Assessing the Argument for Specialized Courts: Evidence from Family Courts in Spain* by Nuno Garoupa** Natalia Jorgensen*** Pablo Vázquez**** DOCUMENTO DE TRABAJO 2008-16

April 2008

The authors acknowledge the financial support of Fundación Rafael del Pino for this research, the help provided by the Spanish Judicial Council [Consejo General del Poder Judicial] in accessing the data and research assistance of Paloma Leal and María Dólores Delgado Inglada for gathering the data. Obviously the views expressed by the paper do not bind in any circumstance those institutions. We thank Matilde Machado for helpful comments and Yeny C. Estrada for editorial assistantship. Nuno Garoupa also acknowledges funding from the Portuguese Government, FCT, PIQS/ECO/5003/2003. All the remaining errors are our own.

University of Illinois College of Law & IMDEA (Madrid). Email: ngaroupa@law.uiuc.edu FEDEA.

FEDEA, Universidad Complutense

Los Documentos de Trabajo se distribuyen gratuitamente a las Universidades e Instituciones de Investigación que lo solicitan. No obstante están disponibles en texto completo a través de Internet: http://www.fedea.es. These Working Paper are distributed free of charge to University Department and other Research Centres. They are also available through Internet: http://www.fedea.es.

ISSN:1696-750X

Jorge Juan, 46 28001 Madrid -España Tel.: +34 914 359 020 Fax: +34 915 779 575 infpub@fedea.es

Abstract

Specialized courts have become a key component of the legal reform packages implemented in civil law countries, particularly, in the area of family law. One argument for this policy is that they are able to reach a decision faster than the regular courts, which are normally congested. We use data from a survey of Spanish family courts in the region of Madrid to test this claim. After controlling for other relevant variables, the econometric results did not provide strong support for specialized courts.

1. Introduction

Specialized courts have become an essential component of legal reform packages in civil law countries.¹ While they are less popular in common law countries, these courts have nevertheless made their way through. The widespread enthusiasm for specialized courts enjoyed by legal policymakers is not always shared in legal academia.³ Moreover, many of the advantages of specialized courts that are emphasized by legal policymakers have not been empirically validated. In particular, it is not clear if specialized courts assure higher quality decisions in a shorter time frame (i.e., more efficiently than in the congested court system).

Family law is one of the areas of law which increasingly uses specialized courts. Not only are the social implications of family law extremely relevant, but the contemporary changes in behavior in marriage and divorces have put courts under pressure. More and more people are getting divorced. Faster decisions are being demanded by the citizens. More and more children are affected by increase in divorces; therefore, their interests are paramount in family law. In civil law countries, the response to the higher rate of divorces, and the consequent impact on an already congested court system, has been the development of a network of specialized family courts.⁴

Based on the existing literature on specialized courts, Table 1 summarizes the advantages and disadvantages of specialized courts in family law. In this paper, we investigate the major argument in favor of faster decisions in divorces and other familyrelated cases. Clearly, the conclusion that specialized courts offer decisions in a shorter period of time does not necessarily imply that family law is better enforced. However, if specialized courts cannot decide cases faster than regular courts, the argument for specialization is certainly less compelling. Therefore this is a crucial test.

¹See Carlo Guarnieri, Judicial Independence in Latin Countries in Western Europe, in Judicial Independence in the Age of Democracy, Critical Perspectives Around the World (ed. P. H. Russell and D. M. O’Brien), University of Virginia Press (2001).

² See Rochelle C. Dreyfuss, ‘Specialized Adjudication’, 377 BYU L. Rev. (1990) at pp. 393-96.

³See, among others, Richard Posner, ‘Will the Federal Courts of Appeals Survive Until 1984? An Essay on Delegation and Specialization of the Judicial Function’, 56 S. Cal. L. Rev (1983) p. 761; Richard L. Revesz, ‘Specialized Courts and the Administrative Lawmaking System’, 38 U. Pa. L. Rev. (1989) p. 1111; Rochelle C. Dreyfuss, ‘The Federal Circuit: A Case Study in Specialized Courts’, 64 NYU L. Rev (1989) p. 1; Nicholas Bagley and Richard L. Revesz, ‘Centralized Oversight of the Regulatory State’, 106 Columbia L Rev (2006) p.1260.

⁴Similarly in the US and in Australia. See, for example, Barbara A. Babb, ‘Where We Stand: An Analysis of America’s Family Law Adjudicatory Systems and the Mandate to Establish Unified Family Courts’, 32 Family Law Quarterly (1998) p. 31 and Alaistar Nicholson and Margaret Harrison, ‘Family Law and the Family Law of Australia: Experience of the First 25 Years’, 24 Melbourne University Law Review (2000), p. 756.

Table 1: Costs and Benefits of Specialized Courts

We use data from a survey of Spanish family courts in the region of Madrid. Spain is a particularly interesting case where specialized courts to address family law cases have been actively pursued by legal policy. Furthermore, recent law reforms have been quite relevant in this matter. First, as a consequence of the new divorce law (Ley 15/2005, of July , there has been a significant increase in divorces by mutual consent, and a considerable reduction of divorces subject to litigation in trial. Second, a new legal framework for domestic violence 1/2004, of December has effectively transferred important competences concerning divorce and family-related cases to the new specialized courts for domestic and gender violence. Intuitively, these changes have reduced the caseload pressure on family courts (since the cases subject to litigation, such as divorce or other family-related cases with issues of custody and alimony, are the ones that require more time). On the other hand, they have also reduced the benefits of specialization since fewer complex cases are actually litigated in family courts, and therefore the demand for specific knowledge from the court has decreased.

In section two, we provide an overview of the family court system in Spain. In section three, we present the dataset. Section four will discuss the regression analysis. And section five provides the conclusion.

(2^nd

⁵For a general overview, see Elena Merino Blanco, Spanish Law and Legal System, Thomson edition, 2006).

2. Family Courts in Spain

In Spain, there are currently 76 specialized family courts across 25 (of the existing 50) provinces. In 2004, there were 65, confirming the current trend to expand the network of family courts. Generally, family courts are located in the capital city of a province, and have permanent teams of psychologists and social workers to assist the judges. Outside of their jurisdiction area, family law is a matter of regular courts. Particularly, in the civil courts of first instance in the provinces where civil and criminal matters are separated. Regular courts share experts on a demand-based system. Madrid, alongside Catalonia and Andalucia, are the three regions with the most volume of family law cases filed. There can be no forum shopping in family law cases since the cases must be filed in the district of residence of the defendant.

3. Dataset

Our sample includes two specialized courts in the capital city of Madrid (Madrid 24 and Madrid 25) and two regular courts (so-called mixed courts in Spain) from the periphery of Madrid (one from Getafe and the other from Majadahonda). The population of the city of Madrid is only served by specialized courts when it comes to family law, whereas the population in the periphery is served by regular courts. These four courts have been assessed by the Judicial Council (the independent body that runs the judiciary) with having meritorious performance and reasonable dimension. The chosen peripheries are reasonably similar to the capital city in terms of demography and socioeconomic variables.

The data was obtained from these four mentioned courts for all cases filed from September 1, 2005 to August 30, 2006. We have excluded all filed cases that were still pending by the end of the period, which are a total of 77 cases (for example, cases filled in the second semester of 2006; this naturally generates a sub-estimation of the average duration of a case), and all filed cases that were rejected by the courts for procedural reasons (which were 21 cases). These exclusions were determined by purely operational means since the access to the court files is manual (not by computer) and confidential (hence limited in time by the Judicial Council).

We have coded all mutual consent cases, as well as, cases subject to litigation. They include divorces, but also other family-related matters such as separations (predivorce, not mandatory under the new law), changes in divorce agreements, alimony and custody agreements, and other preliminary or provisional pre-trial issues.

There are 1549 observations in the dataset, 1434 cases filed in specialized courts (100% of their workload), and 115 cases filed in regular courts (less than 10% of their workload). Also out of 1549 cases, 68% are by mutual consent, and 32% are subject to litigation. However, from the cases subject to litigation, 17% were in later stages settled by mutual consent, hence only 15% of the cases were effectively decided by trial. Divorces represent 84.6% of the mutual consent cases, but only 43.2% of cases to be litigated.⁹ Another important difference is that, while cases by mutual consent seem to be uniformly distributed over the year, cases subject to litigation have peaks in September, October and November.

⁶Consejo General del Poder Judicial.

⁷Information provided orally to the authors.

⁸There are 895 divorces, 88 pre-divorce separations, 49 alimony and custodial agreements, and 26 changes of mutually agreed family-related issues.

The duration of a case is the time period from the appropriate filing of the case until the judge makes a final decision in either the first instance or in a second instance (for those cases that are appealed to a higher non-specialized court). We have excluded the time from the judge’s decision to registration in the civil registrar since that is purely administrative and outside of the control of the courts. Cases by mutual consent have an average duration of 87 days and a standard deviation of 47 days (Table 2). The corresponding figures for cases subject to litigation are 185 and 91 days, respectively (Table 4). It can also be observed that for the cases filed by mutual consent, 25% took 59 days or less and 50% took 73 days or less. These numbers are 124 and 180 days respectively for cases subject to litigation. Further characterization of the dataset is described by Tables 2 to 4.

From Table 2 we can see that there are significant differences in the average duration across courts. This may be due to different strategies the individual judges take to solve cases that are not that complex.

Table 2: Characterization of Dataset for Divorce and other Family Procedures by Mutual Consent

Table 3: Characterization of Dataset for Non-Mutual Consent Divorce and other Litigious Family Procedures (First Instance Only)

⁹There are 212 divorces, 105 pre-trial measures, 73 not mutually agreed family-related issues, 49 alimony and custodial decisions, 23 pre-divorce separations and other 29 family law disputes.

Table 4: Characterization of Dataset for Non-Mutual Consent Divorce and other Litigious Family Procedures (Including Appeals)

The descriptive statistics of the dataset with respect to litigated cases (i.e., those with a higher degree of expected complexity) do not seem to support the hypothesis that specialized courts are able to deliver decisions quicker. In fact, for cases subject to litigation, for which the argument for specialized courts is more compelling, we can observe, in Tables 3 and 4, that one has the lowest (Madrid 24) and another has the highest (Madrid 25) duration.

4. Regression Analysis

To analyze the determinants of the duration for the proceedings in family court, we have developed an econometric exercise that allows the identification of the partial impact of the relevant factors on the duration. We determine what variables explain the likelihood that each case, subject to litigation, will be concluded within a certain period of time. The reason why we concentrate on litigated cases is because they are fundamentally the reason under the development of the specialized family court system.¹⁰ The duration of cases by mutual consent is essentially driven by administrative and other formal procedures, and not the need for specific knowledge in managing complex situations.

After excluding those cases that were dropped by the plaintiff during the process (44 cases in specialized courts and only one in regular courts) and those for which we do not have information on at least one of the relevant variables (9 cases), our final sample consists of 380 observations. For this sub-sample, by reference to Table 3, the average duration in the first instance increased to 190 days, and the standard deviation increased to 97 days.

The econometric exercise is based on an ordered probit following Wooldridge (2002).¹¹ This technique has been chosen given the characteristics of the dataset, namely, the random variable not being normally distributed or symmetrically distributed. A linear regression model for the average duration, conditional on the explanatory variables, could be inadequate under these circumstances. The ordered probit seems more appropriate. We estimate the ordered probit on the probability of a case subject to litigation being concluded before 125 days, between 126 and 175 days, between 176 and 229 days, and more than 229 days. This way we can identify which variables have a statistically significant effect on duration, and assess their partial impact. Table 5 summarizes the data used in the regression analysis.

¹⁰In fact, there is also a pragmatic reason. The court files are not so informative for mutual consent cases. Even for litigated cases, we had to exclude nine cases due to lack of information concerning at least one of the explanatory variables.

¹¹See Jeffrey Wooldridge,. Econometric Analysis of Cross-Section and Panel Data (Cambridge, Massachussets; London, England, MIT Press 2003)

Table 5: Non-Mutual Consent Divorce and other Litigious Family Procedures by Duration [Dependent Variable of Ordered Probit]

We run the regression for the duration of cases in first instance. For robustness, we repeat the exercise for the total duration of the process (the first instance and second instance processes when appealed¹²).¹³

We control for the court and the quarter where the process began.¹⁴ One of the caveats of the study is the small sample size of regular courts, which is only 8.3% of the sampled cases. Although purely driven by operational constraints as explained before, we believe that this information is enough to get an idea about the differences in duration cases. As is shown in Table A1 in Appendix, most of the cases start in the four quarter of the year (40.3% of the cases).

The variables we use to control for the complexity of the case are the existence of minors (because it requires the intervention of family law prosecutors, 69.2% of the cases), the existence of pretrial measures (in 11.3% of the cases), and the request for expert evidence (in 12.3% of the cases). Furthermore, we include variables to account for administrative procedures that can affect the duration of the case. The variables accounted for, which delay trial in administrative procedures, are: locating one of the parties, request for legal aid lawyer, and problems with court services (including change of judge). Finally, error in filing is also included. That is a good proxy for the quality of legal services offered by the lawyers.¹⁵ We have not included the decision to change the procedure from litigation to mutual consent (19.2% of the sample), since most of them occur in one particular court (Getafe), where apparently the judge actively favored this kind of solution.

We also have collected information about individual characteristics of the plaintiff and the defendant such as gender, nationality, and labor status. We include these variables in two of the specifications to evaluate the robustness of our results in the Appendix, see Tables A2 and A3.

We present the main econometric results in Tables 6 and 7, and the extended results in Tables A2 and A3. The baseline case, to which the econometric results must be compared to, is the one filed in the first quarter in the specialized family court Madrid 24; being concluded with preliminary measures or dropped (hence not reaching a final sentence), does not require the intervention of the family law prosecutors (hence no minors were involved), exhibits low complexity (no pretrial measures or expert evidence was requested), no party requested an adjourning trial, no administrative proceedings delayed trial, no request for legal aid lawyer were made, no problems with court services, and there were no errors in the filing.

¹²The number of appealed sentences is 12 (11 in specialized courts and one in regular courts), all confirmed in the second instance.

¹³ We use STATA for the regression analysis.

¹⁴Denoting the beginning of the case the date, the plaintiff brings the lawsuit in the form that the court considers to be procedurally correct.

¹⁵We also have information about the quantity of lawyers for each party as well as if they change lawyers during the proceedings. However, they are very small numbers, and therefore, have not been included in the regression analysis.

The estimated coefficients cannot be directly interpreted usually with probit models. The relevant information is their sign, which permits a qualitative evaluation. The sign of a given coefficient shows the impact of the independent variable on the likelihood that the duration is one of the four categories. In order to facilitate a correct interpretation, the estimated coefficients refer to the marginal effect of each category.

Generally, the estimated coefficients of the variables used as the controlled variables have the expected signs (see Tables A2 and A3 in Appendix). Tables 6 and 7 show (for first instance only and for total duration respectively) that Madrid 24 is faster than all the others, although, other important variables have significant explanatory power. Apart from the court where the proceedings take place, the statistically significant variables with a positive marginal impact are the type of decision taken by the court (final sentence or preliminary measures subject to later review), decision to adjourn at the request of one party, the existence of administrative procedures that delay trial, and existence of errors in filing for divorce or other family-related procedures that need to be corrected before the proceedings start.

A negative sign means that the explanatory variable has a negative impact on the duration of a proceeding in a family law case. For example, the existence of minors also generates interesting results when it is an issue in the different courts. Pre-trial measures as expected have a negative sign, but not statistically significant; the need of expert evidence, representation by legal aid lawyer, and problems with court services as expected have a positive sign , but again not statistically significant impact.

Table 6: Ordered Probit. Dependent Variable: Duration in First Instance of Nonmutual Consent Divorces and other Litigious Family Procedures

Robust Standard errors in parentheses; * significant at 5%; ** significant at 1%.

Table 7: Ordered Probit. Dependent Variable: Total Duration of Non-mutual Consent Divorces and other Litigious Family Procedures

Robust Standard errors in parentheses; * significant at 5%; ** significant at 1%

The econometric model also provides for the probabilities of each category of duration conditional on the explanatory variables. These probabilities are 0.154 (less than 125 days), 0.293 (from 126 to 175 days), 0.290 (from 176 to 229 days), 0.262 (more than 229 days) for the total duration; 0.156 (less than 125 days), 0.273 (from 126 to 175 days), 0.277 (from 176 to 229 days), and 0.292 (more than 229 days) for duration in first instance only.

The estimations are fairly robust to the different specifications (see Tables A2 and A3 in Appendix). We have also developed a further test of robustness by running identical regressions with a dummy for family courts (Madrid 24 and Madrid 25) and regular courts (Getafe and Majadahonda). As expected, the coefficient for family courts is not statistically significant, as seen at the bottom of Tables 6 and 7, and with more detail in tables A4 and A5 of the Appendix.

¹⁶Dummy variable for both Madrid 24 and Madrid 25 family courts.

5. Conclusions

By looking at some family courts in Spain, we have assessed the extent to which specialized courts are actually faster in reaching a decision. This is an important test for any argument that is based on the higher quality of specialized courts. The econometric evidence is not strong but seems to point out that specialized courts (in our sample, particularly Madrid 24) could conclude litigation with a lower average duration than regular courts (in our sample, Getafe and Majadahonda), after controlling for other important variables. However, the overall results do not provide strong support for the claim that specialized courts, when handling family cases, are indeed faster.

Obviously, these empirical results are not to be hastily generalized since we do not have an exhaustive set of information about each case that allows us to explore other potential relevant variables. Furthermore, one should not confuse assessing the duration with a full-fledge evaluation of the merits. Nevertheless, the results are sufficiently important to contrast with the conventional optimism of legal policymakers for court specialization. The results suggest caution, and ask for a more comprehensive assessment. They also point out that current optimism should be carefully restrained.

Appendix

Table A1: Descriptive statistics

(+) heteroskedasticity and correlations between the observations from the same court adjusted for (**) significant at 1 per cent (*) significant at 5 per cent

(+) heteroskedasticity and correlations between the observations from the same court is adjusted for (**) significant at 1 per cent (*) significant at 5 per cent

(**) significant at 1 per cent (*) significant at 5 per cent (+) heteroskedasticity is adjusted for