BIAS AND PREDICTABILITY IN JUDICIAL DECISIONS INVOLVING CORPORATE RESTRUCTURING: EVIDENCE FROM SERBIA***

The paper empirically assesses evidence of bias among bankruptcy judges and predictability of judicial decision making in Serbia. For this purpose, we devised three hypothetical cases that were distributed to judges that preside over bankruptcy cases in Serbia. While the results do not indicate a consistent bias towards either the debtor or the creditor, they show concerning unpredictability of judicial decision making, hence leading to a high level of legal insecurity. Based on responses we observed high heterogeneity among Serbian bankruptcy judges in their interpretation of the bankruptcy laws. This makes allocation of cases to the particular judge a significant determinant for how the debtors and creditors will be treated and how the law will be applied.


INTRODUCTION
In an ideal setting, judges should always be impartial, fair and efficient, set aside their prejudices, and follow the rule of law (Rachlinski, Wistrich 2017). Judicial bias, defined as consistent favouring of particular types of claimants (litigants), and unpredictability of judicial decision making may have profound effects on lowering trustworthiness in courts and increasing legal insecurity. 1 While the prevalence of bias and unpredictability of judicial decision making is constrained by law, there is almost always certain leeway allowing judges to exercise discretionary powers, often provided by the very same law. These issues are especially relevant for bankruptcy laws that are commonly perceived to be either pro-creditor or pro-debtor inclined (Ayotte, Yun 2007;Claessens, Klapper 2005;Davydenko, Franks 2008). Besides the inclination of the law, in most legal systems judicial bias may significantly affect the resolution of financial distress. In a nutshell, ceteris paribus, the more debtor-friendly the judge is, the more likely the outcome of the bankruptcy resolution is restructuring. This judicial debtor friendliness is linked to the continuation bias, i.e. enabling "failing businesses to linger under the protection of the court" (Morrison 2007, 381). On the contrary, bankruptcy judges may be prone to protect specific types of creditors instead of the debtor. Indeed, recent empirical literature finds that, depending on the jurisdiction, judges may be leaning towards the interests of employees (Blazy et al. 2011), state creditors or local authorities (Lambert-Mogiliansky, Sonin, Zhuravskaya 2007), or lawyers and other bankruptcy professionals (LoPucki, Doherty 2004).
So far not only has legal and economic research in Serbia not been accompanied by empirical assessment, but there have been only a handful of studies that examined judiciary bias and/or legal uncertainty (Begović 2016;Mojašević, Nikolić 2018). 2 Anecdotal evidence that support these views is obviously inadequate, and more reliable empirical studies are needed to examine such assertions. To this end, the paper contributes to a better understanding of judicial decision making in a number of ways.
First, the paper contributes to the literature examining how judicial bias affects bankruptcy outcomes and provides a novel instrument for assessing continuation and pro-creditor bias. Second, we show that the interpretation (application) of the law seems to vary considerably across bankruptcy judges and courts. Hence, the paper also provides important implications for Serbian bankruptcy practice. More specifically, the paper tests for empirical support for the claim that bankruptcy judges a) have continuation bias and b) treat vulnerable creditors differently, and c) examines the claim that there is a high level of legal insecurity in Serbian bankruptcy courts. In this respect, it contributes to the previous research that examined omissions by the first instance bankruptcy court and the issue of judicial activism (Radulović, Radović 2019), providing additional empirical evidence.
The paper is organized as follows. Section 2 presents the relevant Serbian bankruptcy framework and the role of a bankruptcy judge. Section 3 reviews several strands of relevant literature. Section 4 presents the research design and documents the results regarding several dimensions of judicial bias. Section 5 provides the conclusion.

THE ROLE OF THE JUDGE IN SERBIAN BANKRUPTCY PROCEEDINGS
Serbian bankruptcy proceedings are overseen by one of 16 commercial courts. 3 The bankruptcy court directs and controls bankruptcy proceedings and randomly appoints a bankruptcy judge who, among other things, rules on the initiation of (preliminary) insolvency proceedings, establishes whether grounds for bankruptcy proceedings exist, approves the proposed reorganisation plan (subject to approval by the requisite majority of classes of creditors) and passes the resolution on the distribution of proceeds. 4 The major reform of the Serbian Law on Bankruptcy (hereinafter: LoB) that took place some 10 years ago introduced the pre-arranged reorganization plans (PARP), as one of the main innovations. 5 The PARP is a court-supervised procedure in which the debtor submits a 3 In Serbia, commercial courts have original jurisdiction over disputes arising from commercial activities including bankruptcy proceedings where the seat of the debtor is registered. See Zakon o uređenju sudova [Law on Organization of Courts], Official Gazette of the Republic of Serbia, 116/2008, 104/2009, 1/2010, 31/2011, 78/2011, 101/2011, 101/2013, 106/2015, 40/2015, 13/2016, 108/2016(1)(4). 4 Zakon o stečaju [Law on Bankruptcy] Art. 18. Official Gazette of the Republic of Serbia, 104/2009Serbia, 104/ , 71/2012Serbia, 104/ , 83/2014Serbia, 104/ , 113/2017Serbia, 104/ , 44/2018Serbia, 104/ , 95/2018 The LoB was subsequently amended several times and many of those amendments were aimed at improving PAPR proceedings. reorganization plan which is then put to a vote, to be accepted or rejected by the debtor's creditors (Radović, Radulović 2018, 397). Besides being a relatively low-cost and reasonably efficient procedure, another advantage is that in the event that the creditors fail to vote for the plan, the PARP does not automatically trigger liquidation procedure. Ahead of the bankruptcy commencement, the debtor and key creditors engage in a voluntary outof-court work-out procedure (Vukelić et al. 2014, 5;Todorović 2016, 8). The bankruptcy judge confirms the fulfilment of formal and material conditions for commencement of bankruptcy proceedings in accordance with the PARP. 6 If approved by the simple majorities of relevant classes, the plan is confirmed by the bankruptcy judge and has the binding effect on all creditors (Radović, Radulović 2018, 397). Radović, Radulović (2018, 397-407) emphasises several key features of the PARP procedure. Namely, court intervention is fairly limited, and the interim administrator has a rather limited role. 7 The judge examines only the overall legality of the plan (including the formation of classes) and the fulfilment of procedural requirements and, at least in theory, should not play an important filtering role (Radulović 2015, 161). The law does not require judges to examine the feasibility of the plan, which is entirely left to creditors and financial advisors. Namely, the law only requires that PARP is accompanied by a statement by the auditor or a licensed bankruptcy administrator regarding the feasibility of the plan. In addition to the auditor's or administrator's opinion on the plan's feasibility, such a statement should contain the evaluation of the appropriateness of measures envisaged by the plan and of the assumptions on which the plan is based. However, "the law substitutes the judicial oversight with a series of disclosure requirements and safeguards" aiming to reduce uncertainty and information asymmetry (Radović, Radulović 2018, 397). 8 In Serbian legislature the role and discretion of bankruptcy judges has been limited for several reasons: first, the specialization and training of judges in the field of valuation and bankruptcy takes time; 6 If PARP contains minor or remedial deficiencies with respect to the fulfilment of formal and material requirements or has a technical error, the judge will order the petitioner to make the necessary corrections within the prescribed period of eight days. Because of the complexity of the actions, additional time is often required to act on the said court order, the bankruptcy judge may, at the reasoned request by the proponent, also grant additional time (Tomić 2016, 269;Radović 2017, 365). 7 The role and the influence of the bankruptcy judge differ significantly from legislation to legislation. A significant number of bankruptcy frameworks give bankruptcy judges a considerable level of discretion. For example, in the US and Germany the ultimate decision rests with a judge (e.g. bankruptcy judges may impose the plan on dissenting classes of creditors). Discretion requires a high degree of specialization and training of bankruptcy judges, so their absence is often cited as one of the key obstacles to the successful implementation of bankruptcy reform. See Ayotte, Hayong (2007, 2). 8 For a detailed discussion see Radović (2017, 354-361). second, the organization of the majority of commercial courts does not allow the narrow specialization of judges so that they deal exclusively with bankruptcy proceedings. However, in the context of PARP, the judge may play an important role as they may assess the amount of the bankruptcy creditor's disputed claim for the purposes of voting, request amendment of the plan, change the formation of creditors' classes, etc. Radulović, Radović (2019, 659) provides an illustrative example of the judicial continuation bias in bankruptcy proceedings. Namely, in several recent cases, after the PARP was not allowed by the bankruptcy judge or it was rejected by the creditors, the petitioner resubmitted the plan. Prior to the resubmission, a creditor had filed for ordinary bankruptcy proceedings, which should lead to liquidation of the debtor. Despite of a failed attempt and competing petition, in practice, bankruptcy judges found that PARP should be given preferential treatment over ordinary procedure (liquidation), i.e. resubmitted PARP should have priority over a bankruptcy petition filed by the creditor. Basically, bankruptcy judges saw PARP as a preferable solution to liquidation because of continued economic activity of the proponent, job retention, social issues, etc. However, due to the filtering failure, as a number of resubmitted plans failed, the legislature took the opposite view, finding that such court practice enabled the debtor's obstructive action aimed to prolong the resolution of financial distress, thus diminishing the creditor recovery rate. Consequently, the amendments to the Law on Bankruptcy explicitly require that priority be based on the first in time, first in right rule. 9

RELEVANT LITERATURE
Our study is related to several strands of research. First, there is an increasing body of empirical bankruptcy research on the effects of judges' characteristics. However, there is only a handful of studies that analyse judicial bias or judicial discretion in bankruptcy proceedings. Most of these studies examine judicial discretion and bias in the context of the US bankruptcy reorganization framework where judicial approval is necessary for most major actions, including the final approval of a plan of reorganization. 10 Evans (2003) was the first to categorize motions adopted by bankruptcy judges into pro-debtor and pro-creditor actions. Using a sample of 290 small closely held debtor firms from a single district in the United States, it reports that pro-creditor discretionary judicial decisions decreased the probability of reorganization but increased the probability 9 Art. 158 LoB. 10 See 11 U.S. Code Title 11 -Bankruptcy. of deviation from absolute priority. 11 Sharfman (2006) focuses on judicial bias in bankruptcy rulings using data on a small number of disputes over the value of debtor-retained collateral in the United States. The main finding is that judges are generally pro-debtor as they "tend to favor loss-averse debtors over gain seeking secured creditors" (Sharfman 2005, 399). Morrison (2007) used a sample of 95 small business reorganizations to find that despite widespread perception practice in the US, bankruptcy courts exhibit no systematic continuation bias in favour of saving nonviable businesses, but that "judges act as if they are seeking out and preserving going-concern surplus" ( Morrison 2007, 411). Chang, Schoar (2013) uses data on Chapter 11 filings for 7,824 private companies in the US between 1989 and 2006. This novel approach uses the random assignment of bankruptcy cases as a natural experiment and heterogeneity in judges' interpretation of the law. The findings include significant differences across judges in the propensity to consistently grant (or deny) specific key motions (e.g. use of cash collateral or lifting the automatic stay). Judge specific fixed effects were obtained, that were used as a prodebtor index and show that pro-debtor judges have worse firm outcomes in reorganization cases i.e. Chapter 11 works more efficiently when presided by a pro-creditor judge. More recently, He, Yu, Wu (2020) used US commercial bankruptcy data to determine that judicial bias varies across states and bankruptcy courts. Other empirical bankruptcy literature mainly deals with the judicial attributes such as experience (Iverson et al. 2018), ideological preferences (Nash, Pardo 2012), time constraints and workload (Iverson 2017), or forum-shopping on bankruptcy outcomes. Using a similar approach as Chang, Schoar (2013), Iverson et al. (2018) finds that judges' experience significantly influences the speed of ruling and managing of corporate restructuring cases.
Outside of the US, empirical bankruptcy literature that examines issues related to judicial bias is scarce. Lambert-Mogiliansky, Sonin, Zhuravskaya (2007) used a sample of Russian firms to find that Russian commercial courts were biased in favour of regional authorities. Blazy et al. (2011) examined a sample of bankruptcy files on French SMEs, showing that employment protection is the key element driving the courts' behaviour (Blazy et al. 2011, 136).
Second, several theoretical papers in law and economics and financial economics literature analyse the role of judicial discretion and bias. Ayotte, Yun (2007) presented a model of optimal bankruptcy law considering the expertise of judges and the quality of contract enforcement. The optimal degree of "creditor friendliness" in the bankruptcy law decreases as judicial ability to recognize firm quality increases, i.e. in order to be effective, debtor-friendly law requires judicial expertise. Gennaioli, Rossi (2010) shows that judicial bias may not be random, due to career concerns of bankruptcy judges. To attract prospective cases, bankruptcy judges tend to "over-reorganize" debtors, signalling their prodebtor inclination. Similar to Ayotte, Yun (2007), it shows that "creditorfriendly" law is crucial for enhancing judicial incentives to resolve a case in an efficient manner.
Finally, another strand of research, closely related to our research design, deals with the behavioural analysis of judicial decision making (see, for example, Guthrie, Rachlinski, Wistrich 2001;Teichman, Zamir 2014). Rachlinski, Guthrie, Wistrich (2006 examined the impact of six phenomena on bankruptcy judge decision making in order to compare them to generalist judges. They presented 113 US bankruptcy judges (approximately one-third of all bankruptcy judges in the US) with a set of hypothetical cases to examine the extent of judges' reliance on common heuristics. On the one side, they found that anchoring (relying on the initial available value to make an estimate) and framing (influencing the decision by the way they are framed through different wordings, settings, and situations) affected judges' assessments and had the similar effect on bankruptcy judges and generalists. On the other side, unlike irrelevant anchor and framing effects, omission bias (reacting more strongly to harmful actions than to harmful inactions), the debtor's race, the debtor's apology, and terror management did not affect bankruptcy judges. Interestingly, there was no evidence found that more experienced judges performed better than less experienced ones.

RESEARCH DESIGN AND RESULTS
The study was conducted during September and October of 2019. Data was collected with the support of the Commercial Appellate Court (CAC). Participating judges had to make decisions in several hypothetical cases. The research experiment was carefully explained, and a pilot questionnaire used to test participants' understanding. A total of 63 judges from 16 Commercial Courts, Commercial Appellate Court, and Supreme Court of Cassation (SCC), provided answers to the questionnaire. Only bankruptcy judges from first instance courts and judges that decide bankruptcy-related cases in the CAC and SCC were asked to participate in the study. The sample represented almost the entire population of judges that preside over bankruptcy-related cases. Judges were asked to provide answers and comment individually. Furthermore, to provide incentives to engage in the experiment (and per request of the CAC), participants provided their answers anonymously i.e. judges were assured that their names or other identifying information were not recorded. This limited the inquiry to some extent, since apart from the region and specialization, we were not able to obtain additional data on judges' experience, gender, law school, etc. Figure 1 shows the distribution of participating judges. The number of participating judges per court depends on the court organisation. Namely, while large courts usually have bankruptcy judges dedicated solely to bankruptcy cases (e.g. Belgrade, Novi Sad, Kragujevac, etc.), some smaller courts do not have judges that preside exclusively in bankruptcy cases (e.g. Užice, Pančevo, etc.). In addition, in several courts it is difficult to make a clear distinction between "specialized" and "generalist" judges (e.g. a half of judges in a particular commercial court are dedicated solely to bankruptcy cases, while others are not or judges were only recently reallocated from the bankruptcy to litigation panel). Each of the bankruptcy judges was provided a questionnaire that contained three hypothetical cases. The judges were assigned to groups randomly. The first case assessed the presence of continuation (pro-debtor) bias. The second case examined the presence of specific pro-creditor bias. Finally, the third case looked at whether judges express judicial activism when deciding cases and whether this is affected by social concerns. Each of these cases is explained below, along with the results.

Case 1: Continuation Bias
To test for the presence of continuation bias, we constructed a problem in which judges had to decide on the opening of bankruptcy proceedings. In preparing this case, we took advantage of recent cases involving debtors already in compulsory liquidation proceedings. We used these circumstances as a sort of natural experiment that we replicated in a questionnaire. 12 The first case provided judges with the following context. Compulsory liquidation proceedings were initiated against the debtor (company) because it failed to submit the annual financial statements to the competent registry by the end of the previous financial year, for two consecutive financial years. 13 After the compulsory liquidation proceedings had been initiated, the debtor filed a motion for initiating bankruptcy proceedings in accordance with a pre-arranged reorganization plan. The PARP contained all of the essential elements enumerated in Art. 156 of the LoB, with the exception of the annual financial statements for the previous three years. However, the plan proponent had also provided the auditor's opinion on the state of the books of accounts, which states that the information contained in the plan is factual and that PARP is viable. Based on this background, the judges were asked whether they would reject the proposed PARP and if so on what grounds. Table 1 shows that the majority of bankruptcy judges would reject the plan. However, the majority (54%) is rather slim, as almost 46% of judges would either accept the plan (29%) or provide the debtor with a second chance (17%), i.e. allow the debtor to provide additional documentation, or not render the decision and wait for the resolution of the compulsory liquidation proceedings. The results demonstrate a very high heterogeneity of answers. This heterogeneity may be observed between commercial courts, but also within commercial courts. Figure 2 reveals that both types of heterogeneity are present. Out of 15 first instance courts (one court with only one response was not included), only in six courts bankruptcy judges provide unison and cohesive answer. In the remaining nine first-instance courts, as well as in the higher instance courts, judges provided substantively different responses. With respect to between courts' differences, four commercial courts seem to be biased towards continuation (i.e. pro-debtor proxied by the acceptance decision) and six towards creditor (with only four courts being evidently pro-creditor, proxied by the rejection decision). The remaining six courts had significant within heterogeneity, so it is not possible to classify them either as having pro-debtor or pro-creditor bias. This makes allocation of cases to the particular judge a significant determinant for how the debtors and creditors will be treated and how the law will be applied. They also provide support for the recent attempts of forum shopping, where debtors changed the seat of their companies in order to be able to file for bankruptcy in different jurisdiction.

Case 2: Specific Pro-Creditor Bias
The second case examined whether characteristics of creditors and the size of the claim affect judges' decisions. Unbeknownst to the respondents, they were randomly distributed to either to a control group or to a treatment group. To examine whether judges are more prone to decide in favour of particular type of creditors, we presented judges with the case in which in the PARP proceedings a creditor, either an entrepreneur (control group) or a public enterprise (treatment group), had disputed claims against the debtor. We asked judges to state how they would assess the value of disputed claims for the purpose of voting. To examine the effect of the size of the claim, the disputed sum significantly differed in the two versions of a hypothetical case (the sum in the control group was set to be much higher). 14 Again, in creating the case our starting point was a legal ambiguity. Namely, amendments to the 2014 LoB stipulated that the assessment of the likelihood of claims for voting purposes may be made solely by an authorized professional (appraiser). However, amendments to the LoB in 2017 removed this norm, creating a dilemma whether a bankruptcy judge alone should assess the likelihood for the purpose of voting. 15 We 14 Additional details of the second hypothetical case are as follows. The creditor's first claim against the debtor originates from damages and amounts to EUR 100,000 (control group) or EUR 5,000 (treatment group). The disputed sum represents the value of real estate that the debtor has misappropriated. Consequently, the creditor filed claim for damages with the civil court, and a preliminary hearing was scheduled. Additionally, the creditor also seeks a sum of EUR 100,000.00 from the debtor, backed by a first instance judgement rendered. The said first instance judgement partially upheld the claim to the amount of EUR 50,000.00, while rejecting it for the remaining EUR 50,000.00. From the judgement's reasoning, it is apparent that the claim has become unenforceable due to the statute of limitations, since said claim became due three years before it was filed. Being that, according to the court of first instance, the statute of limitations that applies to companies (referred to in Article 374 of the Law on Obligations) should apply accordingly to entrepreneurs, this claim was considered unenforceable due to the statute of limitations. The PARP acknowledged said creditor's claim to the amount of EUR 50.000,00, based on the aforementioned first instance judgement, while listing the claim of EUR 100.000,00 for damages and the remaining EUR 50.000,00 (considered obsolete and unenforceable by the first instance judgement) as disputable. 15 Art. 160 LoB.
hypothesise that judges would be prone to protect the entrepreneur with a large disputed claim, rather than public enterprise with a small disputed claim. To avoid random value answers, we avoided asking judges to provide the particular sum and instead asked two questions.
First, we asked judges whether they would perform the evaluation themselves or they would appoint a professional (appraiser) to do the assessment (Case 2a). Note that the problem explicitly states that the creditor that filed said claim had proven with certainty the existence of damages in the amount of real estate's market value.
As the total size of the data set is n=62, we perform the Freeman-Halton extension of the Fisher exact probability test for a 3x2 contingency table. 16 The null hypothesis is that these two categories are not different. The test shows that there is no statistically significant difference with respect to the creditors' type and size of the claim (p= 0.204).  To test for robustness of our findings, we also combined all the responses where judges stated that they would engage an appraiser (Table  3). As we hypothesise that judges are more likely to appoint an appraiser when the size of claim is large, we use a one-sided Fisher exact probability test. Again, the test rejects our hypothesis as p=0.162, and although data showed that judges were more likely to appoint appraisers when the size of the claim is larger, the results were not statistically significant. Second, we also examined how a bankruptcy judge would act with respect to the non-final first instance judgment on a disputed monetary claim that is found to be groundless (Case 2b). More specifically, the judges were asked to provide an answer to whether they would base their decision only by looking at the order of the judgment (in which the court has rejected said claim), or they would assess the merits of the claim by analysing the judgement's reasoning. We assessed whether judges are either pro-state or pro-entrepreneur, depending on whether they also looked at the reasoning, trying to find additional grounds for revaluation of the disputed monetary claim. Table 4 shows that in both the control and treatment group the majority of judges look at the judgement and the reasoning. While the share of judges that look at both judgement and reasoning is higher in the case of entrepreneurs, we again do not find a statistically significant difference. Results, however, reveal another potentially problematic aspect of bankruptcy judges' reasoning. It is objectively expected that due to the small amount the bankruptcy judge should not indulge in the detailed analysis of the correctness of the legal position taken in the reasoning of the first instance basic court. Our results reveal that this is not the case. Radulović, Radović (2019) identified several examples of judicial activism in recent Serbian bankruptcy practice. Under judicial activism in this paper, we refer to bankruptcy rulings that are based on personal opinion, rather than on existing bankruptcy law. 17 We thus examine to what extent judges behave more like legislators than like judges. While the term "judicial activism" is traditionally employed in the context of appellate courts (Green, Roiphe 2019, 365), in the case of Serbian bankruptcy courts there is no such exclusivity, as first instance judges may also use and abuse discretion in how they conduct hearings and rule decisions.

Case 3: Social Concerns and Judicial Activism
We presented judges with a case in which a petitioner had filed a motion for initiating bankruptcy proceedings in accordance with a PARP. We altered the case to test for the effects of both social concerns and judicial activism. To control for the social concerns and ties to the local community, judges were again randomly distributed to a control and a treatment group. In the control group the debtor was a micro-enterprise that employs 10-15 persons, and in the treatment group the petitioner was a large company employing 1,500 workers. Unlike the first case, where due to vague circumstances judges could decide both ways, in the third case the law clearly suggests that judges should confirm the plan. Hence, any rejection of confirmation of the plan could be interpreted as an instance of judicial activism.
The third case provided judges with the following context. The PARP petitioner stated that the reasons for bankruptcy were permanent inability to make payments as well as excessive indebtedness. The debtor's business activity is the production of specialised packaging. During the previous period, the debtor did business only with company A (buyer). Due to problems, the buyer needed less packaging from the petitioner, which left the petitioner insolvent. The PARP, as a part of the financial projection, stated that the business of the proponent's main client, company A, was s of paramount importance. In order to increase business revenue, the petitioner also continuously made efforts to gain other clients as customers. The debtor stated that, should they gain new buyers, a more favourable settlement for creditors could be reached, in comparison to bankruptcy. However, the debtor did not submit any letter of intent or preliminary contract, nor any other evidence that would make the debtor's cooperation with potential partners credible. These circumstances make the debtor's future business extremely uncertain. Notwithstanding these facts, the majority of all creditor classes have voted in favour of the plan. We asked judges to make a decision regarding the confirmation of the plan. Table 5 and Figure 4 show that responses were almost identical in both groups.  The Fisher exact test clearly rejects the null hypothesis. Although we cannot relate judicial activism to social concerns, almost 18% of judges express judicial activism either by promptly rejecting the adopted plan or by stating that further actions would be required before confirming the plan. Interestingly, the share of activist judges is similar both at the first instance and appellate courts (both CAC and SCC). While judicial activism in this particular hypothetical case means that bankruptcy judges strive to confirm only feasible plans, i.e. prevent inefficient reorganisation, which may be a positive step in terms of signalling judicial skills, in the context of the current Serbian bankruptcy framework this only contributes to legal uncertainty (at least until the LoB is harmonized with the recently adopted EU Directive on Restructuring and Insolvency). 18

CONCLUSION
Our study shows that there are some good and some bad news for the Serbian bankruptcy judiciary. judges did resist the continuation bias. However, the majority is rather slim, as only 54% of judges rejected submission of the plan by the dubious petitioner. Second, we did not find sufficient statistical evidence to claim that judges behave differently depending on the type of creditor. In other words, judges are unaffected by the type of the debtors and were indifferent to social concerns proxied by the number of employees. We think that unbiasedness is rather good news.
At the same time, we observed a very high level of heterogeneity among Serbian bankruptcy judges in their interpretation of the bankruptcy laws. Additionally, the judges paid too much attention to irrelevant issues. Such practices by bankruptcy judges are making the process even more costly and time-consuming, further wasting scarce judicial resources. Finally, considering that the whole process of reorganization is based on the out-of-court negotiation between creditors and the manifestation of their will exemplified by the plan approval, it is rather questionable whether the will, clearly expressed by the majority creditors, should be affected by the judge. The share of judges that go beyond what is provided for in the law is close to 20%, i.e. roughly there is a one in five chance that the court may interfere and examine the feasibility of the plan in spite of the fact that the very same plan was approved by the creditors. This type of judicial activism further amplifies the legal uncertainty of the bankruptcy process in Serbia. Our results clearly show that both jurisdiction and the allocation of cases to a particular judge represent significant determinants of how the debtors and creditors will be treated and how the bankruptcy law will be applied.