i
Clarity of cartel law and the deterrence effect
A theoretical model and a Dutch case study
Anna den Boer -10656324 Anna.denboer@student.uva.nl
MSc. Economics - Industrial organization, regulation and competition policy Supervisor: Prof. Dr. B.E. Baarsma
ii Abstract
This paper investigates what type of firms do not find cartel law clear, and its effect on deterrence. First, a theoretical model is developed that aims to capture decisions and behavior of firms regarding participating in an agreement or not. It predicts that there are two undesirable consequences of firms not finding cartel law clear; there will be more deterrence from legal behavior and less deterrence from illegal behavior. Second, a case study of The Netherlands is performed, using data of a survey of Van der Noll et al. (2010). The case study shows that firms that do not find the law clear give, on average, a lower probability of deterrence than firms that do find the law clear when suspecting the agreement is illegal. This is in line with the predictions of the theoretical model. Having a compliance officer and having more employees positively influences clarity of the law as well as probability of deterrence. The final part of this thesis looks at the determinants of probability of deterrence. It finds that a personal fine can be a useful tool to increase probability of deterrence.
iii Preface
‘If you don’t know where you’re going, any road will get you there’
- Lewis Carroll, Alice in Wonderland
For an unknown reason, writing my thesis was something I dreaded for a long time. Fortunately, it was only occasionally a little frustrating and stressful. It was mostly satisfying, fun and educational. After these 3 months I can say that I really enjoy working with Stata, that I know all about cartel law and what firms do not find the law clear, and that I even almost feel like an economist. The quote this preface starts with perfectly describes how I started writing my thesis. After I knew that my topic was going to be clarity of cartel law and the deterrence effect, the hardest part was to define the research questions. This research has never been done before, so anything could be researched. Along the way of writing this thesis, the results showed that more and more side-topics were interesting. This only increased my enthusiasm for the topic, but it also did not make it easy to decide what I wanted to publish in my thesis. I believe a book could be written about this topic. To give some statistics; in total I did 234 T-tests and 39 regressions, and I guess my do-files are around 23 pages. But, like in movies, some of the best parts never make it to the screen because a story cannot last too long.
I am a little sad that with the finishing of this thesis my life as a student will officially end, but I am very excited to start with something new and to further dive into the world of cartels, competition and regulation at the European Commission, my next adventure.
To conclude, I would like to thank SEO Economic Research for the data they gave me and for the opportunity to write my thesis in an environment that was full of great economic ideas. I especially want to thank Barbara Baarsma, my supervisor, and Rob van der Noll for their ideas, feedback and support. Without their input I could never have written this thesis.
iv Table of content 1. Introduction ... 1 2. Literature review ... 3 3. Data ... 6 3.1. The Survey ... 6
3.2. Representativeness of the Data ... 8
3.3. Weaknesses ... 8
4. Theoretical model & hypotheses ... 9
4.1. Firms that find the law clear ... 9
4.2. Uncertainty for firms that do not find the law clear ... 11
4.3. Comparing the situations ... 13
4.4. Hypotheses ... 15
5. Case study – The Netherlands... 18
5.1. Probability of deterrence ... 18
5.2. Determinants of clarity and effects on probability of deterrence ... 19
5.3. Distinguishing between sectors ... 25
5.4. Vignettes ... 31
6. Final remarks ... 38
References ... 41
Appendix A ... 45
1 1. Introduction
A main reason for governments to introduce anti-cartel law is to deter firms from participating in illegal cartels1 (Baker, 2001; Gordon & Squires, 2008). By punishing firms, authorities seek to decrease the incentive for other firms to engage in illegal behavior. Many studies have researched this deterrence effect of cartel legislation, both theoretically and empirically. Most of those papers found evidence for a positive deterrence effect of anti-cartel law on the number of cartels.
Previous research regarding deterrence and compliance with cartel law relies on the assumption that firms find the law clear, and therefore know whether behavior is illegal. However, in reality this seems to not be the case. Parker (2013) interviewed almost 600 Australian compliance officers of firms. Their understanding differed greatly regarding the illegality of participating in agreements and its associated punishments. In the Netherlands this occurs as well. According to a survey of Van der Noll et al. (2010), distributed among firms in the Netherlands in 2010, almost a third2 of the firms did not find it clear when the ACM3 considers agreements and conduct to be prohibited.
A possible explanation for the fact that firms in the Netherlands do not find the cartel law clear, is that not all agreements restrict competition, so not all agreements are illegal. Also, there are several exemptions to the cartel law. According to the Dutch Competition Law, article 6, paragraph 1, ‘Agreements between undertakings, decisions by associations of undertakings and concerted practices between undertakings which have as their object or effect the restriction or distortion of competition on the Dutch market or part of the Dutch market are illegal’. However, article 6, paragraph 3 describes cases in which an agreement does not need to be illegal, for example when it does not eliminate competition and it improves production or distribution4. Because of economies of scale and synergies that arise from firms working together and pro-competitive effects an agreement can have, the benefits of an agreement can outweigh the negative effects. For a firm, this can make it unclear when behavior is illegal or not. Article 7
1 For this paper, I will use the definition of cartels as it is formed by the European Commission; ‘A cartel is a group of similar, independent companies which join together to fix prices, to limit production or to share markets or customers between them’(http://ec.europa.eu/competetion/ cartels/overview/index_en.html). Cartels are a form of collusion. When it is a cooperative agreement between firms, as is a cartel, it is explicit collusion. When firms act in a non-cooperative way it is tacit collusion (Motta, 2004).
2 The hypothesis that every firm finds the cartel law clear can be rejected with 99% confidence, found with a T-test. Only 70% of the firms found the cartel law clear.
3 The ACM is the Autoriteit Consument & Markt, the Dutch competition authority.
4 Article 6, paragraph 3; ‘Paragraph (1) shall not apply to agreements, decisions and concerted practices which
contribute to the improvement of production or distribution, or to the promotion of technical or economic progress, while allowing consumers a fair share of the resulting benefits, and which do not: a. impose any restrictions on the undertakings concerned, ones that are not dispensable to the attainment of these objectives, or b. afford such undertakings the possibility of eliminating competition in respect of a substantial part of the products and services in question.’
2
describes the exemptions from cartel law. This is called the ‘bagatelvrijstelling’ which indicates that companies that do not have a significant market share or revenue above a threshold will not be prosecuted for making price agreements. This makes it more difficult for a firm to completely understand the law.
What is the consequence of the fact that not all firms find the cartel law clear? Is there a difference in deterrence between firms that do and firms that do not find the law clear? What are the determinants of whether a firm finds the cartel law clear? The goal of this thesis is not to see whether there is a deterrence effect in the Netherlands; this has already been done by Van der Noll et al. (2010). This thesis uses data of the survey of Van der Noll et al. to find what type of firms do not find the cartel law clear, and how this affects deterrence. This will be done in two steps. First, a theoretical model will be developed that aims to capture decisions and behavior of firms regarding participating in an agreement or not. Second, a case study will be conducted based on the available data to see whether the perceived clarity of the law affects deterrence in the Netherlands.
The set-up of this thesis is as follows. Chapter 2 describes past empirical research about the deterrence effect and the determinants of the deterrence effect. The results about firms that do not find cartel law clear from Parker (2013) and Van der Noll et al. (2010) will also be discussed, as well as some possible effects of firms not finding cartel law clear. Chapter 3 will describe the data used for the empirical part of this thesis. This chapter will also discuss the representativeness of the data and weaknesses of using survey data. In chapter 4 a theoretical model will be developed. This theoretical model will describe the decisions firms face when they have to decide whether to participate in an agreement or not, separating between firms that do find cartel law clear and that do not find the law clear. It will be a one-stage game with and without uncertainty. It predicts the consequences for society when not all firms find the cartel law clear. In the second part of chapter 4 the hypotheses will be formulated. Chapter 5 will consist of a case study of the Netherlands, where the results for the hypotheses will be presented. In chapter 6 conclusions will be drawn and discussed. Also, the weaknesses of the analyses will be discussed as well as possible further research.
3 2. Literature review
Becker (1968), as well as other researchers, argues that the purpose of a punishment is to deter individuals from illegal behavior. He believes that this can be done by deterrent incentives, such as probability of punishment and sanctions. Since the deterrence effect is one of the main purposes of cartel law, it is a subject that has been researched many times. It is costly to detect and prosecute participants of cartels, since evidence needs to be found to be able to punish participants. If the enforcement of the law does not deter firms from participating in illegal agreements, the costs are not compensated for, which harms society (Buccirossi et al., 2009). Two topics are researched most; several papers measure whether the deterrence effect exists, whereas the other topic often researched is what type of punishments have an influence on the deterrence effect. Both topics will be discussed in this thesis.
When looking at papers that investigated whether the deterrence effect exists, there are papers that did not find a positive deterrence effect. Crandall and Winston (2003) and Konings et al. (2001) did not find evidence that enforcing cartel legislation deters firms from illegal conduct. Veljanovski (2013) used a survey in the United Kingdom and found no robust evidence for a positive deterrence effect. Schildberg-Hörisch and Strassmair (2012) conducted an experiment in which they found that the deterrence effect does not increase when sanctions increase. Most papers, however, did find a positive deterrence effect. These results have been found using several different research methods. Feinberg (1984), Miller (2009), Hylton and Deng (2007), Block et al. (1981) and Buccirossi et al. (2013) used empirical methods to find the deterrence effect. Another example that used and empirical method is Clarke and Evenett (2003). They tried to find the deterrence effect in the vitamins cartel that operated from 1989 to 1999 and involved producers in many countries across Europe, the United States and Japan. Their research differed from past research because they investigated the price increases of the cartel compared over countries with and without strong cartel enforcement. They found that the prices in countries without an anti-cartel enforcement regime rose more than prices in countries with an anti-cartel enforcement regime, which indicates a positive deterrence effect.
Feinberg (1985), Beckenstein and Gabel (1983), Parker and Nielsen (2005) and Hüschelrath et al. (2011) used surveys for their research. Gordon and Squires (2008) discussed the research of Deloitte in 2007 about the deterrence effect in the United Kingdom. Deloitte surveyed around 200 senior competition lawyers and 200 companies with over 200 employees in the UK. With the data from the respondents they calculated ratios of agreements abandoned or modified to agreements that resulted in a Competition Act 1998 infringement. They found a positive deterrence effect. The ratio for companies was 16 to 1; 16 abandoned or modified agreements for each infringement of the Competition Act. They also found that over the period 2000-2007 a total of 202 companies modified or abandoned 144 agreements and 126
4
cartels, which equals to 1.3 cases per respondent. Van der Noll et al. (2010) used a survey in the Netherlands for firms and advisors about the deterrence effect. They found that every firm abandons or modifies an agreement, on average, once every 5 years, which is similar to the findings of Gordon and Squires and indicates a positive deterrence effect.
Aside from researching the actual effect, several papers investigate the determinants of the deterrence effect. Beckenstein & Gabel (1986), West (2012), Morgan (2009), Harrington (2014), Bigoni et al. (2014), Bageri et al. (2012) and Katsoulacos and Ulph (2013) use several different methods for this research. Buccirossi et al. (2009) claim that deterrence is affected by three main determinants; the costs (fines) of being detected when engaging in illegal conduct, the probability of being detected and prosecuted, and the probability of being wrongly prosecuted. Van der Noll et al. (2010) used a conjoint analysis to test the effects of a personal fine, a company fine, a leniency program, whether the firm is on the agenda of the ACM, and the publicity after being detected, on the probability of deterrence. The factors with a significant effect were the personal fine and the company fine. In other words, the probability of deterrence would rise if the personal fine or the company fine would rise. Morgan (2009) found that fines do have a positive deterrence effect, but that it is not enough to completely deter firms from participating in cartels. Morgan argues that the leniency program that is adopted in most countries has a positive deterrence effect. Beckenstein & Gabel (1983) found that personal penalties were much more effective than penalties for a firm. Feinberg (1985) found similar results. Bigoni et al. (2014) conducted an experiment and found that high fines are the main determinants of deterrence but that a leniency program has a positive influence on deterrence as well.
All previous papers rely on the assumption that firms find the cartel law clear and therefore know whether their behavior is illegal or not, which allows them to consciously make a decision on whether to enter an agreement. However, it is likely that not all firms find the law clear, as explained in chapter 1. Parker (2013) found that there is a large variation in knowledge of the anti-cartel laws in Australia. Participants of the survey were presented a hypothetical situation in which a firm is agreeing on prices with its competitors. The respondents were asked whether they knew if the conduct was a civil contravention (in Australia, civil contravention has been the legal status of cartel conduct since 1974). If they thought it was illegal, they were asked whether it was a criminal offence5 and what sanctions were placed upon the illegal conduct. Table 2.1 presents the answers to these questions. As can be seen, 18% argues that the price agreement is not against the law and 19% is not sure about this. Even among firms that agree that the conduct is against the law, there is variation about whether it was a criminal offence. The thoughts on
5 Civil contravention and criminal offence are two classifications of offenses. A civil contravention is a minor offense, whereas a criminal offence is the most serious type of offense.
5
the penalties for the conduct varied greatly among respondents as well. This indicates that a great part of the firms does not know exactly when behavior is illegal.
Van der Noll et al. (2010) found a similar outcome in their paper, as indicated in the introduction. They distributed a survey among firms in the Netherlands, who were asked the question ‘do you think it is clear when the ACM considers agreements and conduct to be prohibited?’ Almost a third of the respondents answered ‘no’ to this question. Since the survey might incentivize respondents to give the socially preferable answer, 30% is expected to be a lower bound.
But what could be problems for society if not all firms find the law clear? Buccirossi et al. (2009) argue that deterrence should only occur when actions are illegal and anti-competitive, and therefore bad for consumers. The optimal level of deterrence does not stop a firm from behavior that is not anti-competitive. If a firm abandons behavior that is not anti-competitive it is called over-deterrence. When firms find the law clear and know whether behavior is illegal, over-deterrence is unlikely to happen. However, when firms are not sure whether behavior is illegal, it is possible that they participate in an illegal agreement or that they deter from a legal agreement, which are both undesirable outcomes.
It is difficult to measure the deterrence effect, since one has to measure infringements that are not committed (Veljanovski, 2013) and thus not reported. It is even more difficult, if not impossible, to measure over-deterrence. Firms will not expect behavior to be legal when they deter from doing it so they will not report it as over-deterrence. The next chapter will explain what type of data will be used to give a measurement of the difference in deterrence between firms that do find cartel law clear and firms that do not.
Agreeing on prices with competitors is.. Percentage of respondents (N=567)
A criminal offence 42
Against the law but not criminal 11
Against the law but not sure if criminal 9
Not sure whether against the law 19
Not against the law 18
Table 2.1: Answers of respondents to the scenario presented in the survey of Parker (2013), it indicates a great variation in knowledge of the law across Australian compliance officers
6
We are going to present you with six hypothetical situations. The following applies in each case. You have discovered that your company has entered into a price fixing agreement with a competitor. You strongly suspect that this agreement is not compatible with the cartel prohibition. The sales department has informed you that your business unit’s annual turnover is 20% higher as a result of this price fixing agreement. Each situation describes a hypothetical regulatory regime. Please indicate how probable it is in each of these situations that you would give instructions to terminate the price fixing agreement. You can express probability by assigning it a score on a scale of 1 to 10, 1 being the lowest probability and 10 the highest.
3. Data
3.1 The survey
The data that will be used for the case study of this thesis comes from an online survey that SEO Economic Research6 developed and used for the paper of Van der Noll et al. (2010). The survey was developed with input from TwijnstraGudde (2005) and Deloitte (2007), that both researched the deterrence effect using a survey. The survey was sent (in slightly different versions) to legal advisors and to companies. For this thesis only the survey for the companies is used. It was sent to 4.381 companies, to either the CEO or the internal company lawyer. Only firms with over 100 employees were invited to participate. The data covers the period from 2005 to 2010, as respondents were asked to take in mind this period when answering the questions. 512 companies participated in the survey (of which 342 answered all questions), which is a response rate of 11%.
The survey consisted of three parts. In the first part questions about mergers were asked, which will not be used for this thesis. The second part contained questions about cartels and price agreements. The third part of the survey was a conjoint analysis7. For the conjoint analysis, participants were presented several hypothetical situations of punishments for a price-agreement, which they had to classify. In Box 3.1 the exact task of the respondents is explained.
When firms were facing the scenario presented in the survey, it is important to notice that they were told that they suspected the conduct to be illegal, so the measurement of deterrence gives the willingness of firms to stop conduct that is expected to be illegal. Therefore, for society, it is best if firms give a high probability of deterrence because this would indicate that firms are not willing to participate in anti-competitive agreements.
6 SEO Economic Research is a company affiliated to the University of Amsterdam that carries out applied and independent economic research.
7 The conjoint analysis is a good way to measure the importance of determinants of the deterrence effect. Since the respondents are shown scenarios, it is a good indicator of how important the determinants are relative to each other. It is also a good way to test whether firms that do not find the law clear act different than firms that do find it clear (Van der Noll et al., 2010).
7
The scenarios were based on attributes, as presented in Table 3.1. For each type of punishment there were several ‘values’ it could get; from value 1 (the best for the firm) to value 4 (the worst for the firm). There were a total of 3 x 4 x 2 x 3 x 3 = 216 different scenarios. Every firm was randomly presented 6 scenarios - 2 at a time - and a total of 1456 scenarios were ranked. For every pair of scenarios that was presented to a respondent it was made sure that they did not diverge too much, so a respondent had to rank them based on their value of the different attributes8. An advantage of using the conjoint analysis is that it gives both the probabilities of a firm deterring in a certain situation and the ranking of the attributes.With this type of analysis it is possible to find which attributes have a larger effect on deterrence than others, and this can be compared between different types of firms. Another advantage is that it minimizes factors that can distort the results, such as strategic responses and social bias.
For this thesis it is useful to compare the probabilities of deterrence that firms assigned to scenarios. Those probabilities do not give an exact measure of deterrence, but they give the relative deterrence effect when comparing two groups. When the outcome would be that the probability of deterrence is higher for firms that do not find the law clear, it means that they deter from behavior that is illegal. When the outcome is that the probability of deterrence is lower for firms that do not find cartel law clear, it indicates that those firms are less likely to stop with illegal conduct.
To be able to see if firms that do not find the law clear behave differently regarding deterrence, two separate analyses will be done in this thesis;
1) Characteristics of firms, like having a compliance officer, firm size and sector, will be used to see what type of firms do not find cartel law clear,
8 The exact way of assigning the scenarios to respondents is explained in Van der Noll et al. (2010).
Value 1 Value 2 Value 3 Value 4
Personal Fine None €450,000 €650,000
Company Fine 2% annual turnover 10% annual turnover
20% annual turnover
30% annual turnover The sector is: Not listed in the
ACM agenda
Listed in the ACM agenda
Leniency Company expects
to be the first applicant Company expects to be the second or subsequent applicant Not possible
Publicity Only on the ACM
website
ACM website and trade journals
All newspapers and television news
8
2) The answers to the conjoint analysis will be used to see how clarity of the law and the other characteristics influence probability of deterrence.
3.2 Representativeness of the data
Van der Noll et al. (2010) tested whether the data collected with the survey is representative for the population. Using descriptive data as well as statistical analysis they tested whether the sample was representative for the total population of firms. It was tested for four different factors; geographical spread, sector, company size and company age. The statistical tests for testing representativeness were mostly significant, which indirectly this could suggest that some of the factors tested are underrepresented or overrepresented. For example, the sectors financial institutions and health care appear to be overrepresented. This is hardly surprising, because firms in those sectors more often deal with mergers, or it is a more discussed theme (Van der Noll et al., 2010). When looking at the descriptive statistics (percentage distribution and graphs) there is no cause for concern about the representativeness of the sample. All the differences are small, which indicates that the sample is representative (Van der Noll et al., 2010)9.
3.3 Weaknesses
There are several weaknesses in using a survey for an empirical analysis. A survey does not give revealed preferences, it merely gives stated preferences. Using only stated preferences has several disadvantages. Because firms are presented hypothetical situations may not behave the way they would in real life, which is a hypothetical bias. Also, since respondents give their view, they may have the incentive to give socially optimal answers or might interpret the question differently. Another disadvantage is that there can be a strategic bias in the answers. This can happen when firms try to influence the results of the survey, which is more likely to happen when questions are asked directly. When questions are asked indirectly, for example through a conjoint analysis, strategic bias is less likely to occur.
If you want to investigate the deterrence effect in the way this thesis does, only stated preferences can be used. The actual deterrence effect is not directly measurable since there is no revealed data about it. When using only stated preferences, the conjoint analysis is the most preferable method.
9 4. Theoretical model and hypotheses
In this chapter a theoretical model will be developed that reflects the choice that firms face when they decide to participate in an agreement or not. The first part of the model is based on Van der Noll et al. (2010). The base of this model is a one-stage game, with and without uncertainty. After the model is developed, the hypotheses of this thesis will be formulated.
4.1 Firms that find the law clear
Firms that have to make a decision about whether or not to engage in a price agreement or other type of cartel, face the decision tree presented in Fig. 4.1. First, a firm has an initiative to meet with another firm for an agreement. This agreement can either be illegal or legal, which is determined by nature. An assumption of this model is that firms that find the law clear know if their agreement is illegal or not. The next step is to decide whether to enter the agreement, or whether to deter from it. If a firm decides to deter from the agreement it will get no extra pay-off. If a firm decides to start, it will get a pay-off from participating in the cartel. The pay-off from the agreement when the behavior is illegal is 𝑅𝐼 and the
pay-off is 𝑅𝐿 if the behavior is legal. We can assume 𝑅𝐼> 𝑅𝐿 because anti-competitive behavior will come at
the costs of consumers, therefore it is possible for producers to gain more. When firms enter an agreement, the ACM will either detect the cartel or not. When behavior is illegal the ACM detects the agreement with probability 𝑞. It can occur that the ACM does not detect an agreement even though it is illegal (with probability 1 − 𝑞), which is a type II error. It can also happen that the ACM detects a cartel and terminates it whilst it is not illegal. This is a type I error, which occurs with probability 𝑝. An assumption made in this thesis is that 𝑞 > 𝑝, i.e. that the probability an agreement will be detected and prosecuted is higher if the agreement is illegal. If an agreement is detected, the firm will get a fine (𝐹). The fine the ACM sets is a percentage of the revenues of a firm. For simplicity, the fine in this thesis will be a proportion of the additional revenue from the agreement (𝑅𝐼 & 𝑅𝐿) for which we will assume 𝐹 > 1,
so the fine is more than the extra revenue a firm will make10. The pay-offs for the firm are presented in the lowest row of Fig. 4.1. If a firm knows whether the initiative is legal or not it is easy to make a trade-off between starting with the agreement and abandoning it, given the probability of detection, the fine and the profit. When looking from society’s point of view, there are two desirable outcomes; when behavior is anti-competitive, it is best if a firm abandons the behavior. When behavior is not anti-competitive, it is best if firms continue with the behavior since it can benefit society.
10
10 1 - q q p 1 - p Initiative Anti-competitive
Start
Detection+ R
I- F*R
I No detection: Type II Error+ R
IAbandon
Deterrence0
Not Anti-competitiveStart
Detection: Type I Error+ R
L- F*R
L No detection+ R
LAbandon
Over-deterrence0
A firm will not start with illegal behavior if the expected pay-off for not starting is higher than the expected pay-off for starting;
𝑞(𝑅𝐼 − 𝐹𝑅𝐼) + (1 − 𝑞)𝑅𝐼< 0,
or 𝐹 > 1/𝑞
Similarly, firms will decide to enter an agreement when behavior is legal if the expected pay-off from starting is higher than the expected pay-off from abandoning the legal agreement;
𝑝(𝑅𝐿− 𝐹𝑅𝐿) + (1 − 𝑝)𝑅𝐿> 0,
or 𝐹 < 1/𝑝
Because all variables are exogenous except for the fine, the fine is the only tool the ACM has to influence behavior of firms in this model. The optimal situation is therefore to set the fine such that 1
𝑞< 𝐹 < 1 𝑝,
which is possible since 𝑞 > 𝑝 and 𝐹 > 1. When the ACM decides to set the fine as to create the optimal situation, the optimal strategy for firms that find the law clear is simple; start when behavior is legal and deter when behavior is illegal. In this thesis there are 𝑛 firms that all face this decision once (so there are 𝑛 possible agreements). Nature chooses whether behavior is anti-competitive or not, where 𝑥 is the probability that an agreement is anti-competitive. It is not necessary for firms to know 𝑥 as they know if their own behavior is illegal or not11. It is only necessary to know the true value of 𝑥 when calculating the outcomes for society. When the fine is set such that 1
𝑞< 𝐹 < 1
𝑝 and when all firms know whether behavior
is illegal and play their optimal strategy, 𝑥 can be seen as the percentage of cases that are illegal and will
11 This assumption is plausible, since it is not likely that firms know exactly what proportion of all cases are illegal.
Fig. 4.1. Decision tree on whether to participate in an agreement, of a firm that finds the cartel law clear. Based on Van der Noll et al. (2010)
11 𝒙̇ 1 - 𝒙̇ 1 - q q p 1 - p Initiative Anti-competitive Start Detection + RI - F*RI No detection: Type II Error + RI Abandon Deterrence 0 Not Anti-competitive Start Detection: Type I Error + RL - F*RL No detection + RL Abandon Over-deterrence 0
be deterred from. Because there are 𝑛 possible agreements, a total of 𝑛 ∗ 𝑥 illegal agreements will be deterred from. A total of 𝑛 ∗ (1 − 𝑥) cases will start, which are all the cases that are not anti-competitive. 4.2 Uncertainty for firms that do not find the law clear
This paragraph will adjust the model of section 4.1 to find the optimal fine when firms do not find cartel law clear. When firms do not find the law clear, they do not know if their conduct is anti-competitive or not; uncertainty is introduced in the model. Fig. 4.2 presents the adjusted version of Fig. 4.1, allowing for this uncertainty. Because a firm does not know in which state (anti-competitive or not anti-competitive) it is, the trade-off firms make is more complex than without uncertainty. The decision or strategy of a firm in this game cannot simply be ‘start when behavior is legal and deter when behavior is illegal’, because they do not know in what state they are. The pay-offs and the probabilities of detection remain the same as in the model without uncertainty.
Also, firms do not know the true value of 𝑥 so they cannot calculate their expected pay-off using this probability that a case is illegal. It is probable, however, that firms do have a belief on whether behavior is illegal or not. In Fig. 4.2, 𝑥̇ is the perceived𝑥; it can be seen as the belief of a firm about their behavior being illegal, but it can also be interpreted as a measurement of risk-aversion. The 𝑥̇ of a firm is independent of 𝑥, because no firm knows the actual value of 𝑥.
For this thesis, the belief 𝑥̇ of a firm is drawn from a uniform distribution on the interval [0,1], independent of the other firms12. When the 𝑥̇ of a firm is high, it either believes that the probability of
12 One could argue that 𝑥̇ is more likely to be normally distributed than independently uniformly distributed. However, to keep the model and calculations simple and clear, the uniform distribution is used in this model. In both cases the effects are similar; 𝑥̇∗ is always positive but lower than 1 for 1
𝑞< 𝐹 < 1
𝑝. When comparing the normal distribution to the uniform distribution the deterrence effect will be weaker (i.e. more firms will start) when 𝑥̇∗ is
12
conduct being illegal is high, or it weighs the consequences of conduct being illegal as important. The firms are therefore risk-averse. When 𝑥̇ is low it is the other way around; a firm believes it is likely that the behavior is legal, or it weighs the consequences of the behavior being illegal as less important. A firm will enter an agreement when the expected pay-off from starting is higher than the expected pay-off from abandoning the agreement, which is represented in the next formula;
𝑥̇[𝑞(𝑅𝐼 − 𝐹𝑅𝐼) + (1 − 𝑞)𝑅𝐼] + (1 − 𝑥̇)[𝑝(𝑅𝐿− 𝐹𝑅𝐿) + (1 − 𝑝)𝑅𝐿] > 0,
or 𝑥̇(𝑅𝐼− 𝑞𝐹𝑅𝐼) + (1 − 𝑥̇)(𝑅𝐿− 𝑝𝐹𝑅𝐿) > 0
A firm will deter from the agreement when
𝑥̇(𝑅𝐼− 𝑞𝐹𝑅𝐼) + (1 − 𝑥̇)(𝑅𝐿− 𝑝𝐹𝑅𝐿) < 0.
The strategy of a firm will either be ‘start’ or ‘deter’, independent of whether behavior is illegal. The choice of a firm depends on the variables explained before as well as its 𝑥̇. Because firms have different 𝑥̇’s, some firms will enter and some firms will abandon an agreement. For every combination of 𝐹, 𝑅𝐼,
𝑅𝐿, 𝑞 and 𝑝 it is possible to determine which firms will start and which will deter. This is done by
calculating for which 𝑥̇thepayoff from starting is zero;
𝑥̇(𝑅𝐼− 𝑞𝐹𝑅𝐼) + (1 − 𝑥̇)(𝑅𝐿− 𝑝𝐹𝑅𝐿) = 0,
or 𝑥̇∗ = 𝑝𝐹𝑅𝐿− 𝑅𝐿
𝑅𝐼−𝑞𝐹𝑅𝐼−𝑅𝐿+𝑝𝐹𝑅𝐿
According to this formula all firms that have a 𝑥̇ smaller than 𝑥̇∗ will enter the agreement; it gives the
maximum 𝑥̇ in order to have a positive expected pay-off from entering an agreement. Because 𝑥̇ is uniformly and independently drawn from the interval [0,1], 𝑥̇∗ represents the percentage of firms that will
enter the agreement given the exogenous variables and 𝐹. The fine can take on several values13. If the fine is set above 1 𝑝� no firm will enter the agreement, which is never optimal because there will be over-deterrence14. If a fine is set below 1 𝑞� all firms will enter the agreement, which is also not optimal15. If 𝐹 is somewhere in between, a part of the firms (the firms with a low 𝑥̇) will enter the agreement, and a part above 0.5, and the deterrence effect is stronger (i.e. more firms will abandon behavior) when 𝑥̇∗ is below 0.5. Because this model only wants to prove the difference in behavior, using the uniform distribution is possible. 13As found earlier in this chapter, the optimal fine when firms do find cartel law clear is 1
𝑞< 𝐹 < 1
𝑝 which could therefore be taken as boundaries of the fine the ACM will set. However, this part tries to find the optimal fine for firms that do not find the law clear so fines below and above these boundaries have to be considered as well, even though it is not likely that those fines will occur in real life.
14 When F =1/p, 𝑥̇∗becomes 0 so no firm will start. 15 When F =1/q, 𝑥̇∗becomes 1 so all firms will start.
13 Firms that do
start with behavior
Firms that do not start with
behavior 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1 1,5 2 2,5 1/q 3 3,5 4 4,5 5 1/p 5,5 6 Fine Firms that do start with behavior Firms that do not start with
behavior 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1 1,5 2 2,5 1/q 3 3,5 4 4,5 5 1/p 5,5 6 Fine
of the firms will abandon the agreement. Theoretically, the line 𝑥̇∗= 𝑝𝐹𝑅𝐿− 𝑅𝐿
𝑅𝐼−𝑞𝐹𝑅𝐼−𝑅𝐿+𝑝𝐹𝑅𝐿, will be
decreasing from 1 to 0 for a fine between 1 𝑞� and 1 𝑝� , i.e. the number of firms starting with an agreement is positive but decreases when the fine increases between 1 𝑞� and 1 𝑝� . This is shown in Fig. 4.3, which has the fine on the horizontal axis and 𝑥̇∗, representing the percentage of firms that will start, on the
vertical axis16. The grey area indicates the firms that will enter the agreement, given the variables, and the white area indicates the firms that will abandon the agreement. For this example, when the fine is set at 3.5, approximately 35% of the firms will enter the agreement and when the fine is set at 4.5 approximately 10% of the firms will start.
𝑥̇∗= 𝑝𝐹𝑅𝐿− 𝑅𝐿 𝑅𝐼−𝑞𝐹𝑅𝐼−𝑅𝐿+𝑝𝐹𝑅𝐿
Because firms that do not think the law is clear cannot distinguish legal from illegal cases, there will always be firms that enter an agreement that is illegal, firms that abandon an agreement that is not illegal, or both. Therefore, there is no optimal fine.
4.3 Comparing the situations
The conclusion from section 4.1 is that firms that know whether behavior is legal will always choose to play the optimal strategy for society, as long as 1
𝑞 < 𝐹 < 1
𝑝. Section 4.2 finds that for firms that do not
know whether behavior is illegal, there is no optimal fine. Not knowing whether behavior is illegal has
16 This graph is calculated with the following variables; 𝑅𝐼
= 50, 𝑅𝐿= 40, 𝑞 = 0.4 and 𝑝 = 0.2. Therefore, 1 𝑞� = 2.5and 1 𝑝� = 5.
Fig. 4.3. Graph that shows the firms that will and will not enter an agreement, depending on 𝑭, 𝑹𝑰, 𝑹𝑳, 𝒒 and 𝒑
The horizontal axes represents the fine, the vertical axes represents the percentage of firms that will start given the variables.
14
two undesirable effects; there will be more deterrence from legal behavior (over-deterrence), and less deterrence from illegal behavior.
It is possible to calculate the impact of this outcome on society. As stated before, the decision of a firm does not depend on the actual 𝑥 in this model, but the outcome for society does. When firms think cartel law is clear, 𝑛 ∗ 𝑥 cases are illegal and are deterred from, and 𝑛 ∗ (1 − 𝑥) cases are legal and will be entered, so there is no undesirable behavior. However, when looking at the firms that face uncertainty it can easily be calculated how many firms enter illegal behavior and how many firms are deterring from legal behavior. The results are presented in Table 4.1.
Since 1 > 𝑥̇∗> 0, 𝑛 ∗ 𝑥 > (1 − 𝑥̇∗) ∗ 𝑛 ∗ 𝑥, so fewer firms deter from illegal behavior. Similarly, since
(1 − 𝑥̇∗) ∗ 𝑛 ∗ (1 − 𝑥) > 0, more firms abandon legal behavior when there is uncertainty. Both are
undesirable outcomes. When looking at which effect dominates, so whether there is more or less deterrence in total, the results show that if𝑥 > (1 − 𝑥̇∗) firms that do not find the law clear deter less, and
when𝑥 < (1 − 𝑥̇∗) firms that do not find the law clear deter more.
Tables 4.2 and 4.3 present two numerical examples. For Table 4.2 the same inputs as for Fig. 4.3 are used17. Also, 𝑛 = 1000, 𝑥 = 0.8, and 𝐹 = 3. Therefore, 𝑥̇∗ is 0.615 which means that𝑥 > (1 − 𝑥̇∗), so there should be less deterrence. With the formulas from Table 4.1 the number of abandoned and started cases can be calculated. Table 4.2 shows that, indeed, there is less deterrence with uncertainty; with uncertainty 385 cases are deterred from, whereas 800 cases will be deterred from when there is no uncertainty. In total 492 illegal cases will start, and in 75 cases there is over-deterrence.
Table 4.3 presents a situation where the probability that the government detects an illegal agreement is high and the gains from an agreement are low; 𝑅𝐼= 20, 𝑅𝐿= 15, 𝑞 = 0.7 and 𝑝 = 0.4, 𝐹 = 3,
𝑛 = 1000 and 𝑥 = 0.6. Therefore, 𝑥̇∗ is 0.27. This means that 𝑥 < (1 − 𝑥̇∗), so in this scenario there
17
𝑅𝐼= 50, 𝑅𝐿= 40, 𝑞 = 0.4 and 𝑝 = 0.2. Therefore, 1 𝑞� = 2.5and 1 𝑝� = 5.
No uncertainty Uncertainty
Start Abandon Start Abandon
Anti-Competitive 0 𝑛 ∗ 𝑥 (𝑥̇∗) ∗ 𝑛 ∗ 𝑥 (1 − 𝑥̇∗) ∗ 𝑛 ∗ 𝑥
Not Anti-Competitive 𝑛 ∗ (1 − 𝑥) 0 (𝑥̇∗) ∗ 𝑛 ∗ (1 − 𝑥) (1 − 𝑥̇∗) ∗ 𝑛 ∗ (1 − 𝑥)
Total 𝑛 ∗ (1 − 𝑥) 𝑛 ∗ 𝑥 𝑛 ∗ (𝑥̇∗) 𝑛 ∗ (1 − 𝑥̇∗)
Table 4.1. Number of (il)legal agreements that will be entered and abandoned compared between firms that do find the law clear and firms that do not
15
should be more deterrence overall. When calculating the results, it can be seen that this is true. In total 730 cases will be abandoned when there is uncertainty, whereas 600 cases will be deterred from when there is no uncertainty. In 292 cases there is over-deterrence, and 162 illegal cases will start.
4.4 Hypotheses
This thesis will answer three questions for the Netherlands; 1) What type of firms think the cartel law is not clear?
2) Do those firms behave differently with respect to deterrence? 3) What factors influence the level of deterrence?
The model developed in this chapter found that firms that do not know exactly if behavior is illegal or not will deter less from behavior that is illegal and deter more from behavior that is not illegal.
With the data from the survey of Van der Noll et al. (2010) it is not possible to exactly replicate the model developed in this chapter. The data does not contain all variables needed to test the theoretical model, such as probability of detection. Also, it is not possible to distinguish between illegal and legal behavior in the case study. This does not mean this model is not helpful for analyzing the Dutch case. It can still be used to reason what outcome is expected from the data, and several aspects of this model can be tested with the data. The part of the model this thesis focuses on is behavior if an agreement is illegal, since respondents ‘suspect their behavior to be illegal18’ when they fill in the survey. The model predicts that there is less deterrence from firms that do not find cartel law clear. There are several reasons why this can be the case. One reason is that those firms did not take the effort or invest the time to get informed about the law. Another reason could be that those firms are less risk-averse. They might weigh the consequences of being punished less than firms that find the law clear. Therefore, in this case study of the Netherlands we expect less deterrence from firms that do not find the cartel law clear. The data provides probabilities of deterrence, which will be used to measure deterrence.
18 See Box 3.1.
No uncertainty Uncertainty No Uncertainty Uncertainty
Start Abandon Start Abandon Start Abandon Start Abandon
Anti-competitive 0 800 492 308 Anti-competitive 0 600 162 438
Not anti-competitive 200 0 123 75 Not anti-competitive 400 0 108 292
Total 200 800 615 385 Total 400 600 270 730
Table 4.2. Number of cases abandoned and started with and without uncertainty. There is less deterrence with uncertainty in this scenario
Table 4.3. Number of cases abandoned and started with and without uncertainty. There is more deterrence with uncertainty in this scenario
16
Hypothesis 1 – Firms that do not think the law is clear will give a lower probability of deterrence.
Factors that are expected to influence the clarity of the law are whether the firm has a compliance officer and the amount of employees the firms has. There are two reasons why having a compliance officer is related to the clarity of the law of the CEO or internal company lawyer of a firm. The first reason is that firms that are more concerned with knowing the law take more care of compliance, and therefore might hire a compliance officer. The second reason is that compliance officer probably has to report to the internal company lawyer or CEO in a firm, so because of the presence of a compliance officer the CEO and the internal company lawyer will get to know the law better.
If a firm is large it is relatively cheaper to hire a compliance officer which therefore influences the clarity of the law. Also, when a firm as more employees, there may be a higher probability that at least one employee of the firm has some knowledge of cartel law.
Firm size can be linked to probability of deterrence directly as well. The exemptions for the cartel law19 apply mostly for smaller firms; the larger the firm, the less likely it is to qualify for exemption of cartel law. When separating between large and small firms and relating to the model, for small firms the value of 𝑥 is lower, i.e. fewer cases are illegal. This means that firms that are small deter less. However, in theory this should not apply for firms that do not find cartel law clear, since they do not know about the exemptions.
Hypothesis 2 - Firms that have a compliance officer and/or more employees will find the law clear more often and give a higher probability of deterrence.
The first two hypotheses are about whole sample, but one can expect that results differ across sectors. In the survey every respondent had to indicate in which sector their firm was operating. Because sectors are different – one example is that some sectors are capital intensive whereas others are labour intentsive – the way of managing a firm can differ among sectors. Firms in some sectors will be more carefull about knowing every aspect of the law – for example because the sector is specified as focus area on the agenda of the ACM – whereas in other sectors firms do not pay much attention to the law. Therefore the clarity of cartel law between sectors may differ, as well as firm size and the number of compliance officers. Because sectors are expected to be different, the probability of deterrence is expected to differ across sectors too.
Hypothesis 3 – The probability of deterrence differs across sectors.
17
The hypotheses above test what type of firms do not think cartel prohibition is clear and what the effect on deterrence is. It is also interesting to see what types of punishments are important in determining the probability of deterrence and if there is a difference between firms that do find the cartel law clear and firms that do not. Firms that do not find the law clear are expected to give a lower probability of deterrence. If firms are less risk-averse, most of types of punishments might not have a large effect on the probability of deterrence. The leniency program, for example, is expected to have little effect; firms that do not find the cartel law clear might not know about this. Previous literature, as described in chapter 2, found that a personal fine affects deterrence. Even though firms that do not find cartel law clear might be less risk-averse, one can still expect the personal fine to have a deterrence effect.
Hypothesis 4 – Firms that find the law unclear indicate the personal fine as most important reason for deterrence.
The last hypothesis in this thesis is about firms that do find the law clear. Because those firms understand the law they know about the leniency program, an efficient instrument to incentivize firms from deterring illegal behavior (Bigoni et al., 2014). Therefore, the leniency program is expected to be an important determinant of probability of deterrence.
Hypothesis 5 – Firms that find the law clear view leniency as an important factor of probability of deterrence.
18 5. Case study – The Netherlands
The hypotheses are based on the theoretical model, the available data, literature and economic intuition. In this chapter, the hypotheses will be tested with data from the Netherlands.
5.1 Probability of deterrence
The first analysis that is needed for this thesis is to check whether perceived clarity of the law has an effect on deterrence. As explained in the previous chapters, this thesis will use probability of deterrence as a measurement for deterrence. Those probabilities are an accurate measurement of deterrence because it indicates how likely a firm is to deter from an agreement. When looking at those probabilities it is difficult to interpret the exact meaning of, for example, an 8. However, the probabilities are useful when comparing them between groups, because it shows whether there is a difference in behavior between groups.
Hypothesis 1 – Firms that do not think the law is clear will give a lower probability of deterrence.
The dataset of the survey contains all the probabilities that the respondents assigned to the scenarios. When looking at the average probability respondents assigned and comparing those probabilities between firms that do think cartel law is clear and do not think cartel law is clear, a first indication of behavior is given. Table 5.1 shows the results, using a T-test. It shows that the average probability of deterrence over the whole sample is 8.6, but when separating between the two groups of firms a significant difference is found. The average probability of a firm that does not think the law is clear is 8, whereas a firm that does think the law is clear gave on average an 8.9 as probability of stopping the conduct. It is difficult to precisely interpret the meaning of those numbers but, as argued before, the relative ranking does indicate a difference in behavior. Therefore, it can be said with 99% confidence that firms that do not think the law is clear gave, on average, a lower probability of stopping conduct that they were presented in the scenarios.
The average probability of deterrence, separated for firms that find the law clear and firms that do not find the law clear. *Significant at 0.1 level **Significant at 0.05 level ***Significant at 0.01 level.
20‘Probability of deterrence’ in this and following tables indicates the average stated probability of deterrence of the respondents of the survey.
The law is… N Probability of deterrence20
Not clear 408 7.966***
Clear 1080 8.87***
Combined 1488 8.6
19
We do need to check for the type of scenarios both groups faced. Because the values of the vignettes were randomly assigned, there is a chance that one group faced worse punishments than the other group. To test for this bias each value of each vignette is given a number between 1 and 4, where 4 is the least desirable for a firm and 1 is the most desirable for a firm. When adding those numbers and testing for the differences in scenarios between the groups, the outcome is that firms that do not think the cartel law is clear faced worse scenarios than the firms that did think the law was clear21. Therefore, the difference between probabilities of deterrence could be even more when both groups face the same scenarios. Another interesting outcome is that the standard deviation of the probabilities of firms that do not think cartel law is clear, is significantly larger than that of firms that do think the law is clear22. This indicates that there is a larger spread in the answers of firms that do not find the law clear. This is predicted by the model in chapter 4, because it indicates that firms that do find the law clear all have the same strategy, whereas firms that do not find the law clear do not have similar strategies.
5.2 Determinants of clarity of the law and effects on probability of deterrence.
The data shows a difference in behavior between firms that do not find cartel law clear and firms that do find the law clear. The next step is to find characteristics of firms that do not find the law clear and firms that give a low probability of deterrence.
Hypothesis 2 - Firms that have a compliance officer and/or more employees will find the law clear more often and give a higher probability of deterrence.
Tables 5.2 and 5.3 present the results for the first part of this hypothesis. Table 5.2 shows that the average size of firms that do find the law clear is 377 employees, whereas the average number of employees of a firm that does not think the law is clear is 218, which is a significant difference. Therefore it can be concluded that, indeed, firms that do think cartel legislation is clear are larger on average. To test whether larger firms find the law clear more often, the firms are divided into 3 groups; small firms (the 25% smallest firms), medium firms (the 25% - 75% percentile), and large firms (the 25% largest firms). Table 5.3 shows that 81.82% of the large firms find the law clear, which is significantly more than the medium and small firms. Among medium and small firms, a less than average percentage finds the law clear.
21 Found with a T-test. The average value of the scenarios of firms that do not think cartel law is clear was 11.06, whilst firms that did think cartel law is clear faced a scenario with an average value of 10.77. This difference is significant at a 1% level.
22 The standard deviation of the average probability of deterrence of firms that find the law clear is 1.525, and for firms that do not find the cartel law clear it is 2.76. This differs significantly at a 1% level.
20
When looking at whether firms have a compliance officer and clarity of the law, Table 5.4 shows that 77% of the firms that that have a compliance officer think the law is clear, which is significantly more than in the group of firms that do not have a compliance officer. Table 5.5 shows that 48% of the firms that find the law clear have a compliance officer, whereas only 33% of the firms that do not find the law clear have a compliance officer. With 99% confidence it can be concluded that the firms in this survey that do not have a compliance officer less often find the law clear, and that firms that do not think the law is clear less often have a compliance officer. Therefore, not having a compliance officer seems to go hand in hand with not finding the law clear.
As there are arguments why large firms more often have a compliance officer, it is interesting to combine the factors to see whether this is true for this dataset. When testing if firms with a compliance officer are larger than the ones without, the results (Table 5.6) show that the difference in average firm size between firms that do not have a compliance officer and firms that do is large and significant at a 0.01 level.
The law is.. N Firm size Firm size N Clarity
Not Clear 108 218*** Small 100 65%
Clear 250 377*** Medium 172 66.86%
Combined 358 329 Large 88 81.82%***
Combined 360 70%
N How many firms find the law clear? The law is.. N How many have a compliance officer? No compliance officer 203 64.5%*** Not clear 108 33.3%***
Compliance officer 157 77%*** Clear 152 48%***
Combined 360 70% Combined 360 43.60%
Group N Firm size
No compliance officer 294 260***
Compliance officer 210 406***
Combined 504 321
Clarity of the law and firm size. ***Significant at 0.01 level.
Average firm size of firms with and without a compliance officer. ***Significant at 0.01 level.
Table 5.4. Firms that have a compliance officer more often find the law clear
Table 5.5. Firms that find the law clear more often have a compliance officer
Table 5.6. Firms with a compliance officer are larger on average Table 5.2. Firms that find the law clear are
larger than firms that do not, on average
Table 5.3. Large firms more often find the law clear
% clarity of the law when distinguishing between firms that do and do not have a compliance officer. ***Significant at 0.01 level.
% compliance officers when distinguishing between firms that do and do not find the law clear. ***Significant at 0.01 level.
Firm size and clarity of the law. ***Significant at 0.01 level.
21
One way of reasoning is that it is relatively cheaper for large firms to hire a compliance officer. Another reason is that there are fewer exemptions from the cartel law for large firms23. Therefore, more cases are illegal for large firms which is an incentive to hire a compliance officer, and thus having more employees has an effect on hiring a compliance officer. This result is shown in Fig. 5.1 with an arrow from firm size to compliance officer accompanied by the positive sign. If firm size increases, a firm is more likely to have a compliance officer.
As shown, both having a large firm and having a compliance officer have a positive effect on clarity of the law. The larger a firm becomes, the more likely it is to find the law clear. It is possible to reason the other way around as well, but those reasons are farfetched and will only have a small effect, so this thesis will ignore this effect24. The relationship between having a compliance officer and finding the law clear goes in both directions, as argued in chapter 4. Firms that understand the law better might want to comply with it more and therefore hire a compliance officer. The other way around, firms that have a compliance officer will more often find the law clear because of the information flows within the company.
Probability of deterrence
The next step is to see whether firms that are larger and/or that have a compliance officer give a higher probability of deterrence. Afterwards, the effects will be separated to see whether both factors affect deterrence directly, and not only through clarity of the law.
Table 5.7 shows that firms that have a compliance officer give on average a probability of 9.14 whereas firms that do not have a compliance officer give a probability of 8.2, which is a significant difference. In the previous part of this chapter it was found that firms that have a compliance officer more often think cartel law is clear, and firms that think cartel law is clear give a higher probability of deterrence, so this result is not surprising.
23 For most firms, more employees indicates more revenue. Revenue is a criterion for the exemptions.
24 For example, finding the law more clear could ensure that a firm only engages in legal behavior which would decrease the probability of a fine. This, in turn, can increase the expected revenue which in the long term can increase the number of employees.
+
+
+
+
Fig. 5.1. The relationships between firm size, whether firms have a compliance officer, and if firms find the law clear
Firm size
Compliance Officer
22
In this paper there will be several tables similar to Table 5.8. The stars on the left-hand side of the ‘forward slash’ indicate a column level significance. The stars on the right-hand side indicate a row level significance. For example, when looking at firms with a compliance officer, those who find the law clear give a 9.2 on average, which is a significantly higher probability of deterrence (at 0.05 level) than firms that do not find the law clear (they give an 8.9). When looking only at firms that find the law clear, the probability of deterrence of firms with a compliance officer is higher (at 0.01 level) than that of firms with a compliance officer. When looking at firms that do not find the law clear, firms with a compliance officer also give a significantly higher probability of deterrence (at 0.01 level) than firms without.
It is interesting to see if having a compliance officer in itself has an effect on the probability of deterrence, or if this higher probability only occurs because those firms more often think the law is clear. The results in Table 5.8 prove that both clarity of the law and having a compliance officer have a significant individual effect. Box 5.1 explains how one should read these types of tables.
When separating between firms that find the law clear and firms that do not find the law clear (Table 5.8), results show that for both groups having a compliance officer increases the probability of deterrence. It is noticeable that the firms that do not think the law is clear but do have a compliance officer give a probability of 8.9, which is higher than the average probability of the overall sample (8.6). It therefore suggests that having a compliance officer is an important factor for the probability of deterrence, and maybe even more important than whether a firm thinks the law is clear.
Law is clear Law is not clear
N Probability of deterrence N Probability of deterrence
No Compliance officer 534 8.55***/*** 282 7.55***/*** Compliance officer 546 9.2***/** 126 8.9***/** Combined 1080 8.87 408 7.96 N Probability of deterrence No Compliance officer 816 8.2*** Compliance officer 672 9.14*** Combined 1488 8.6
Explanation how to read the significance of this table is presented in Box 5.1. ** Significant at 0.05 level, ***Significant at 0.01 level.
Box 5.1. Explanation on how to read the significance of results in Table 5.8, 5.10, 5.11 and 5.13
Table 5.8. Both clarity of the law and having a compliance officer have a direct positive effect on probability of deterrence
Table 5.7. Firms with a compliance officer give, on average, a higher probability of deterrence than firms that do not have a compliance officer
23
Table 5.8 also shows that for both firms with and without a compliance officer, clarity of the law gives a significantly higher probability of deterrence. Therefore, both clarity of the law and having a compliance officer have a positive effect on probability of deterrence.
When looking at firm size, the same groups for small, medium and large firms are used as earlier in this chapter. Table 5.9 presents the results from a T-test when testing for the average probability of deterrence. It shows that large firms give on average a higher probability of deterrence. This is not surprising since larger firms more often think cartel law is clear (Table 5.3) and more often have a compliance officer (Table 5.6). Less than average small and medium sized firms find the law clear and the average probability of deterrence of small and medium sized firms is lower than average, which is in line with previous findings. But before being able to draw any conclusion from these results, it is necessary to check if firm size has a direct effect on probability of deterrence.
Table 5.10 shows the results when comparing firm size and clarity of the law. For small firms, clarity of the law does not seem to make much difference for the average probability of deterrence – both groups give a lower probability of deterrence than the overall sample. For medium and large firms it does make a difference. Firms within those groups that find the law clear give a significant higher probability of deterrence. When looking only at firms that find the law clear it shows that there is a significant difference (at 0.01 level) of the probability of deterrence between small, medium and large firms that find the law clear; large firms give the highest probability of deterrence, then the medium firms, and the small firms give the lowest probability of deterrence – which is in line with the expectation. When looking at firms that do not find cartel law clear, the differences are less significant and the averages are closer to each other. This is in line with the predictions of the theoretical model. Firms that do not find the law clear generally do not know about the exact rules of the exemptions so their behavior is considered to be independent of the size of a firm. Firms that do find the law clear know about the exemptions and therefore large firms deter more than small firms, as explained in section 4.
Firm size N Probability of deterrence
Small 402 8.41**
Medium 708 8.55*
Large 378 8.98***
Combined 1488 8.6
Average probability of deterrence when distinguishing between small, medium and large firms. * Significant at 0.10 level, ** Significant at 0.05 level, ***Significant at 0.01 level.
24
The last combination of characteristics in this section is firm size and whether firms have a compliance officer, to see if those factors affect probability of deterrence independently of each other. Table 5.11, which has the same type of lay-out as Tables 5.8 and 5.10, presents the results. For every group of firm size, having a compliance officer significantly increases the probability of deterrence. It is noticeable that for firms that have a compliance officer, firm size does not seem to matter for the probability of deterrence. Not one outcome is significant. For the firms that do not have a compliance office firm size does make a difference; large firms give a significantly higher probability than smaller firms, and medium sized firms give a significantly lower probability than small and large firms.
Small Medium Large Total
N Probability of deterrence N Probability of deterrence N Probability of deterrence No Compliance officer 282 8.15***/. 360 7.96***/*** 174 8.79***/*** 816 8.2
Compliance officer 120 9.03***/. 348 9.17***/. 204 9.14***/. 672 9.14
Combined 402 8.41 708 8.55 378 8.98
Conclusion
From the results found section 5.2, one can conclude that all three factors – clarity of the law, having a compliance officer and firm size – have a positive effect on probability of deterrence. This does not only seem to go through the ‘clarity’ variable; all factors have an individual effect. With these findings Fig. 5.1 can be extended to Fig. 5.2.
Small Medium Large Total
N
Probability of deterrence N
Probability
of deterrence N Probability of deterrence Not Clear 126 8.23./* 222 7.7***/* 60 8.4**/** 408 7.96 Clear 276 8.49./*** 486 8.94***/* 318 9.1**/*** 1018 8.87
Combined 402 8.41 708 8.55 378 8.98
Explanation how to read the significance of this table is presented in Box 5.1. * Significant at 0.10 level, ** Significant at 0.05 level, ***Significant at 0.01 level.
Explanation how to read the significance of this table is presented in Box 5.1. * Significant at 0.10 level, **Significant at 0.05 level, ***Significant at 0.01 level.
Table 5.10. For firms that do not find the law clear firm size does not matter for probability of deterrence. For firms that do find the law clear firm size does matter
Table 5.11. Having a compliance officer has a positive influence on probability of deterrence for all firm sizes
