DEPARTMENT OF ECONOMICS Uppsala University
Bachelor’s Thesis (Examensarbete C) Authors: Katarina Hagstedt & Julia Proos Supervisor: Lars Lindvall
Spring Semester 2008
Has the recent restructuring of the Swedish district courts improved efficiency?
A DEA analysis
ABSTRACT
This paper conducts an efficiency study on the Swedish district courts following the restructuring in the late 1990s using DEA.
The objective is to establish whether the efficiency has improved after a reform that significantly decreased the number of courts.
The data is county-based with wages and administrative costs as input and cases settled as output. The results show an overall increase in efficiency. Moreover, many units seem to be operating at decreasing returns to scale.
I. INTRODUCTION
According to the welfare theorem, competitive markets allocate resources in an efficient manner. Assuming this occurs in an ethically acceptable way, the government’s main purpose would therefore be to establish law, order, property rights and a court system (2006, Morgan).
Thus the effectiveness of the legal apparatus is important for the efficiency of the economy.
This study aims to find out whether a recent restructuring of the Swedish district court system has improved its efficiency. The task of the district courts is to handle crime and civil cases, and their main objective is to do this in an efficient fashion while guarding the legal rights of the individuals (Statskontoret, 2007). Due to lack of information and difficulties in ensuring its efficiency, there is a problem in monitoring district courts. The situation warrants a study of the efficiency of the courts.
Due to a government decision in the late 1990s, there has been a major restructuring of the Swedish district courts throughout 1999-2006. The mergers have resulted in a decrease in the number of district courts from 96 in 1999 to 55 in 2006, with the greatest changes made throughout 2001-2002 and 2004-2005. The major goal with this reform was to in the long run make the district courts more efficient by merging several of them together into larger units.
The main motive for these mergers was the need for larger courts that would have a greater judicial and administrative supporting capacity (Statskontoret, 2007). A contributing reason was the preceding centralisation of the police force, detention facilities, correctional facilities and prosecutors’ offices. A consolidation of the courts therefore seemed like a natural development, which hopefully enables the courts to provide better and more efficient services to Swedish citizens. The government has put in place some more concrete goals for the short run (Statskontoret, 2007): i) create opportunities for a stronger organisation ii) increase opportunity for specialisation iii) create development of special competences iv) secure opportunity to hire competent staff v) ameliorate the regional coordination with other judicial services.
The focus of this study will be on the years 1998-1999, representing the very beginning of the restructuring, and the years 2006-2007 at the very end of the reform process. By observing four different years, situated in different stages of the restructuring, it is possible to make a
comparison before and after the reform, given that a comparison will show whether these new larger units are more cost-efficient than the old structure.
The examination conducted by the Swedish Agency for Public Management (Statskontoret, 2007) concludes that the government’s short term goals, previously mentioned, have been fulfilled to a certain extent. However, there is still more work to be done, and there are some differences between the different district courts. A general trend is that the amount of cases handled by the new larger units is greater, and the amount of staff hired by the courts is also larger. Larger district courts would therefore not only increase the competence and quality of its services, it would also enable the courts to gain legitimacy since they are handling more cases than before. The Swedish Agency for Public Management (2007) has in its study excluded any conclusion of the general efficiency of the new system. Empirical studies have already been conducted by the Swedish Agency for Public Management (2007) and Riksrevisionsverket (2001), where the results of the restructuring have been argued to be favourable.
This paper adds to these studies by focusing on the efficiency of the district courts at the county level through a quantitative study. Data Envelopment Analysis (DEA) is the method used to evaluate efficiency with cost statistics as input and data on cases settled as output.
Through this other aspects are considered, such as efficiency in terms of costs and whether the reform was successful from an economic point of view. Previous studies conducted in Brazil and Norway have examined the efficiency of courts and suggest economies of scale (Kittelsen
& Førsund, 1992 and Sambaio & Schwengber, 2005). The Norwegian study court efficiency in order to compare generalized rural courts with specialized city courts, their results show that the optimal court size is larger than the current average court size. The Brazilian study calculated the efficiency of justice courts in the region of Rio Grande do Sul. They concluded that smaller courts waste resources and have higher average costs, suggesting economies of scale due to the specialization possible in larger court units. Given the similarities in court systems, an economy of scale is used as an underlying assumption. However, there must be a limit to the economies of scale. In this study, the indication is that the results will show increased efficiency, since the restructuring lowers the production cost of legal services.
II. MEASURING EFFICIENCY AND DEA
One problem that arises when handling data concerning the judicial system is the fact that courts are subject to monopoly. The efficiency of a district court depends on what is measured. Focus can be directed towards number of cases settled, hours per case, caseload accumulated, or convictions, etc. The problem is that efficiency cannot easily be measured since different cases are of different nature. A case is not a standardized unit of production.
If one would consider the amount of time spent per case this would vary depending on case type, number of witnesses, evidence and other factors. Convictions are likewise an unsuitable measure of efficiency as it would be favourable for courts with many convictions, which is a reflection of their cases rather than their efficiency. Courts generally have an accumulated caseload, studying this does not consider their production, but rather the demand for court services. Therefore accumulated caseload is not desirable as a measure of efficiency. Total hours of court services would be an alternative, yet there is no available data on this. Cases settled represent the production of court services, it is also imperfect, as cases are of different nature, but is the preferable alternative for an efficiency study. The data is available, and is consequently used as output measure.
When choosing inputs for the efficiency analysis, there are several options. Since it is difficult to estimate the actual value of the production of the district courts, given that there is no market evaluation of the output, using the inputs is the best way to quantify the output.
The main input into this production process is labour which is composed of judges and administrative staff. Due to the lack of data on the size of staff, and the difficulty in comparing the value of a judge to that of administrative staff, focus will be on their costs.
Another considerable expense for the court system is that of buildings. It is quite likely that court buildings are not used in the most efficient way, considering alternative costs. The government may choose to keep a court in a historical structure instead of locating it in a cheaper building. Moreover, when attempts are made to manage costs, the courts have incentives to over-report actual costs (Morgan, 2006) enabling them to collect rents.
Investments, educational costs and rents paid for buildings are not included as these are handled at a central level, and do not debit specific district courts. Therefore labour costs are used as input.
Social costs that may increase as plaintiffs have to travel farther to reach their nearest district court are disregarded. At the same time savings that can be made as the police force, detention facilities, correctional facilities, prosecutors and district courts can be synchronized are ignored (Collin, 2007). It is assumed that the savings from police transports to different civil courts and other synchronization effects offset the increased social cost incurred. Social costs may or may not be larger than the savings; this assumption does, however, substantially simplify the model.
A non-technical description of the DEA programme follows, based on Coelli. DEA constructs a non-parametric envelop frontier over a set of data points such that all observed points lie on or below the production frontier. The economic efficiency measured consists of two components; technical efficiency and allocative efficiency. The technical efficiency of a firm reflects its ability to obtain the maximal output from a given set of inputs, and the allocative efficiency reflects its ability to use its inputs given its respective prices.
Table 1 and Figure 1 illustrate a data envelopment frontier. The variable returns to scale (VRS) frontier goes from (1,0) to point A, then continues from point A to point B, finally extending endlessly to the right from point B. At point A it is implied that since the production of 2 units is possible with 1 unit of input, it must also be possible to produce less.
Table 1. Example data
Firm Input Output A 1 2 B 2 3 C 1,5 2
Figure 1. Example of a data envelopment frontier using data from Table 1.
At point B the production of 3 units requires 2 units of input, meaning that 3 units could also be produced using more inputs. The points on the frontier are efficient with VRS, whereas C is inefficient. The dashed line shows possible efficiency improvement; theoretically C could either increase its output while keeping its input constant, or decrease its input while maintaining its level of output.
The constant returns to scale (CRS) frontier goes from the origin through point A, and is defined by the most efficient point out of a set of data, in this case point A. CRS assumes that production should be possible with the same ratio between input and output as the most efficient point. With CRS both points B and C are inefficient. All points to the right of A beneath the CRS line have decreasing returns to scale, whilst all the points to the left of A and beneath the CRS line have increasing returns to scale.
All DEA models are available in an input or an output orientation. While the input oriented version calculates how inputs can be proportionally reduced without changing the output quantities produced (Coelli, 1996), the output orientation focuses on how outputs can potentially be expanded without changing inputs. When calculating data under the assumption
A
B
C
0 0,5 1 1,5 2 2,5 3 3,5
0 0,5 1 1,5 2 2,5
Input
Output
CRS
VRS
of CRS, the input- and output oriented DEA will provide equivalent results. Under some circumstances the observed unit is not operating at its optimal scale, due to imperfect competition, monopoly, financing difficulties, etc., and it will therefore be more suitable to conduct calculations under the assumption of VRS. As the district court system is a monopoly this study uses VRS.
The results from DEA are displayed in terms of constant returns to scale technical efficiency (CRSTE) and variable returns to scale technical efficiency (VRSTE). The measure of VRSTE is predominantly used since the assumptions are made based on VRS. DEA treats each county as its own decision making unit that has the ability to reorganise its activities in order to increase efficiency. The results suggest ways in how these units can improve its efficiency, unfortunately, these values are of little value significance to the study, since decisions are made by the district courts and not on a county level.
DEA is a useful method for the study since it makes it possible to utilize the limited available information. Once the production possibility frontier has been established, it can be used it for various efficiency measures. This technique also makes it possible to make a simple comparison of the efficiency of district courts before and after the reform. This is a simple model, but as the data is imperfect, a complex model might not provide any viable results.
Among the different DEA programmes the DEAP computer program is used (for a more technical description of the DEAP programme see Coelli).
Output oriented DEA with technical-, scale- and cost efficiencies is used, since estimates can then be made of how much output quantities can be proportionally expanded without altering the input quantities. District courts have a fixed quantity of resources and are asked to produce as much output as possible. Focus will be on the costs, in order to see how efficiently the courts are using these funds in terms of cases settled before the reform. A comparison will be made with the current reformed system of district courts, where fewer, larger district courts will receive more funds, but at the same time will have more cases to handle. This will enable us to draw a conclusion of whether the new system is more cost efficient or not in terms of settled cases. The calculations are made on a multi-stage DEA.
this case. Quality in the production of court services means that trials are settled in a fair and just manner. This assumption implies that courts with lower costs are not less professional.
III. DATA
The data is collected from three different sources; the Swedish Central Bureau of Statistics (SCB), National Courts Administration and the Swedish National Council for Crime Prevention (BRÅ) (see Table 3 in Appendix I which displays the data covering cases settled and costs).
While the number of cases settled is readily available at the BRÅ homepage for the years of 1998-1999, the National Courts Administration are now handling the statistics of the district courts, and was therefore used as a source for the years of 2006-2007. Since the statistical duty has been restructured in between these two periods, the data contains slightly varying information as it has been collected for different purposes. For example, the data from BRÅ categorizes according to type of crime per county, whereas the National Courts Administration categorises according to type of court case per district court.
As this subject relates to a government body, much of the data has been accessible on the internet. Since there are no statistics on the activities of the separate district courts for 1998-1999, this study is based on the district courts grouped into Sweden’s 21 counties.
In order to normalize the results a parallel calculation on efficiency per capita has been conducted. In this manner it is possible to make a comparison to see if the same district courts remain efficient when removing the fact that the counties are of varying sizes. Due to the limitation to statistics on a county basis, this comparison attempts to eliminate those differences and make a robustness check. The population statistics are available on the SCB homepage. The county figures may not be completely comparable between periods, for example Uppsala county included Heby from the first of January 2007, which was considered part of Västmanland previously. This however, is a slight change and will not affect the study.
Aside from this change there may have been other minor changes between 1999 and 2006.
The statistics on the costs of the district courts has been kindly supplied by Ann-Christin Sundell, a controller at the National Courts Administration in Jönköping. These costs correspond only to the amounts that the district courts themselves are responsible for;
wages and administrative costs. Consequently these costs are used to assess the value of output and thus the counties efficiency. If there are economies of scale and the competence has increased due to the increased caseload after the reform the courts will prove more efficient.
Unfortunately, some of the figures from the Swedish National Council for Crime Prevention also comprise the cases of county administrative courts, rent tribunal and/or National Legal Aid Authority. Västernorrland is the only county that incurred costs for rent tribunal and National Legal Aid Authority, in the cost figures for 2006 and 2007. Västernorrland was excluded for these years, because it was not only burdened by those costs but also by the costs of county administrative courts. The costs of county administrative courts were included for the years 2006-2007 for the following courts: Kronoberg, Blekinge, Västernorrland, Jämtland and Västra Götaland. All of these, except Västra Götaland, had only one district court during this period, and the cost increase was dramatic when compared to other counties of similar size. As Figure 2 displays, results are presented without Kronoberg, Blekinge, Västernorrland and Jämtland. For Västra Götaland the district court (Mariestad) is deducted since it was debited with county administrative courts’ costs from the total costs and settled cases.
The motive for this is the difficulty in estimating the costs for the county administrative courts, and the extent of their effect on the district courts. When removing these counties, the problem of Västernorrland’s specialized court costs is eliminated as well. By excluding them only costs concerning district courts are included, which makes the data more reliable. Aside from this, five of the district courts handle environmental cases during 2006-2007:
Västerbotten, Jämtland, Västra Götaland, Kronoberg and Stockholm. As the data for 1998- 1999 was supplied by a different authority it is unknown whether they also handled environmental cases these years, and no special adjustments have been made to compensate for this. These cases have been treated as all other cases, in the same way that the National Courts Administration includes them.
a base year to make the costs comparable. Finally, in order to minimize potential extreme values, mean values of cases settled and costs for 1998/1999 and 2006/2007 are calculated.
IV. ANALYSIS
Figure 2 and Table 2 display the results from DEA. Gotland is the first county defining the VRS frontier. According to DEA it operated at an efficient level both before and after the reform. This figure is based on the assumption that the production possibilities have not changed through the time period. The methods of production have not improved technically.
Figure 2. Production frontier from DEA analysis of Swedish District Courts on county level for the years 1998/1999 and 2006/2007
The county of Blekinge was according to DEA operating at an optimal level in terms of efficiency before the reform. It defines the CRS- as well as the VRS frontier. Compared to Blekinge, some other district courts have increasing returns to scale, while most have
0 5000 10000 15000 20000 25000 30000 35000 40000 45000
0 200000 400000 600000 800000 1000000 1200000 1400000 1600000
Costs in SEK
Settled Cases
1998-1999 2006-2007 CRS
VRS
CRS = Constant returns to scale VRS = Variable returns to scale
decreasing returns to scale. This suggests that most counties would benefit from downsizing their district courts. However, since this study is on a county basis, there is a risk that one inefficient district court may deteriorate the results for the entire county. In order to find the optimal scale a study should be conducted per district court, which exceeds the purpose of this study. Therefore the indication that most district courts have decreasing returns to scale is inconclusive.
The county of Östergötland is the third point establishing the VRS frontier in 1998-1999, but has decreased its efficiency according to VRSTE after the reform. This is due to a decrease in caseload while at the same time incurring increasing costs. Östergötland therefore differs from most of the other counties.
Skåne is efficient after the reform, as the fourth point defining the VRS frontier. Skåne’s costs have increased somewhat while the caseload has increased even more, causing this improvement in efficiency.
Stockholm’s county is also more efficient after the reform and defines the last point along the VRS curve. This might be due to its substantial increase in settled cases, with approximately 50% more cases (Swedish National Council for Crime Prevention 1998/1999, National Courts Administration 2006/2007), while the costs have increased at a slower rate. When looking at other big city regions such as Göteborg (Västra Götaland) and Malmö (Skåne) there is a pattern that these regions have improved their efficiency.
From Table 2, one can conclude that more than half of the counties have increased their efficiency in between the two periods, while 30 % have decreased their efficiency. Overall, the VRSTE has increased by 0.547, which implies that the counties in general have moved towards the VRS frontier. The counties of Jämtland, Västernorrland, Blekinge and Kronoberg are excluded from the VRSTE values of the second period. With cost data excluding the county administrative courts, it is credible that they would support the general results. Given the unavailability of this data, these counties are omitted from the analysis.
Table 2. Results from DEA using average values of cases settled and deflated costs
County VRSTE 1998-1999 VRSTE 2006-2007
Difference 2006/2007 – 1998/1999
Stockholm¤ 0.78 1 0.220
Uppsala 0.774 0.864 0.090
Södermanland 0.886 0.798 -0.088
Östergötland 1 0.797 -0.203
Jönköping 0.649 0.543 -0.106
Kronoberg* ¤ 0.584
Kalmar 0.683 0.76 0.077
Gotland 1 1 0
Blekinge* 1
Skåne 0.915 0.99 0.075
Halland 0.613 0.747 0.134
Västra Götaland* ¤ 0.848 0.892 0.044
Värmland 0.742 0.702 -0.040
Örebro 0.587 0.837 0.250
Västmanland 0.772 0.712 -0.060
Dalarna 0.559 0.609 0.050
Gävleborg 0.656 0.808 0.152
Västernorrland* 0.746
Jämtland* ¤ 0.472
Västerbotten¤ 0.616 0.446 -0.170
Norrbotten 0.565 0.687 0.122
Total 0.547
Notes:
* = costs of county administrative courts were included for the years 2006-2007, the counties with only one district court are therefore excluded from the average VRSTE for these years.
¤ = courts handling environmental cases
The results support the thesis, since it suggests that efficiency has improved as the units have become larger (economies of scale).
Another factor that might enhance the results is that costs of the government buildings are not counted into the costs of the district courts. These capital costs have probably decreased since many district courts have shut down, resulting in decreased costs with the same amount of cases settled. This would suggest a further increase in efficiency, but since these expenses are not included in the costs of the districts courts, it cannot be used in the analysis. This cost and its effects, nevertheless, needs to be considered.
As mentioned earlier there is a lack of consistency in the activity of courts, since there are different types of cases which furthermore vary in complexity. Some counties that displayed decreased efficiency in DEA may have encountered more complicated cases, whereas others may have had less, or there may have been a general trend that affected all the counties. This can cause misleading results, which is important to bear in mind.
Gotland appears to be an efficient court in the VRSTE results, but disappears when making the analysis per capita. The county only has one district court both before and after the restructuring, it is appearing as efficient merely because it is the smallest production unit. In spite of having increasing returns to scale, it would be difficult to merge it with another unit because of its location. There is a social value in having a district court in each county. Skåne also defines the VRS frontier in Figure 2, but disappears entirely when normalized in a per capita analysis. In Figure 3, where the VRS frontier is based on cases settled and costs per 1000 inhabitants Blekinge defines the VRS frontier in the per capita comparison, as the first point on the frontier.
Figure 3. Production frontier from DEA analysis of Swedish District Courts per 1000 inhabitants for the years 1998/1999 and 2006/2007
5 10 15 20 25
Cases settled per 1000 inhabitants
As seen in the examples of Skåne and Gotland, the results are slightly different when conducted on a per capita basis. In one way it undermines the research purpose by breaking up the larger production units. On the other hand, it offers an opportunity to normalize and compare the results. According to Figure 3, Stockholm appears to have increased its efficiency even more. This would imply that Stockholm is efficient even when distributed per capita, while Gotland and Skåne are not. A population decrease in Skåne and Gotland would increase their efficiency according to the per capita study, and place them closer to the frontier. Consequently, Skåne and Gotland are inefficient per capita although the caseload is not directly related to the population. The per capita study suggests that the caseload in these counties is low in relation to the population. On these grounds the per capita analysis is only used for comparative purposes.
V. CONCLUDING REMARKS
The objective of this study is to evaluate whether a reform has approved the efficiency of Swedish district courts using DEA. The results show that the reform has improved the efficiency with the majority of the district courts showing positive results.
One suspects that the division of statistics of district courts into counties complicates the ability of monitoring. The National Courts Administration faces difficulties in evaluating the efficiency improvement in between these two periods for the activities and costs of the separate district courts since the statistics are collected on a regional basis. The new data collection method enhances the ability to make future efficiency studies.
It would be interesting to conduct a study where the comparison is based on the separate district courts instead of the counties in order to find the optimal scale. The total number of cases handled per county will not differ although several district courts have been merged into larger production units. Therefore efficiency cannot be measured in the same way per county since the distribution of cases does not increase in the same manner as it would in a separate district court analysis.
The results support the conclusions drawn by the Swedish Agency for Public Management, and suggest that this reform was successful also when considering cost-efficiency.
Consequently, the reform was advantageous from both a qualitative and quantitative point of view.
VI. REFERENCES
Coelli. T.J.. (1996) A Guide to DEAP version 2.1: A Data Envelopement Analysis (Computer) Program; No. 8/96. Centre for Efficiency and Productivity Analysis (CEPA) Working Papers.
University of New England.
Collin. L; (2007) ”Fyra tingsrätter i Värmland blev en”. Domkretsen nr. 3. Intellecta Tryckindustri AB. pp. 4-5
Da Conceição Sampio de Sousa. M; Battaglin Schwengber. S; (2005) “Efficiency Estimates for Judicial Services in Brazil: Nonparametric FDH (Free Disposal Hull) and the Expected Order-m Efficiency Scores for Rio Grande Do Sul Courts’” Departamento de Economia.
Universidad de Brasília.
Kittelsen. S; Førsund. F; (1992) ”Efficiency Analysis of Norwegian District Courts” The Journal of Productivity Analysis 3; Kluwer Academic Publishers pp. 277-306
Morgan. W; Katz. M; Rosen. H; (2006) Microeconomics (European Edition) ISBN-139780077109073. McGraw-Hill Education
Statskontoret (Swedish Agency for Public Management). (2007) “Sammanslagna Tingsrätter – En Utvärdering”. Stockholm.
Internet:
National Courts Administration:
Courts Statistics – Official Statistics of Sweden 2006 Courts Statistics – Official Statistics of Sweden 2007
http://www.dom.se/templates/DV_InfoPage____868.aspx (collected April 24th)
Data on the costs of district costs are available upon request from the National Courts Administration, domstolsverket@dom.se.
Swedish Central Bureau of Statistics:
CPI (Consumer Price Index)
http://www.scb.se/templates/tableOrChart____33847.asp (collected April 24th) Population sorted by age. civil status and sex:
http://www.h.scb.se/scb/bor/scbboju/cgi-bin/bjssd/sok_link.asp?sokord1=efter+l%E4n&xu=c 5587001&yp=duwird&lang=1&prodid=BE0101 (collected April 24th)
Population sorted by county for the years 2006 and 2007:
2006 http://www.scb.se/templates/tableOrChart____193256.asp (collected April 24th) 2007 http://www.scb.se/templates/tableOrChart____228181.asp (collected April 24th)
Swedish National Council for Crime Prevention:
Statistics on settled cases for 1998:
http://www.bra.se:80/extra/pod/?action=pod_show&module_instance=8&id=32&statsType=4 60&Year=1998&type=0 (collected April 24th)
Statistics on settled cases for 1999:
http://www.bra.se:80/extra/pod/?action=pod_show&module_instance=8&id=31&statsType=4 60&Year=1999&type=0 (collected April 24th)
Appendix I.
Table 3. Average values of deflated costs and settled cases per county
County 1998/1999 2006/2007
Costs Cases Costs Cases
Stockholm¤ 1230431 25549.5 1473796 38761.5
Uppsala 125590.5 3812 133119 4503.5
Södermanland 122236.5 4251 128872.5 4030.5
Östergötland 169966 6605.5 191084.5 5676.5
Jönköping 116939 2983.5 185366 3793
Kronoberg* ¤ 99044.5 2290.5
Kalmar 98073.5 2652 93609.5 2823.5
Gotland 24509.5 697.5 24364 669
Blekinge* 45938 1907.5
Skåne 664668 17214 717285.5 19894
Halland 122025 2933.5 129214.5 3781.5
Västra Götaland* ¤ 796642 18698.5 872222.5 21352
Värmland 118918.5 3467 119230.5 3287
Örebro 144946.5 3322 123523 4057.5
Västmanland 116467.5 3537 149285.5 4144
Dalarna 135418 2959 137541 3276
Gävleborg 120514 3106 121174 3842.5
Västernorrland* 114179 3353
Jämtland* ¤ 72495.5 1374.5
Västerbotten¤ 107040 2600.5 173215.5 2982.5
Norrbotten 134844 2979 121661 3279.5
Total 4680887 116293 4894565 130154 Notes:
* = costs of county administrative courts were included for the years 2006-2007, the counties with only one district court are therefore excluded from the average VRSTE for these years.
¤ = courts handling environmental cases
