason, we usually report federal and state court data separately. *82 Readers concerned with a statistical profile of all middle range civil cases in the United States should treat our state data as most representative. The reader should also bear in mind that we have not included cases from small claims courts or other specialized courts; this is a sample of cases that involve genuine disputes in state courts of general jurisdiction and the federal courts.

The main source for the data reported here was telephone interviews with attorneys. Conducted in 1981, these interviews lasted about an hour and covered all aspects of the case. We contacted the attorneys in advance and asked them to review their records prior to the interview. In many cases, since the events in question might have occurred some years before, it is inevitable that there were problems with selective memory ("recall bias"'). Nonetheless, our data base is the best (indeed, probably the only) source of information on litigation costs and their effects currently available. The reader, aware of the limits of our data, will have to assess the plausibility of our conclusions and the policy results our findings suggest.

B. Extraordinary and Ordinary Litigation

One advantage of our data base is that it allows us to focus on what might be called the "typical"' civil lawsuit. Much of the discussion of litigation in general and the "costs of litigation"' in particular deals with the extreme ends of the range of litigation phenomena, not with what statisticians would call the central tendencies.⁽²²⁾ Three themes dominate the literature. First, the public hears a lot about the very large, complex case, usually involving major businesses, in which legal titans clash in forests thick with briefs, motions, discovery and endless trials.⁽²³⁾ Second, the unusual or problematic subjects for litigation are well publicized, with commentators questioning if the courts are not unwisely intruding into complex public policy issues⁽²⁴⁾ or privileged spheres of private life.⁽²⁵⁾ Finally, researchers have paid a good deal of attention to the minor dispute, where the conflict is too small to justify the investment of lawyer time and for which the existing court system *83 may be either too costly or ineffective or both.⁽²⁶⁾

Large cases and small claims raise important policy issues. But the large case is a rare phenomenon in our civil courts of general jurisdiction, and small claims do not reach these courts. Before we as ess the "costs"' of litigation in the United States or discuss the need for reform, it is important to examine the ordinary and typical cases. The CLRP data represent the bulk of what is going on in the courts. The "middle range"' civil disputes we studied mostly involve routine legal business; many are standard tort and contract suits so familiar to the litigating bar and the bench, yet rarely discussed in the media or by proponents of reform.⁽²⁷⁾

These data portray an image which varies from the picture of litigation projected in much popular and some professional discussions of the dispute resolution "problem."' A casual reader of the literature on courts and court reform in the United States might conclude that Americans litigate with great frequency, and that the typical lawsuit is complex, costly, and time-consuming. Further, the reader might think that litigation involves the unconstrained exercise of adversary skills by legions of lawyers who devote many hours to such "lawyerly"' activities as preparing motions, conducting discovery, writing briefs and conducting trials. All this effort might appear to impose a vast burden on judges who must rule on numerous motions, supervise extended discovery, conduct lengthy trials and render difficult judgments. Finally, it might seem that clients pay an exorbitant cost for services rendered in litigation and that fees eat up a substantial portion of recoveries.⁽²⁸⁾ While all this occurs, and probably occurs with some frequency, the typical case, as we observed it, is very different.

To reduce our data to a single, composite case that might be considered typical is difficult. But if we were to do this, we would describe the "'typical"' case as follows: first, the very fact that a dispute has reached the court and not been settled without litigation makes it unusual. Viewed against the baseline of potential lawsuits, litigation is not frequent, since for every dispute in the court records there are nine others that never even reach the filing *84 stage.⁽²⁹⁾ Second, the cases in courts of general jurisdiction are modest. The parties are usually fighting over money, and the amounts at stake are $10,000 or less. Third, the typical case is procedurally simple and will be settled voluntarily without a verdict or judgment on the merits. This case will involve some pretrial activity, but no trial. Each side's lawyer spends about thirty hours on the case, mostly gathering facts and negotiating a settlement. Judicial involvement, either ruling on motions or rendering judgment, will be rare. The typical case is a "paying"' proposition for the parties. The average plaintiff will recover some portion of the amount claimed, and the amount recovered will significantly exceed the money and the value of time spent on the case. Even the defendants can be said to have "gained"' from the litigation, at least in the sense that their litigation expenditures are less than the amount by which plaintiff's claim was reduced during litigation.

This composite picture of ordinary litigation helps correct biases in discussions which focus on the extraordinary lawsuit or the very small claim. But these findings do not mean we question the need for reform in the civil justice system; indeed, quite the contrary is true. First, even if ordinary litigation is less problematic than the extraordinary case, cost-related problems still exist in this area, especially in the smaller or middle range claims where costs may exceed benefits. Second, even if the typical case is less problematic than the statistically unusual ones, at least six million civil cases are annually filed in the United States,⁽³⁰⁾ so that, even if only 5% are ''extraordinary,"' these cases could consume substantial resources. Third, while our data do not deal with the small claim, what data we have confirm the view that litigation in conventional courts is not a cost-effective way to deal with many minor disputes.⁽³¹⁾ Fourth, we show that litigation "'pays"' in the sense of yielding net monetary benefits. But we cannot say if these gains are wiped out by negative non-monetary features of the litigation experience. Moreover, we are not saying litigation clients do as well as they might if the system were changed. In the first place, our data provide some support for the view that lawyers appropriate some of the gains that might otherwise accrue to clients under more competitive market conditions.⁽³²⁾ Second, we cannot say *85 whether or not the clients of the lawyers we studied would have been ever better off if their disputes had been handled in some other way.

C. Some Dimensions of Litigation

In this section we describe five "dimensions"' of litigation: frequency, stakes, activities, lawyer time and costs. Within the limits of our sample of middle range civil disputes from five parts of the United States, we seek to give some idea of how often people in disputes actually use the courts, how much money is involved in those cases in which the basic dispute can be treated as a conflict over money, what actually occurs once a lawsuit is filed, what lawyers spend their time doing, and how much money is spent by litigants.

1. The Frequency of Litigation

Is litigation a frequent response to the disputes that arise in society? Some would say yes, arguing that Americans are unusually prone to resort to the courts when problems arise.⁽³³⁾ Marc Galanter has analyzed this view, which he calls "hyperlexology,"' and demonstrates that the view that Americans are unusually litigious is based more on myth than careful analysis of the data.⁽³⁴⁾ Our contribution to this debate is based on the survey we conducted of over 5,000 households in the five judicial districts studied. Our data cannot answer the question of whether we are litigating "too much,"' but it can suggest how frequently litigation occurs.⁽³⁵⁾

Any empirical discussion of the frequency of litigation must employ a baseline--some measure of the number of opportunities to use the courts against which actual filing rates can be compared.⁽³⁶⁾ *86 The litigation baseline is the number of transactions of a particular type which might ultimately lead to lawsuits. In studying medical malpractice litigation, for example, noting an increase in the number of cases filed is inadequate. This increase must be compared to rates of professional contacts which might generate lawsuits. Thus, one might employ for a baseline the number of visits to doctors, the number which result in injury or the number which result in perceived grievances. The frequency of malpractice litigation would then be a percentage of the number of visits, injuries or grievances.

Our baseline was the incidence of disputes which occurred in eight selected general areas--tort, consumer, debt, discrimination, property, government, post-divorce and landlord-tenant. A "dispute,"' for our purposes, occurs when an individual perceives a grievance, seeks redress and is at least initially rebuffed by the other party. To determine litigation frequency, we compared the number of "disputes"' with the number of complaints filed in federal or state courts by disputants. (Note that by using filing as our measure of litigation, rather than some index of substantial court activity, we are overstating the rate at which disputes lead to judicial intervention.) We measured the incidence of both disputes and litigation by randomly sampling the general population in each of five geographic areas. This study provides a rough estimate of the frequency of litigation involving individuals in the United States.⁽³⁷⁾

Overall, we found that 71.8% of individuals with grievances complained to the offending party, and that a dispute arose in 63.0% of these situations. Of these disputes, 11.2% resulted in a court filing. Figure 2 shows the overall disputing pyramid that emerges.

These figures show that lawsuits are filed in just over 10% of the disputes involving individuals where $1,000 or more is at issue. Approximately 90% of the cases were settled or abandoned without a court filing. When one realizes that in many lawsuits little or nothing occurs except filing the complaint, an 11.2% litigation rate does not seem particularly high compared to the potential baseline. Of course, in a country as large as the United States, *87 even at such a rate there will be numerous lawsuits which will involve substantial judicial activity. Nevertheless, it is clear that litigation, even in the limited sense of starting a lawsuit, is by no means the most common response to disputes. As Table 1 shows, the lowest litigation rate is in the consumer area, a field in which the amount at stake is often comparatively small. The post- divorce disputes (e.g., adjustments in custody and support) have the highest litigation rate. This rate in part reflects the fact that many times even consensual arrangements must be ratified by the court.⁽³⁸⁾

We now turn to describing what goes on in the ordinary lawsuit once it is filed. These data are from two sources: the lawyer survey and the court records.

2. Stakes

We first sought to estimate the amount of money the litigants thought was in dispute during the life of the lawsuit. To measure this, we used the lawyers' estimate of how much their client would *88 have accepted or paid to settle the case. By using this variable, we tried to capture as accurately as possible the amount that the parties considered in controversy as they made decisions about investments in litigation. We preferred settlement amount to actual economic loss or potential exposure since we believe that lawyers, at least, in calculating case value in practice discount both loss and exposure for risk and transaction costs. We call this measure the "stakes"' in the case, and report the highest figure if the attorney's view changed during the case.⁽³⁹⁾ In some of our cases the *89 attorneys could not give a monetary estimate of the stakes, but we did get monetary figures in 859 of our attorney interviews.

Figure 3 shows the distribution of stakes for the cases in our sample. Overall, 56% of the cases involved $10,000 or less. Only 12% of our cases involved stakes of $50,000 or more. Not surprisingly, state court cases tend to be "smaller"' than cases filed in the federal courts. For state cases the median stakes is $4,500; for federal cases, $15,000. Most often courts do not manage cases in which vast amounts of money are involved. While the prospect of transferring 5, 10, 15, or 20 thousand dollars is by no means trivial, it does not convey an image of a court system overwhelmed with blockbuster, megacases.

3. Activities

What happens in ordinary litigation? There is a popular image that litigation involves extensive pretrial activity and protracted trials.⁽⁴⁰⁾ Our data suggest the contrary. Trials are rare, pretrial activity modest, and most cases terminate through settlement negotiations.

Less than 8% of the cases in our sample went to trial. In another 22.5%, the judge dismissed the complaint or rendered judgment on the merits without a trial.⁽⁴¹⁾ The most frequent mode of termination is voluntary agreement between the parties, which occured in over 50% of the cases. Our data suggest civil judges and juries provide final, authoritative third party dispute processing in less than a third of the cases. More often, the courts serve as the background for bargaining between the parties. Bargaining occurs "in the shadow of the law,"' but is conducted primarily, if not exclusively, by the parties and their lawyers.⁽⁴²⁾

Pretrial activity is much more common than trials, but modest nonetheless. Discovery, for example, is widely thought to be a cause of delay and spiraling costs.⁽⁴³⁾ Our data, however, suggest *90 that relatively little discovery occurs in the ordinary lawsuit. We found no evidence of discovery in over half our cases. Rarely did the records reveal more than five separate discovery events. While our data are limited to the court records, these findings confirm the conclusion of an earlier study that even in federal courts discovery is used intensively only in a small fraction of civil lawsuits.⁽⁴⁴⁾

4. Lawyer Time

The third dimension of litigated cases we measured was the time lawyers spent on cases and the way they allocated that time. Once again, these data demonstrate the differences between ordinary and extraordinary litigation. Our data show that a typical case involves relatively few lawyer hours and that attorneys spend almost half of this time in conferences with clients, factual investigation other than discovery, and settlement negotiation.

We asked all lawyers to estimate the number of hours they and other lawyers in their firm spent working on the cases which we were studying. The number of hours spent per case by each attorney varied from eight or fewer (our lowest category) to 2,200 hours. In the median case the lawyer spent 30.4 hours, while the mean for all cases (total hours in all cases divided by the number of cases) is 72.9. The distribution is set forth in Table 2.⁽⁴⁵⁾

*91 Sixty percent of the lawyers (or firms) spent less than one person- week on the cases we asked them about. In 13% of the cases they devoted eight or fewer hours to the case.

In addition to asking for the total number of hours lawyers spent, we sought to determine how time was allocated among a series of different litigation activities. We asked the lawyers in our sample to indicate the percentage of time they spent in that case on each of nine separate activities, such as pleadings, discovery and settlement discussions. We then determined the average percentage for each activity for all the lawyers in our sample. Table 3 sets forth these percentages.

The lawyers in our sample on average spent a relatively small portion of their time on legal research and formal procedural matters. They devoted less than 10% of the time in the case to trials or hearings, and more time to settlement negotiation than to legal research.

5. Monetary Costs

We secured information from clients and lawyers about the expenditures of time and money the clients make on litigation. Clients reported that the bulk of their expenditures were payments made to lawyers to cover fees and expenses. Payments to lawyers constituted 99% of out-of-pocket litigation expenditures for individual clients and 98% for organizations. Even when we add in the monetary value of the time clients spend on cases, fees and related expenses equal 88% of the median individual's costs. (For organizations, the comparable figure is 72%.) We found that expenses,⁽⁴⁶⁾ *92 as distinguished from fees, make up a very small percentage of the total bill for most lawyers, and probably are closely correlated to the fees. We chose, therefore, to concentrate description and analysis on legal fees as a reasonable proxy for total monetary cost.

The data in Table 4 reveal that legal fees in the world of ordinary litigation are modest. In almost half the cases we studied, the fees were under $1,000. In only 8% of the cases were fees over $10,000. Not surprisingly, fees are lower in state cases than in the federal courts. Twenty-five percent of the lawyers in our federal sample reported fees over $5,000, while lawyers in only 6% of the state cases received this level of compensation.

6. The Litigators

In addition to examining these dimensions of litigation activity, we sought information on the litigators. We wanted to get some idea of the settings in which they practice, their experience, specialization, and income. We report data only on private attorneys, house counsel and legal services lawyers. Government attorneys *93 were surveyed separately, but this data has not yet been analyzed.

Seventy-eight percent of the lawyers practice with firms (2 or more lawyers); the modal size firm is 5-9 lawyers. The distribution by firm size is shown in Table 5. In addition, 17% of the sample are solo practitioners, 3% are house counsel and 2% work for a legal services or legal aid program.

Some lawyers in our sample had practiced less than one year, and some had been at the bar over fifty years. Thirty-four percent had practiced less than five years, and well over half (58%) had been in practice ten years or less.

Our indicators point to a young but specialized litigating bar. The lawyers surveyed spent most of their time on litigation: the average was 75%. Twenty percent (274) of the lawyers devoted 95% or more of their time to litigation and only 2% reported spending less than 10% of their professional time litigating. The lawyers also tend to specialize in one area of law. The average lawyer reported spending half her time on the type of case we were interviewing her about. Over half the lawyers reported having already handled at least 200 such cases before. When asked to evaluate their own expertise in the field in question, 78% said they were "expert"' or "somewhat expert."D'

Lawyers in the sample were asked about their average annual income from practicing law for the three years preceding the 1980 interview. The median lawyer's income from practice was $45,000. Most lawyers (60%) earned between $25,000 and $75,000. Only 5% reported incomes above $100,000 and only 3% made $15,000 or less.

What explains the amount of resources invested in lawsuits? We focused on the number of hours the lawyer spends, rather than the dollars the client pays. We had three reasons for this choice. First, we accepted Johnson's⁽⁴⁷⁾ theory that lawyers are principal decision makers in litigation investment decisions, and one of the decisions they make is how much time to spend on the case. Second, the patterns of our data dictated a focus on hours. Seventy-one percent of all our plaintiff lawyers and 41% of all lawyers surveyed were paid on a contingent fee basis. Since fees in these cases were determined exclusively by the amount recovered, they were not a good measure of the resource investment decisions we wanted to study. Finally, in a national study like this, we can more easily compare hours than fees, as fees vary on a regional basis.

A. The Model of Lawyer Time Investment

The number of hours spent per case ranged from less than eight to over 2,000. This figure served as our dependent variable--the factor to be explained. To explain hours, we constructed a model of the investment process. This model consists of a number of independent variables, which we thought should increase or decrease the number of hours a lawyer will spend on any case, and can be seen as a series of related hypotheses about what will influence investments. We tested these hypotheses against our data by a statistical technique called multiple regression analysis. By using this method, we can both determine how well the whole model (all the independent variables) explains variation in hours, and assess the relative importance of specific variables.

1. Factors and Variables

The dependent variable in the model is the number of hours the lawyers reported they or their firm spent on the case in question. To explain variation in hours, we selected 29 independent variables which we could measure and which we had reason to believe would explain variation in time spent on the case. For exposition and analysis, we have grouped these variables into five major factors. These factors are: (a) characteristics of the "case"' itself, such as stakes, complexity and duration; (b) the procedural events which occur; (c) the participants; (d) the goals of the participants, including the lawyers; and (e) certain strategic choices *95 made in case processing and management. The factors and the detailed indicators used to measure them are explained below.

a. Case characteristics. We reasoned that the amount of money (or monetary equivalent) involved in the case, its overall complexity, and the length of time it took to process would have a significant impact on hours spent. These factors are related more to the nature of the "case"' itself than to choices made in processing it, such as procedural events and management decisions, or to characteristics of the participants in it, such as client motivation, lawyer's goals, abilities and background.

We considered stakes to be one of the most important determinants of time investment. If litigation is the process of investing time to secure an expected "return,"' "stakes"' is the measure, for plaintiffs, of what they could realistically gain by litigation expenditures, and for defendants, of what they expected they might lose from an adverse outcome. We reasoned that the higher the stakes, the more time parties would invest in the case.

When we began our work we expected to find that stakes would be the primary factor that would determine the amount of time spent on cases. At the same time, we recognized that factors other than stakes were likely to influence litigation investment decisions. We conceived of these factors as modifying an investment of time or money that would primarily be determined by stakes.⁽⁴⁸⁾ Some of these modifying factors, like those which measured lawyer expertise and planning, would reduce the time needed because they increased the productivity of the service. Others, however, like the legal complexity of the case, would increase the amount of time. But as Figure 4 suggests, in our original conception the modifying factors would increase or reduce an investment amount primarily determined by estimates of stakes.

Early analysis of our data made clear that the stakes model, at least in the simple form that we had envisioned it, was an inaccurate picture of the litigation investment process. In constructing the model shown in Figure 4, we thought of stakes as "driving"' *96 the investment process. We quickly realized that stakes do not drive the investment process, but merely put a "'cap"' on the level of investment that will occur. This can be seen in the information displayed in Figure 5, a "scatterplot"' of lawyers' hours and stakes. The vertical axis of the figures shows the number of hours spent on a case and the horizontal axis shows the stakes. Each point represents the hours- stakes combination for one case in our sample of lawyers. (Figure 5 includes all cases in which stakes were less than $100,000 and hours were less than 200.) While the figure shows a general rise in the level of investment as the stakes increase, the rise primarily occurs in the upper limits of time spent on the case; that is, the range of investment level increases as stakes rise, but there are still many cases with high stakes for which the level of investment is very low. The stakes set the upper limit on the hours a lawyer will spend, but other variables are more important in determining the actual hours that are invested.⁽⁴⁹⁾

Some cases involve simple and clear-cut issues of law and easy questions of proof. In others, the law or facts may be complicated or unclear. The more complex the law involved, or the more difficult the problems of proof, the more time it should take *97 to conduct the litigation. Our measure of complexity was based on the lawyer's evaluation.

Much of the discussion of the "costs"' of litigation is in terms of "delay."' We hypothesized that the length of time a case takes from filing to termination would have an independent effect on the number of hours worked by lawyers. If, for example, the case stretched over a long period of time, the lawyer would periodically refresh her memory of the case, or "find"' things to do. We measured duration simply as the number of days elapsed from filing the case to its termination, whether by settlement, adjudication or abandonment.

b. Events in the case. A lawyer's time will be influenced by the "events"' that occur in a case. Is there substantial pretrial activity? Does the case go to trial? By using the cluster of "events"' variables, we sought to measure the relationship between the presence of several "events"' and the number of hours attorneys spend on cases.⁽⁵⁰⁾ The cluster includes the number of (i) pleadings, (ii) non-discovery motions and briefs and (iii) discovery related events, including depositions, interrogatories, requests for admissions, medical exams and the like, plus discovery related motions. In addition, we examined whether there was a trial and whether there were settlement negotiations.

c. Nature of participants. The participants in the case are the lawyers and the clients. We classified clients as individuals or organizations because the literature⁽⁵¹⁾ suggested that organizations would devote more resources to litigation than individuals.

With respect to lawyers, our classification was more complex. We created six separate indicators designed to measure variation in lawyer characteristics. Specialization measures whether the case in our sample fell within the lawyer's specialty or not. Law school performance is the lawyer's self-report of performance as a law student. Amount of general experience is the number of years the lawyer had been practicing law. Litigation experience is the proportion of the lawyer's time devoted to litigation. Personal capacity is a measure of the lawyer's feelings of efficacy based on a standard measure. Lastly, craftsmanship is the likelihood (self-reported) of spending extra time to make marginal improvements on legal documents; the more likely this was, the higher the "craftsmanship"' score.

*98 We expected that the first five variables, which measure ability and self-confidence, would be inversely related to the amount of time lawyers spend on cases. We reasoned that a more experienced, specialized, and confident lawyer would not have to spend as much time on a case as would an attorney who was newer to the field of law, to the courtroom, or to practice in general. We expected the craftsmanship variable to work the other way; that is, lawyers who were more oriented toward "craftsmanship"' would spend more time on their cases, other things being equal.

d. Participant goals. We measured participant goals for both lawyers and clients by using data from the lawyers. We asked lawyers what they thought their client's goals were in the case. The "goals"' variable, in a sense, modifies the "stakes"' variable. We asked lawyers if they thought their clients were seeking as much money as possible, or just a "fair amount"' (for defendants, to pay the least or pay a fair amount). We expected to find that the lawyer whose clients wanted to get the most (or pay the least) to put in more time on a case than the lawyer in an otherwise identical case whose client only wanted "fairness."' We assumed that those clients (about 24% of our respondents' clients) who wanted to neither "get most/pay least"' nor "get fair/pay fair"' were primarily concerned with goals other than money.⁽⁵²⁾

To get information on lawyer goals, we asked our respondents why they had taken the case in question. We reasoned that lawyers may have motives independent of their clients' which would affect the amount of time they spend on cases. From the answers, we constructed five lawyer goal variables designed to measure the predominance of different factors in the lawyer's decision to take the case. These are:

While we felt variations in these goals were likely to affect hours, we did not have strong expectations concerning the nature of all these effects. For example, we thought that the professional visibility and challenge goals might influence a lawyer to spend more hours than the making money goal, but we were not sure what effect "public service"' would have.

e. Processing and management. Since we thought that differences in procedures, judges, administration, etc., between state and federal courts might have an independent effect on the amount of time lawyers would spend, we included a court type (federal or state) variable. We did not have any expectations about the direction of this effect. We thought lawyers might vary in the case management techniques used and this would affect hours. We used three indicators: standard operating procedures, plans, and client control. We thought that the lawyers who developed standard operating procedures (SOPs) for estimating case worth and pretrial activity would be able to reduce the number of hours spent on a case, other things being equal. Explicit planning should also increase lawyer efficiency and thus decrease time spent. The variables plans for motions, plans for settlement, and plans for discovery measure if planning occurred or not.

We expected to find that client control and participation would influence hours spent, but the effect would differ for hourly fee lawyers and non-hourly fee lawyers. We measured the client control variable by using data provided by lawyers on (1) reporting procedures to the client and (2) the client's participation in key decisions in the case. We felt fee arrangements would affect the incentives of lawyers. Hourly fee lawyers, who can pass their time costs on to the client, would be more likely to spend time than would contingent fee lawyers. Further, following Johnson,⁽⁵³⁾ we thought that it would often be in the client's interest to reduce the hours spent by the hourly fee lawyers and try to increase the time spent by those on contingent fees. For these reasons we expected that a high level of client control for hourly fee lawyers would reduce the number of hours those lawyers worked on a case. In contrast, we expected that for non-hourly fee lawyers (most of whom were paid by contingent fees) high client control would lead to an increase in the number of hours the lawyer would work in the case.⁽⁵⁴⁾

2. Expected Results

Our complete model includes the dependent variable "hours"' and the five major factors we expected would explain variation in hours. We measured these factors by 29 independent variables organized in eight clusters. We have suggested the reasons these variables were included and the nature of the effects we expected. *101 In the process of constructing this model we relied on "'empirical feel"' as well as on existing theory; the theory we had was partial and largely untested. Thus, we were prepared to find--as we did-- that some of our variables had no effect, and others had effects opposite to the ones we anticipated.

Table 7 sets forth the complete model, including all the individual variables. The signs in the table show the expected direction; "zero"' designates variables we thought would have an effect, but for which we could not predict if the effect on hours would be positive or negative. For analytic purposes, we divided some of the factors into "clusters"' of related individual variables, and tested the relative importance of the cluster. Thus Factors I and II were each treated as a cluster, but Factors III-V were each subdivided into two clusters. There are therefore eight clusters, which are italicized in the table.

A key variable--fee arrangements--is excluded from the model as presented. One would expect fee arrangements to influence lawyer hours. Why then did we exclude this variable?

The answer is based on the nature of the data. The economic incentives for hourly fee lawyers, who charged fixed sums per hour whether they win, lose or draw, are very different from those for contingent fee lawyers, who are paid an agreed proportion of the recovery if they win and nothing except expenses if they lose. Because economic incentives appear to differ, theorists argue that in comparable cases the hourly lawyer will spend more time than the contingent fee lawyer.⁽⁵⁵⁾ These considerations would suggest that fee arrangements should be included among the variables which influence hours. But at an early stage we saw that our model "worked"' differently for non-hourly (contingent fee) and hourly lawyers. Early tests showed that, overall, very different factors explain hourly fee lawyer investment than those which govern the contingent fee attorney's decisions. The patterns were so different that we chose to analyze these categories separately.

B. Findings

We tested the model against our data, using techniques which permit us to measure three things:

1. Overall

The measure for overall explanatory power is the R2 statistic, which indicates what percentage of the actual variation in hours is explained by the variables included in our model. The relevant R2 statistics we report are .45 for hourly and .35 for non-hourly lawyers. These results mean that we have succeeded in identifying and measuring factors that account for about half of the differences in hourly lawyer investment, and a third of the differences in non-hourly lawyer time decisions. By social science standards, R2s of .35 to . 45 are quite respectable. The difference between the R2s suggests that we have been more successful in modeling the hourly than the non-hourly lawyer investment process.

Which variables have an effect on hours and which are most important? This information is set forth in Table 8, which contains several key items of information. First, it shows whether the variables have a statistically significant effect on hours. Significance tests measure the degree of confidence one has in inferring that data from a sample reflect actual patterns in the underlying population. Where our results for this analysis fell below the 95% confidence level we show a zero. Second, where the variable has an effect, we show whether the presence of the variable increases (+) or decreases (-) hours spent. Finally, the table shows which clusters of variables were most important; that is, which clusters had the largest overall effect on hours, plus or minus. The table separates hourly and non-hourly lawyers; almost all of the latter are paid on a contingent fee basis.⁽⁵⁶⁾

Several things stand out. The first is the difference between hourly and non- hourly lawyers. Not only are the overall R2 statis ics different, but many more of our variables have a measurable influence on the hourly lawyer's time investment than on the non-hourly lawyer's time, and the relative importance of the several variable clusters is quite different. The second notable feature of the table is that many variables we thought would affect hours do not.⁽⁵⁷⁾ Finally, some variables had an effect, but in the opposite *104 direction from the one we had expected. These variables are indicated by the letter "a." D'

2. Hourly Lawyers

Some findings are not surprising. Case characteristics and events have a significant effect on hours. The higher the stakes, and the more complex the case (as reported by the lawyer), the more hours the case takes. But note that the relative importance of the case characteristics cluster is low (fifth out of eight). This confirms our initial finding that stakes do not "drive"' investments. Moreover, contrary to our expectations, duration does not have a substantial effect on hours.

Events are obviously important; this cluster has the highest relative score. Looking at the individual variables, we find, not surprisingly, that the more motions filed and discovery conducted, the more hours spent. But the trial variable did not have a statistically significant influence on hours. This finding may at first seem to be counter-intuitive. But trials are rare and when one occurs it typically takes a short time. Our data indicate that a trial will add, on average, 6.7 hours to the time lawyers spend on a case.

None of the lawyer characteristic-variables, with the exception of craftsmanship, has any statistically significant effect on hours. The other five lawyer variables were introduced into the model to test lawyer "'productivity."' We thought that more specialized, qualified and experienced lawyers would be able to do a task more quickly, and, ceteris paribus, these variables would reduce hours. Our expectation was not confirmed; these variables have no significant effect on hours, one way or the other. One explanation of this result is that increased capacity can cut two ways: better lawyers do things faster, but can also think of more things to do. Another purely statistical explanation is that there is not enough variation in our sample to catch the effect which lawyer experience and specialization actually have on hours.⁽⁵⁸⁾

The impact of the lawyer's own goals is interesting. We found that two of the lawyers' goals had an effect on hours. The lawyers who said they took the case for public service reasons devoted *105 fewer hours to the case. In contrast, lawyers who included among their goals enhancing their own or their firm's reputation put in more hours than they would have in an otherwise similar case. Moreover, this cluster was one of the more important ones in explaining overall variation in hours. Full assessment of these findings would require us to compare billings in these cases with the amount charged clients in other cases. But if the hourly lawyer who spends more time on cases that advance the lawyer's own career also charges the client for these hours, our findings would confirm Johnson's theories about the divergence of lawyer and client interests in the litigation situation, and his view that lawyers are able to charge for time that provides little benefit to the client.⁽⁵⁹⁾

A related finding, albeit a negative one, is that client control and participation in a case has no effect, one way or the other, on the amount of time the lawyer spends. We hypothesized that the more control the client exercised on decisions, the fewer hours the hourly lawyer, and the more the non-hourly lawyer, would spend on the case. Quite to the contrary, the data show that client control and participation, at least as we measure it, have no effect whatsoever on the number of hours invested in a case.⁽⁶⁰⁾

A noteworthy element in Table 8 is the independent significance of court type. We found that after controlling for all the other variables in the model, including stakes, complexity of the case, and type and number of events, hourly fee lawyers spent about 13 more hours on a case litigated in federal court than on an "essentially similar"' case in the state courts.⁽⁶¹⁾

We sought to determine whether this effect could be explained by the formal rules of procedure in the two types of courts. We found, however, that cases in state courts which use the Federal Rules of Civil Procedure take less time than similar federal cases. This led us to wonder if variations in practice, rather than the formal rules, account for the court effect. Perhaps systematic variation in what the judges expect from the lawyers, or how lawyers treat their federal as opposed to state cases, explains why federal cases take more time. Do federal judges demand more work from lawyers? Do lawyers look at the federal case as the "big time"' and invest more time in case preparation? We have not yet tested these propositions statistically, but there is *106 anecdotal information to support them. Our field coding staff reported that not only were the federal courts more likely to require briefs (or, in New Mexico, a written statement that a brief would not be filed), but that federal court cases were generally more complex. "Reading a federal court file was like reading a story,"' one coder reported; "there was a discernible 'plot' and a conclusion."' Reading a state court file, on the other hand, "was like reading a recipe."'⁽⁶²⁾

We have already noted that the events cluster is the most important in explaining the number of hours lawyers invest in a case. While this finding may seem tautological, since an "event"' is something that takes time, we think it has an independent meaning and is quite important. In the first place, no more than half the time lawyers spend on cases can be attributed to these procedural "events."' Second, detailed analysis of the data shows that the amount of time per "event"' varies significantly among our cases.⁽⁶³⁾ Thus, there is no one-to-one correlation between events and hours.

Rather than seeing events as a reflection of hours, we view this variable as a surrogate for the effect of strategic interaction in litigation. It seems obvious that one of the factors that influences how much time a lawyer spends on a case is what the other side chooses to do in the case. If one side takes a deposition, the other usually attends and participates. If one side moves for summary judgment, the other will have to respond or risk an adverse ruling on the merits. Our events variables come from the court records, so they include events initiated by either party. We see the relative strength the events cluster has in the explanation of hours as reflecting, in part, the effect on the lawyer of strategic moves made *107 by the other side.⁽⁶⁴⁾

Figure 6 reflects this understanding of the dynamics of litigation and summarizes the causal model of litigation investment which emerges from our analysis. Causal relations are shown by arrows. Note that we show two types of causal influence. For *108 each party, the general variables directly affect the hours invested and influence the events initiated by that side. But the events initiated by the other side also influence the other's events and thus the total hours.

3. Non-Hourly Lawyers

We have already stated that the factors explaining non-hourly lawyer decisions seem to be quite different from those which influence the hourly lawyer.⁽⁶⁵⁾ (Only our case characteristic and events clusters were statistically significant for the non-hourly lawyer.) Johnson⁽⁶⁶⁾ and others⁽⁶⁷⁾ have hypothesized that the non-hourly lawyer would spend less time on a case, other things being equal. We conducted independent tests to see if this difference appeared in our cases.

Our findings are ambiguous. We found that the median hourly lawyer spent slightly more time on a case than the median non-hourly lawyer, but the difference is not statistically significant. We calculated ratios of hours per event and hours per dollar of stakes in the case. In both cases the results were higher, not lower, for the non-hourly lawyer. That is, the hourly lawyer spent fewer hours per event and less time per dollar of stakes. These findings are not statistically significant. The data are summarized in Table 9.

31 UCLA Law Review 72-127 (October, 1983)