Difference between revisions of "Cartel simulation"

From Simulace.info
Jump to: navigation, search
 
(12 intermediate revisions by the same user not shown)
Line 33: Line 33:
 
==Model==
 
==Model==
  
[add text]
+
I start with the general setup, where I create the following turtle breeds:
  
'''Method description'''
+
* enforcement-officials (white persons)
 +
* stable-cartels (red houses)
 +
* unstable-cartels (yellow houses)
 +
 
 +
 
 +
[[File:General-setup.gif]]
 +
 
 +
 
 +
The enforcement officials are together tackling some stable and some unstable cartels. To take down a stable cartel, it takes 4 years (= four ticks in NetLogo). To take down an unstable cartel, it only takes 2 years (= two ticks in NetLogo).
 +
 
 +
There are unstable cartels when cheating on the collusive agreement results in a bigger gain in profit. An indicator for more unstable cartels is, for instance, that there are more businesses or that the demand function for the firm has a higher price elasticity (demand function has not been included in the simulation, we just see the effect in the lower life expectancy). The differential effect is driven by the dying cartels that use the leniency program. In industries with less stable cartels, the rate of dying cartels is higher.
 +
 
 +
We can then see the birth, death, and prosecution pre leniency in the following simulation:
 +
 
 +
 
 +
[[File:General-simulation.gif]]
 +
 
 +
 
 +
The number of all the cartels is constantly around the number of about 30 cartels, on the left hand side there is a monitor showing how the number of the stable and the unstable cartels are changing over time.
 +
 
 +
Cartels do have an average life expectancy between 7.5 and 7.7 years according to Combe et al. (2008) as well as the data from the European Commission. Harrington and Chang (2015) explain that stable cartels in general have a higher life expectancy than unstable cartels. Thus, in my simulation, I have assigned a life duration of 6 to 11 years to stable cartels and 3 to 6 years to unstable cartels. This should more or less result in an average life expectany of 7.5 to 7.7 years for all the cartels taken together.
 +
 
 +
'''With leniency program'''
 +
 
 +
Sp now I simulate the situation when there is a leniency program. Since the unstable cartels want to avoid getting fined, there is a race for leniency application (they want to be the first of the cartel members to apply for it since in that case their fine will be reduced). This decreases the life duration of unstable cartels. At the same time, the life duration of stable cartel increases, since the enforcement officials are more focused on the unstable cartels. The below simulation shows that as soon as the leniency program is added, the number of unstable cartels (yellow) decreases, while the number of stable cartels (red) increases:
 +
 
 +
[[File:Leniency-simulation.gif]]
  
  
Line 41: Line 67:
 
== How to use the program ==
 
== How to use the program ==
  
[add text]
+
So, how to use?
 +
 
 +
* First step: Open the model in NetLogo
 +
* Second step: Click on the "Setup" button
 +
* Third step: Click on the "go" button
 +
* Fourth step: Click on the "add leniency program" button
 +
 
 +
And then, as a "special", you can also add the number of enforcement officials. It shows that when there are more enforcers, there will be less cartels in the total simulation, since there is more workforce to take down the cartels. Below, see the comparison between having a 100 officials versus 200 officials, including the leniency program applied in the middle of the simulation and then stopped again:
 +
 
 +
 
 +
[[File:Number-of-enforcers.JPG]]
 +
 
 +
 
 +
It is clearly visible that with more enforcement officials, the total number of cartels is significantly lower then with less officials.
  
 
==Conclusion==
 
==Conclusion==
  
[add text]
+
I have tested the simulation various times and have gotten the same results overall, which shows a certain robustness of the model, as well as accuracy. It is in line with the study of Harrington and Chang (2015). The following remarks can be derived from the outcome of the simulation in Netlogo:
  
 +
* The leniency program influenced non-leniency enforcement. If the leniency program is applied, stable cartels have a higher life expectancy than before, because enforcement officials are focused on unstable cartels which might have died soon, anyway.
 +
* The more enforcement officials, the less cartels there are in total. This can be seen when changing the number of enforcers in the input field.
 +
 +
Note: These conclusions have been drawn from extensive literture review. They have merely been implemented in the simulation. From the simulation itself, no new conclusions can be drawn, as it is simply a representation of the market itself. The author hereby disclaims any responsibility for any discrepancies with further research, should inconsistencies with other studies arise.
  
 
==Code==
 
==Code==
[add text]
+
 
 +
Here the NetLogo file can be downloaded: https://www.simulace.info/index.php/File:Simulation-assignment-baur00-final.nlogo
 +
 
 +
[[File:Simulation-assignment-baur00-final.nlogo]]
 +
 
 +
==References==
 +
 
 +
* Brenner, S. (2009). An empirical study of the European corporate leniency program. International Journal of Industrial Organization, 27(6), 639-645.
 +
* Combe, E., Monnier, C., & Legal, R. (2008). Cartels: The probability of getting caught in the European Union. Available at SSRN 1015061.
 +
* Harrington Jr, J. E., & Chang, M. H. (2015). When can we expect a corporate leniency program to result in fewer cartels?. The Journal of Law and Economics, 58(2), 417-449.
 +
* Hoang, C. T., Hüschelrath, K., Laitenberger, U., & Smuda, F. (2014). Determinants of self-reporting under the European corporate leniency program. International Review of Law and Economics, 40, 15-23.
 +
* Marvao, C. (2016). The EU leniency programme and recidivism. Review of Industrial Organization, 48(1), 1-27.
 +
* Mihai, B. (2008). Cartels–between theory, leniency policy and fines. Annals of Faculty of Economics, 1(1), 549-552.

Latest revision as of 19:41, 22 January 2023

Introduction

The leniency program of the European Commission offers the companies involved in a cartel either complete or partial immunity from fines if they self-report and hand over evidence. It was introduced in 1996, following the surge in amnesty applications in the wake of the 1993 revision of the Corporate Leniency Program of the US Department of Justice’s Antitrust Division. Reports from various implemented leniency programs showed that such programs led to numerous applications. However, despite the clear increase in leniency applications, the question poses itself as to whether the programs were also successful in a sense that the actual cartel rate in those countries declined. The simulation will be based on a study of Harrington and Chang from 2015, in which they concluded the following:

• The actual cartel rate decreases in case that the leniency program does not affect the non-leniency enforcement

• But: if the non-leniency enforcement is affected because resources are shifted to the prosecution of leniency application cases, there might be two possibilities, the cartel rate might increase.

This simulation focuses on the latter case. Assuming endogenized non-leniency enforcement, the introduction of a leniency program might have a differential impact on different industries. If a leniency program is introduced, the cartels that are about to collapse will seek to self-report. This in turn shifts resources from exposing active cartels to prosecuting cartels that are already collapsing. This creates more work for the authorities, who, instead of focusing on active cartels may now focus on dying cartels. This crowding-out effect coming about with the introduction of a leniency program shall be simulated in this project.

Goal

The simulation will have the following objectives:

  • Illustrate the change in cartel rates and the change in the average life expectancy of a cartel triggered by the introduction of a leniency program in case of endogenized non-leniency enforcement for industries with unstable cartels (e.g. industries with a high number of competitors, or demand with more price elasticity) and for industries with stable cartels (e.g. industries with less competitors and demand with less price elasticity).
  • Illustrate how many resources may be shifted from non-leniency enforcement to prosecuting leniency application cases without it having an undesired effect on the actual cartel rate.

Simulation and data

The dynamic that controls N(t), the number of cartels alive at time t, was first described by Bryant and Eckard (1991) using the concept of birth and death. I use this framework to model cartel detection as well. Contrary to these authors, I take into account three processes in the model that follows: (1) one that has to do with cartel formation, (2) another with their natural demise, and (3) a third that handles cartel detection. Some cartels are discovered while they are still in operation, and I hypothesize that this discovery causes the cartel to disband as a result of antitrust intervention. Ex post cartel detection, after a cartel has died, is not considered in this model. Cartel duration (life expectancy) is considered "natural".

Framework-cartel-lifecycle.JPG

The leniency program was introduced in 1996. I have selected all data from the EU commission from 1964 until 2010 (short preview below):


Preview-of-data.JPG


In total, there are 196 prosecuted cartel cases for those years. According to Combe at al. (2008), 12.9% to 13.3% of all the cartels are not detected. Using this in my calculation, this would result in an approximate average of 31 cartels per year for the complete period of time from 1964 to 2010. Now, splitting the data into the period from 1964 to 1996 (without leniency) and the period from 1996 to 2010 (with leniency), I get an approximate of 27 cartels per year before the introduction of leniency and 45 cartels per year after the introduction of leniency. This can be simplified and used for the simulation.

Model

I start with the general setup, where I create the following turtle breeds:

  • enforcement-officials (white persons)
  • stable-cartels (red houses)
  • unstable-cartels (yellow houses)


General-setup.gif


The enforcement officials are together tackling some stable and some unstable cartels. To take down a stable cartel, it takes 4 years (= four ticks in NetLogo). To take down an unstable cartel, it only takes 2 years (= two ticks in NetLogo).

There are unstable cartels when cheating on the collusive agreement results in a bigger gain in profit. An indicator for more unstable cartels is, for instance, that there are more businesses or that the demand function for the firm has a higher price elasticity (demand function has not been included in the simulation, we just see the effect in the lower life expectancy). The differential effect is driven by the dying cartels that use the leniency program. In industries with less stable cartels, the rate of dying cartels is higher.

We can then see the birth, death, and prosecution pre leniency in the following simulation:


General-simulation.gif


The number of all the cartels is constantly around the number of about 30 cartels, on the left hand side there is a monitor showing how the number of the stable and the unstable cartels are changing over time.

Cartels do have an average life expectancy between 7.5 and 7.7 years according to Combe et al. (2008) as well as the data from the European Commission. Harrington and Chang (2015) explain that stable cartels in general have a higher life expectancy than unstable cartels. Thus, in my simulation, I have assigned a life duration of 6 to 11 years to stable cartels and 3 to 6 years to unstable cartels. This should more or less result in an average life expectany of 7.5 to 7.7 years for all the cartels taken together.

With leniency program

Sp now I simulate the situation when there is a leniency program. Since the unstable cartels want to avoid getting fined, there is a race for leniency application (they want to be the first of the cartel members to apply for it since in that case their fine will be reduced). This decreases the life duration of unstable cartels. At the same time, the life duration of stable cartel increases, since the enforcement officials are more focused on the unstable cartels. The below simulation shows that as soon as the leniency program is added, the number of unstable cartels (yellow) decreases, while the number of stable cartels (red) increases:

Leniency-simulation.gif


How to use the program

So, how to use?

  • First step: Open the model in NetLogo
  • Second step: Click on the "Setup" button
  • Third step: Click on the "go" button
  • Fourth step: Click on the "add leniency program" button

And then, as a "special", you can also add the number of enforcement officials. It shows that when there are more enforcers, there will be less cartels in the total simulation, since there is more workforce to take down the cartels. Below, see the comparison between having a 100 officials versus 200 officials, including the leniency program applied in the middle of the simulation and then stopped again:


Number-of-enforcers.JPG


It is clearly visible that with more enforcement officials, the total number of cartels is significantly lower then with less officials.

Conclusion

I have tested the simulation various times and have gotten the same results overall, which shows a certain robustness of the model, as well as accuracy. It is in line with the study of Harrington and Chang (2015). The following remarks can be derived from the outcome of the simulation in Netlogo:

  • The leniency program influenced non-leniency enforcement. If the leniency program is applied, stable cartels have a higher life expectancy than before, because enforcement officials are focused on unstable cartels which might have died soon, anyway.
  • The more enforcement officials, the less cartels there are in total. This can be seen when changing the number of enforcers in the input field.

Note: These conclusions have been drawn from extensive literture review. They have merely been implemented in the simulation. From the simulation itself, no new conclusions can be drawn, as it is simply a representation of the market itself. The author hereby disclaims any responsibility for any discrepancies with further research, should inconsistencies with other studies arise.

Code

Here the NetLogo file can be downloaded: https://www.simulace.info/index.php/File:Simulation-assignment-baur00-final.nlogo

File:Simulation-assignment-baur00-final.nlogo

References

  • Brenner, S. (2009). An empirical study of the European corporate leniency program. International Journal of Industrial Organization, 27(6), 639-645.
  • Combe, E., Monnier, C., & Legal, R. (2008). Cartels: The probability of getting caught in the European Union. Available at SSRN 1015061.
  • Harrington Jr, J. E., & Chang, M. H. (2015). When can we expect a corporate leniency program to result in fewer cartels?. The Journal of Law and Economics, 58(2), 417-449.
  • Hoang, C. T., Hüschelrath, K., Laitenberger, U., & Smuda, F. (2014). Determinants of self-reporting under the European corporate leniency program. International Review of Law and Economics, 40, 15-23.
  • Marvao, C. (2016). The EU leniency programme and recidivism. Review of Industrial Organization, 48(1), 1-27.
  • Mihai, B. (2008). Cartels–between theory, leniency policy and fines. Annals of Faculty of Economics, 1(1), 549-552.