Abstract
Our novel game-theoretic Principal/Two-Agent model ensures that the Principal has a reliable internal signal about the Agents’ invested work and effort. Analysing the dominant strategies deductively proves that suboptimal results cannot be prevented with focus on evaluation, implying that quality of work, not its evaluation, must be considered as the most important outcome of the process. The objective of the paper is to establish a new game theoretic model that can be used as a tool for policy makers and managers to motivate Agents and ensure high quality results. Additionally, the model can be used to determine the awards in a company bonus system. The newly developed model is an extended Principal-Agent model with an internal game between two Agents, whose payoff structure can be set in order to ensure truthful implementation of the internal signal to the Principal. By analysing the dominant strategies, we determined the conditions that ensure each Nash equilibrium of the game to manifest the desired outcome. The model presents a novel approach to alignment of interests, for example, in economy, social problems (e.g. policy making for educational process), management, project management, and the mentor-apprentice relationship.
Similar content being viewed by others
References
Aliprantis C, Chakrabarti S (2000) Games and decision making. Oxford University Press, Oxford
Andreoni J (1990) Impure altruism and donations to public goods: a theory of warm-glow giving. Econ J 100:401
Baker M (2016) Is there a reproducibility crisis? Nature 533:353–356
Bebchuk L, Fried J (2004) Pay without performance. Harvard University Press, Cambridge
Becker W, Watts M (1999) How departments of economics should evaluate teaching. Am Econ Rev 89:2
Berrett D (2014) Scholars take aim at student evaluations’ air of objectivity. The Chronicle of Higher Education. http://chronicle.com/article/Scholars-Take-Aim-at-Student/148859/. Accessed 15 May 2015
Bhattacharyya N (2004) Student evaluations and moral hazard. J Acad Ethics 2:263–271
Bokal D, Steinbacher M (2019) Phases of psychologically optimal learning experience: task-based time allocation model. Cent Eur J Oper Res 27:3
Bowles S (2016) The moral economy: why good incentives are no substitute for good citizens. Yale University Press, New Haven
Brown B, Saks D (1987) The microeconomics of the allocation of teachers’ time and student learning. Econ Educ Rev 6:4
Chang C (2013) Principal-agent model of risk allocation in construction contracts and its critique. J Constr Eng Manag 140:04013032
Dembe A, Boden L (2000) Moral hazard: a question of morality? New Solut J Environ Occup Health Policy 10(3):257–279
Dietl H, Grossmann M, Lang M, Way S (2013) Incentive effects of bonus taxes in a principal-agent model. J Econ Behav Organ 89:93–104
Ferris J (1992) School-based decision making: a principal-agent perspective. Educ Eval Policy Anal 14(4):333–346
Harris M, Raviv A (1978) Some results on incentive contracts with application to education and employment, health insurance, and law enforcement. Am Econ Rev 68:20–30
Holt C, Roth A (2004) The nash equilibrium: a perspective. PNAS 101:3999–4002
Johnson V (2003) Grade inflation: a crisis in college education. Springer, New York
Krautmann A, Sander W (1999) Grades and student evaluations of teachers. Econ Educ Rev 18(1):59–63
Laffont J, Martimort D (2009) The theory of incentives: the principal-agent mode. Princeton University Press, Princeton
Lawrence J, Celis S, Ott M (2014) Is the tenure process fair? What faculty think. J High Educ 85:2
Lazear E, Rosen S (1981) Rank-order tournaments as optimum labor contracts. J Polit Econ 89:841–864
Lichty R, Vose D, Peterson J (1978) The economic effects of grade inflation on instructor evaluation: an alternative interpretation. J Econ Educ 10:3–11
Lin T (2009) Application of a static game of complete information: economic behaviors of professors and students. Econ Bull 29:1678–1686
Liu H, Liu S (2013) Salesperson compensation contract design based on multi-agent principal agent model. Nankai Bus Rev Int 4(1):78–86
Maroulis S, Guimerà R, Petry H, Stringer M, Gomez L, Amaral L, Wilensky U (2010) Complex systems view of educational policy research. Science 330:38–39
McAfee R, McMillan J (1996) Competition and game theory. J Mark Res 33(3):263–267
Neal D (2011) The design of performance pay in education. http://www.nber.org/papers/w16710.pdf
Newman H, Wright D (1992) Negligence versus strict liability in a principal-agent model. J Econ Bus 44(4):265–281
Poundstone W (2007) How would you move mount Fuji. Back Bay Books, New York
Uttl B, White C, Gonzal D (2017) Meta-analysis of faculty’s teaching effectiveness: student evaluation of teaching ratings and student learning are not related. Stud Educ Eval 54:22–42
Wachtel H (1998) Student evaluation of college teaching effectiveness: a brief review. Assess Eval High Educ 23:191–212
Waterman R, Meier K (1998) Principal-agent models: an expansion? J Public Adm Res Theory 8(2):173–202
Acknowledgements
Drago Bokal was funded in part through Slovenian Research Agency grants J1-8130, J1-2452, and P1-0297.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Proof of Theorem 1
Appendix: Proof of Theorem 1
Lemma 1
If for each \(x\in \lbrace A,B \rbrace \), \(df_x\) and \(dg_x\) are strictly positive numbers and that \(e_x>o_x>u_x>i_x \) , then the following is true:
The pairs of strategies \( ( \lbrace C\; H, C\; L \rbrace , \lbrace C\; \cdot \rbrace ) \) are pure-strategy Nash equilibria of the Principal/Two-Agent model with internal signal, if and only if the following conditions hold:
Proof
First, we assume that (i, j) is a pure-strategy Nash equilibrium for each \(i \in \{1,2\}\) and \( j \in \{ 1,\ldots ,8\}\), i.e. all pairs resulting in the desired outcome I.
-
As the strategy (1, 2) is a pure-strategy Nash equilibrium, it dominates the strategy (5,2). Therefore, \(\pi _A(1,2)\ge \pi _A(5,2)\), expanded \(f'_A+g_A+e_A \ge f_A+g_A+o_A\), implying (5).
-
As the strategy (2, 1) is a pure-strategy Nash equilibrium, it dominates the strategy (2, 9). Therefore, \(\pi _B(2,1)\ge \pi _B(2,9)\), expanded \(f_B'+g_B+e_B \ge f_B+g_B+u_B\), implying (6).
For the converse implication, we assume (5) and (6). For each (i, j), \(i\in \{1,2\}\), \(j\in \{ 1,\ \ldots ,\ 8 \}\), and for each \(k\in \{1,\ \ldots ,\ 8\},\) we need to prove
and, for each \(l \in \{ 1,\ \ldots ,\ 16\},\) also
-
1.
We prove (1) as follows:
-
(a)
The equality \(\pi _A(i,j)= \pi _A (k,j)\) holds for \(i,k \in \{ 1,2\}\), and \(j\in \{1,\ \ldots ,\ 8\}. \)
-
(b)
(5) implies \(f_A'+g_A+e_A> f_A+g_A+o_A\) and therefore \(\pi _A(i,j)> \pi _A(k,j)\) for \(i\in \{1,2\}\), \(j \in \{2\}\), and \(k \in \{5,6\}\).
-
(c)
\(dg_A> 0\) implies \(f_A'+g_A+e_A> f_A'+g_A'+e_A\), therefore \(\pi _A(i,j)> \pi _A(k,j)\) for \(i\in \{1,2\}\), \(j \in \{1,\ \ldots ,\ 8\}\), and \(k \in \{3,4\}\).
-
i.
\(dg_A>0\) with (5) implies \(f_A'+g_A+e_A>f_A+g_A'+o_A\), therefore \(\pi _A(i,j)> \pi _A(k,j)\) for \(i\in \{1,2\}\), and for \(j \in \{1,3,4\}\) and \(k \in \{5,\ \ldots ,\ 8\}\), as well as for \(j \in \{2\}\) and \(k\in \{7,8\}\).
-
ii.
\(e>o>u>i\) implies \(dei_A> deo_A\). \(dei_A> deo_A\) with (5) implies \(f_A'+g_A+e_A>f_A+g_A+i_A\), therefore \(\pi _A(i,j)> \pi _A(k,j)\) for \(i\in \{1,2\}\), \(j \in \{6\}\), and \(k \in \{6,8\}\).
-
A.
\(dg_A>0\) and \(dei_A> deo_A\) with (5) implying \(f_A'+g_A+e_A>f_A+g_A'+i_A\), therefore \(\pi _A(i,j)> \pi _A(k,j)\) for \(i\in \{1,2\}\) and for \(j \in \{6\}\) and \(k \in \{5,7\}\); as well as for \(j \in \{5,7,8\}\) and \(k \in \{5,\ \ldots ,\ 8\}\).
-
A.
-
i.
-
(a)
-
2.
We prove (8) as follows.
-
(a)
The equality \(\pi _B(i,j)= \pi _B (i,l)\) holds for \(i \in \{ 1,2\}\) and \(j,l \in \{1,\ \ldots ,\ 8\}. \)
-
(b)
(6) implies \(f_B'+g_B+e_B>f_B+g_B+u_B \), therefore \(\pi _B(i,j)> \pi _B (i,l)\) for \(i \in \{ 2\}\), \(j\in \{1,\ \ldots ,\ 8\}\), and \(l\in \{ 9,\ \ldots ,\ 16 \}\).
-
i.
\(dg_B>0\) with (6) implies \(f'_B+g_B+e_B>f_B+g_B'+u_B\), therefore \(\pi _B(i,j)> \pi _B (i,l)\) for \(i \in \{ 1\}\), \(j\in \{1,\ \ldots ,\ 8\}\), and \(l\in \{ 9,\ \ldots ,\ 16 \}\).
-
i.
-
(a)
Since the strategies (i, j) for \(i\in \{1,2\}\) and \(j \in \{1,\ \ldots ,\ 8\}\) dominate respective rows and columns, and have equal utilities by themselves, each of these pairs is a pure-strategy Nash equilibrium. Table 5 summarizes the applied arguments. \(\square \)
In Theorem 1, we identify sufficient and necessary conditions for these pairs to be the only Nash equilibria, hence eliminating any rational desire for other pairs of strategies to become stable behaviour.
Theorem 1
If, for each \(x\in \lbrace A,B \rbrace \), \(df_x\) and \(dg_x\) are strictly positive numbers and \(e_x>o_x>u_x>i_x \) , then the following is true:
The pairs of strategies \( ( \lbrace C\; H, C\; L \rbrace , \lbrace C\; \cdot \rbrace ) \) are the only pure strategy Nash equilibria of the Principal/Two-Agent model with internal signal, if and only if all of the following conditions hold:
Proof
Using the notation form of the proof of the previous lemma, we assume that (i, j) is a pure-strategy Nash equilibrium precisely for each \(i \in \{1,2\}\) and \( j \in \{ 1,\ \ldots ,\ 8\}\), and we prove the conditions (9), (10), (11), and (12):
-
The strategy (5, 11) is not a pure-strategy Nash equilibrium. Therefore, one of the following conditions should be satisfied:
-
(a)
\(\pi _A (i,11) > \pi _A (5,11)\) for \(i\in \{1,\ \ldots ,\ 4\}\), expanded \(f_A'+g_A'+u_A> f_A+g_A'+o_A\);
-
(b)
\(\pi _A (i,11) > \pi _A (5,11)\) for \(i\in \{6,\ \ldots ,\ 8\}\), expanded \(f_A+g_A'+o_A> f_A+g_A'+o_A\);
-
(c)
\(\pi _B (5,j) > \pi _B (5,11)\) for \(j\in \{1,2,3,4,9,10,12\}\), expanded \(f_B'+g_B+o_B> f_B'+g_B+o_B\);
-
(d)
\(\pi _B (5,j) > \pi _B (5,11)\) for \(j\in \{5,6,7,8,13,14,15,16\}\), expanded \(f_B+g_B'+i_B> f'_B+g_B+o_B\).
Conditions (b) and (c) are false, and Condition (a) is in contradiction with the assumptions \(df_A>0\) and \(dou_A>0\). Therefore, (d) holds, implying (12).
-
(a)
-
The strategy (6, 14) is not a pure-strategy Nash equilibrium. Therefore, one of the following conditions should be satisfied:
-
(a)
\(\pi _A (i,14) > \pi _A (6,14)\) for \(i\in \{1,3\}\), expanded \(f_A'+g_A'+u_A> f_A+g_A+i_A\);
-
(b)
\(\pi _A (i,14) > \pi _A (6,14)\) for \(i\in \{2,4\}\), expanded \(f_A'+g_A+u_A> f_A+g_A+i_A\);
-
(c)
\(\pi _A (i,14) > \pi _A (6,14)\) for \(i\in \{5,7\}\), expanded \(f_A+g_A'+i_A> f_A+g_A+i_A\);
-
(d)
\(\pi _A (i,14) > \pi _A (6,14)\) for \(i\in \{8\}\), expanded \(f_A+g_A+i_A> f_A+g_A+i_A\);
-
(e)
\(\pi _B (5,j) > \pi _B (6,14)\) for \(j\in \{1,2,3,4,9,10,11,12\}\), expanded \(f'_B+g_B+o_B> f_B+g_B+i_B\);
-
(f)
\(\pi _B (5,j) > \pi _B(6,14)\) for \(j\in \{5,6,7,8,13,15,16\}\), expanded \(f_B+g_B+i_B> f_B+g_B+i_B\).
Conditions (d) and (f) are false, and Condition (c) is in contradiction with the assumption \(dg_A>0\). Condition (e) is in contradiction with (12). Condition (b) implies (a), therefore (a) holds, implying (10).
-
(a)
-
The strategy (4, 16) is not a pure-strategy Nash equilibrium. Therefore, one of the conditions:
-
(a)
\(\pi _A (i,16) > \pi _A (4,16)\) for \(i\in \{1, 3\}\), expanded \(f'_A+g_A'+u_A> f'_A+g_A+u_A\);
-
(b)
\(\pi _A (i,16) > \pi _A (4,16)\) for \(i\in \{ 2\}\), expanded \(f'_A+g_A+u_A> f'_A+g_A+u_A\);
-
(c)
\(\pi _A (i,16) > \pi _A (4,16)\) for \(i\in \{5,\ \ldots ,\ 8\}\), expanded \(f_A+g_A'+i_A> f'_A+g_A+u_A\);
-
(d)
\(\pi _B (4,j) > \pi _B (4,16)\) for \(j\in \{1,\ \ldots ,\ 8\}\), expanded \(f'_B+g_B'+e_B> f_B+g_B+u_B\);
-
(e)
\(\pi _B (4,j) > \pi _B (4,16)\) for \(j\in \{9,\ \ldots ,\ 16\}\), expanded \(f_B+g_B+u_B> f_B+g_B+u_B\); should be satisfied.
Conditions (b) and (e) are false, and Condition (a) is in contradiction with the assumption \(dg_A>0\). Condition (c) is in contraction with (10). Therefore, (d) must hold, implying (11).
-
(a)
For the converse implication, we assume the conditions (9), (10), (11), and (12), and we need to prove that (i, j), where \(i\in \{1,2\}\) and \(j\in \{1,\ \ldots ,\ 8\}\) are the only Nash equilibria. Therefore, for each \((m,n)\notin \{1,2\}\times \{1,\ \ldots ,\ 8\}\), we establish one of
or
-
1.
Condition (13) is satisfied for the pairs of strategies in the following cases:
-
(a)
(10) implies \(f'_A+g_A'+u_A>f_A+g_A+i_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{1,3\}\), \(l \in \{14\}\), and \( m \in \{6,8\}\).
-
(b)
\(dg_A> 0\) implies \( f'_A+g_A+e_A>f'_A+g_A'+e_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{1,2\}\), \(l \in \{1,\ \ldots ,\ 8\}\), and \(m \in \{3,4\}\).
-
i.
\(dg_A>0\) with (9) implies \( f'_A+g_A+e_A>f_A+g_A'+o_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{1,2\}\) and either \(l \in \{1,3,4\}\) and \(m \in \{5,\ \ldots ,\ 8\}\); or \(l \in \{2\}\) and \(m \in \{7,8\}\).
-
ii.
\(dg_A> 0\) with (10) implies \( f'_A+g_A'+u_A>f_A+g'_A+i_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{2,4\}\) and either \(l \in \{13,15,16\}\) and \( m \in \{5,\ \ldots ,\ 8\}\); or \(l \in \{14\}\) and \( m \in \{5,7\}\).
-
i.
-
(c)
\(dei_A> deo_A\) with (9) implies \( f'_A+g_A+e_A>f_A+g_A+i_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{1,2\}\), \(l \in \{6\}\), and \(m \in \{6,8\}\).
-
i.
\(dg_A>0\) and \(dei_A> deo_A\) with (9) imply \( f'_A+g_A+e_A>f_A+g'_A+i_A\), therefore \(\pi _A(k,l)> \pi _A(m,l)\) for \(k\in \{1,2\}\) and both \(l \in \{6\}\) and \(m \in \{5,7\}\), as well as \(l \in \{5,7,8\}\) and \(m \in \{5,\ \ldots ,\ 8\}\).
-
i.
-
(a)
-
2.
Condition (14) is satisfied for the strategies in the following cases:
-
(a)
(12) implies \( f_B+g'_B+i_B>f'_B+g_B+o_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k\in \{5\}\), \(l \in \{5,\ \ldots ,\ 8\}\) and \(n\in \{ 2,9,10,11,12\}\).
-
i.
\(dg_B>0\) with (12) implies \( f_B+g'_B+i_B>f'_B+g'_B+o_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k\in \{7\}\), \(l\in \{5,\ \ldots ,\ 8\}\), and \(n\in \{ 9,\ \ldots ,\ 12\}\).
-
ii.
\(dg_B>0\) with (12) implies \( f_B+g_B+i_B>f'_B+g'_B+o_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k\in \{8\}\), \(l\in \{5,\ \ldots ,\ 8\}\), and \(n\in \{ 9,\ \ldots ,\ 12\}\).
-
iii.
\(dg_B>0\) with (12) implies \( f_B+g_B+i_B>f'_B+g_B+o_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k\in \{6\}\), \(l\in \{5,\ \ldots ,\ 8\}\), and \(n\in \{ 2, 9,10,11,12\}\).
-
i.
-
(b)
(11) and \(dg_B>0\) imply \( f'_B+g_B+e_B>f_B+g_B+u_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k \in \{2\}\), \(l\in \{1,\ \ldots ,\ 8\}\), and \(n\in \{9,\ \ldots ,\ 16\}\).
-
(c)
(11) implies \( f'_B+g_B'+e_B>f_B+g_B+u_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k \in \{ 4\}\), \(l\in \{1,\ \ldots ,\ 8\}\), and \(n\in \{9,\ \ldots ,\ 16\}\).
-
i.
\(dg_B>0\) with (11) implies \( f'_B+g_B+e_B>f_B+g'_B+u_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k \in \{ 1\}\), \(l\in \{1,\ \ldots ,\ 8\}\), and \(n\in \{ 9,\ \ldots ,\ 16 \}\).
-
ii.
\(dg_B>0\) with (11) implies \( f'_B+g'_B+e_B>f_B+g'_B+u_B\), therefore \(\pi _B(k,l)>\pi _B(k,n)\) for \(k \in \{ 3\}\), \(l\in \{1,\ \ldots ,\ 8\}\), and \(n\in \{ 9,\ \ldots ,\ 16 \}\).
-
i.
-
(a)
We have shown for each \((m,n)\notin \{1,2\}\times \{1,\ \ldots ,\ 8\}\) that the pair of strategies (m, n) is not a Nash equilibrium, and we summarize the arguments in Table 6.
Note that if all the pure-strategy Nash equilibria exhibit the desired outcome, so will all mixed-strategy Nash equilibria. Thus each Nash equilibrium of the Game yields the desired outcome, whenever the necessary and sufficient conditions from Theorem 1 are satisfied.
Rights and permissions
About this article
Cite this article
Smole, A., Jagrič, T. & Bokal, D. Principal/Two-Agent model with internal signal. Cent Eur J Oper Res 29, 791–808 (2021). https://doi.org/10.1007/s10100-020-00719-0
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10100-020-00719-0