
Citation: Benoît Perthame, Edouard Ribes, Delphine Salort. Career plans and wage structures: a mean field game approach[J]. Mathematics in Engineering, 2019, 1(1): 38-54. doi: 10.3934/Mine.2018.1.38
[1] | Diogo Gomes, Julian Gutierrez, Ricardo Ribeiro . A mean field game price model with noise. Mathematics in Engineering, 2021, 3(4): 1-14. doi: 10.3934/mine.2021028 |
[2] | Yves Achdou, Ziad Kobeissi . Mean field games of controls: Finite difference approximations. Mathematics in Engineering, 2021, 3(3): 1-35. doi: 10.3934/mine.2021024 |
[3] | Piermarco Cannarsa, Rossana Capuani, Pierre Cardaliaguet . C1;1-smoothness of constrained solutions in the calculus of variations with application to mean field games. Mathematics in Engineering, 2019, 1(1): 174-203. doi: 10.3934/Mine.2018.1.174 |
[4] | Zhiwen Zhao . Exact solutions for the insulated and perfect conductivity problems with concentric balls. Mathematics in Engineering, 2023, 5(3): 1-11. doi: 10.3934/mine.2023060 |
[5] | Boubacar Fall, Filippo Santambrogio, Diaraf Seck . Shape derivative for obstacles in crowd motion. Mathematics in Engineering, 2022, 4(2): 1-16. doi: 10.3934/mine.2022012 |
[6] | Mario Pulvirenti . On the particle approximation to stationary solutions of the Boltzmann equation. Mathematics in Engineering, 2019, 1(4): 699-714. doi: 10.3934/mine.2019.4.699 |
[7] | Pablo Blanc, Fernando Charro, Juan J. Manfredi, Julio D. Rossi . Games associated with products of eigenvalues of the Hessian. Mathematics in Engineering, 2023, 5(3): 1-26. doi: 10.3934/mine.2023066 |
[8] | Simone Paleari, Tiziano Penati . Hamiltonian lattice dynamics. Mathematics in Engineering, 2019, 1(4): 881-887. doi: 10.3934/mine.2019.4.881 |
[9] | Karishma, Harendra Kumar . A novel hybrid model for task scheduling based on particle swarm optimization and genetic algorithms. Mathematics in Engineering, 2024, 6(4): 559-606. doi: 10.3934/mine.2024023 |
[10] | Daniela De Silva, Giorgio Tortone . Improvement of flatness for vector valued free boundary problems. Mathematics in Engineering, 2020, 2(4): 598-613. doi: 10.3934/mine.2020027 |
The mean field game (MFG in short) approach introduced in [1,2,3] (see also [4,5] for recent presentations), describes the relationship between individual decisions and the full population behavior. Here, we use the MFG formalism to analyze the relationship between individual career decisions and the labor market structure. More precisely, we wish to describe the ideal wage structure which allows firms, whatever its size and the employee career stage, to guarantee the individual's optimal career track in terms of his income. Note that if MFG applications to the labor market exists ([6]), and to the general labor theory (see [7] for example on promotion theory), we have not found elements associated to notion of workforce dynamics and labor mobility (see [8] for a review on behavioral economic application to the labor theory). Also MFG have been used in other fields of economics as banks [9], mining [11], see [10] for a survey. Another class of problems related to MFG are the so-called finite horizon planning problem, which has been studied in [12,13,14]. Other related problems have also been explored, as optimal hiring policies from a cost-experience [15] and optimal organizational structure design [16].
Note that the principal difficulty associated with the MFG methodology lays in its inherent analytical complexity. Closed formula or even analytically tractable conclusions are rare and the MFG field heavily relies on numerical solutions [17,18]. However, this is not the case with our present framework. If we provide a couple of general results and properties, we have adopted a more empirical approach. We have indeed directly shaped our problem based on standard assumption on the labor market structure to infer analytic solutions on the wage distribution and the search cost structure. To this extent, the interest of the paper relies on the modeling issue and we leave open the complex mathematical questions which range in the field of inverse problems.
The paper is organized as follows. In the second section, we present our model and detail the intuitions upon which it was constructed. We then compute analytical solutions and discuss results under specific hypotheses. The last section discusses potential implications and shortcomings and offer perspectives on how to expand this research.
Important Legal Remarks. The findings and opinions expressed in this paper are those of the authors and do not reflect any positions from any company or institution.
Career optimization. In a given uniform economy, ignoring outsourcing and offshoring, a labor group
maxv(⋅)E[∫∞te−ρ.(s−t)(ω(t,Q(s),Z(s))−C(s,Q(s),Z(s),v(s)))ds]. | (2.1) |
Here
Labor market hypothesis. We assume that the hierarchical evolution of an individual is assumed to be random and position dependent. This will be transcribed by the following formulation:
dZ(s)=r(s,Z(s),Q(s))ds+σ(Z(s),Q(s))dW(s),Z(s=t)=z, | (2.2) |
Where
The inter firm evolution (on the
dQ(s)=v(s)ds,Q(s=t)=q. | (2.3) |
Based on data that we will review in the last section of this paper, we assume wages to increase with respect to the experience level and the company size (see [19] or [20] for an empirical justification),
∂zω>0,∂qω>0. |
The cost
∂vC(q,z,v)>0,∀q>0, |
which means that cost are incurred because of job changes.
Associated PDEs. This optimization problem eq. (2.1) can be solved by using classical control theory tools such as dynamic programming [22,23]. To do so, we introduce the Bellman's average cost function
Jv(⋅)(t,q,z)=E[∫∞te−ρ.(s−t)(ω(s,Q(s),Z(s))−C(s,Q(s),Z(s),v(s)))ds] | (2.4) |
and
u(t,q,z)=maxv(⋅)Jv(⋅)(t,q,z). | (2.5) |
It is standard that this optimal cost solves the Hamilton-Jacobi-Bellman (HJB in short) equation
∂tu+maxv(Av(u)−ρu+ω(t,z,q)−C(t,z,q,v))=0, | (2.6) |
where
Av(J)=r∂zu+v∂qu+12σ2∂2zzu. | (2.7) |
The solution
ˆv(t,q,z)=ArgMaxv[Av(u)−ρu+(ω(t,z,q)−C(t,z,q,v,r))] | (2.8) |
At any time
Mean field games. Mean field games come when the above optimal control problem is coupled to the full worker population dynamics, which we describe now. Assume that in the labor group
∂tm(t,z,q)+∂q(ˆv(t,q,z)m(t,z,q))+∂z(r(z,q)m(t,z,q))+μm(t,z,q)=12∂2zz(σ2m). | (2.9) |
Here
−12∂z(σ2m)+rm(t,z=0,q)=g(q)∫∫b(t,z,q′)m(t,z,q′)dzdq′+b0(q). | (2.10) |
The framework detailed in the previous section will now be leveraged to understand possible labor market behaviors. Individuals are assumed to know the average wage structure across company size. First, the framework is used in its one dimensional version (e.g. only the company size variable
Assume a one dimensional framework in
C(q,v)=v22F(q)≥0. | (3.1) |
We will sometime use that
∂qu(q)=ˆv(q)F(q). |
Then, the HJB equation can be solved with
−12F(q)(∂qu(q))2+ρu=ω(q). |
Differentiating the above equation with respect to
−ˆv(q)F(q)∂qˆv(q)−12ˆv(q)2∂qF(q)+ρˆv(q)F(q)=∂qω. | (3.2) |
This is a scalar conservation law which is understood in the entropy sense since the HJB equation is understood in the viscosity sense.
The Fokker-Planck equation (2.9) can be used with the appropriate adaptation of the boundary condition to determine the overall working population repartition. We take
{∂q(ˆvm)+μm=0,ˆvm(0)=∫∞0b(q)m(q)dq+b0. | (3.3) |
Note that, because the total population is finite and we expect that
μ∫∞0m(q)dq=∫∞0b(q)m(q)dq+b0. | (3.4) |
The following statement asserts that, as expected for an inverse problem, the appropriate wage structure exists under some conditions on the population distribution
Proposition 3.1. Assume (3.1) and that
∂qln(m)<12∂qln(F), |
0≤μm(q)∫∞qm<vM(q):=μ+ρ12∂qln(F)−∂qln(m), |
then, up to an additive constant, there is an unique wage structure
It has to be noted that adding a constant to
Proof of Proposition 3.1. On the one hand, we can solve the Fokker-Planck equation as an equation on
ˆv(q)=1m(q)[∫∞0b(q)m(q)dq+b0−μ∫q0m]. |
In view of the relation (3.4) this also written
ˆv(q)=μm(q)∫∞qm. | (3.5) |
On the other hand, we may also rewrite (3.3) as
∂qˆv+ˆv∂qln(m)+μ=0. |
Therefore the equation (3.2) can be replaced by a second order polynomial in
ˆv[ˆv∂qln(m)−12ˆv∂qln(F)+μ+ρ]=∂qωF(q), |
that is also written
ˆv[12∂qln(F)−∂qln(m)][vM−ˆv]=∂qωF(q)>0. | (3.6) |
From our assumption, the expression
Example 1. Assume a firm wants to structure its wage to create a specific internal labor market (represented as a continuum of departments of various size
m(q)=m0e−αq,F=F0(1+q)−β,α>0,β>0, |
assuming here that
∂qln(m)=−α,∂qln(F)=−β1+q,ˆv=μα<vM(q)=μ+ρα−β2(1+q), |
Then, the second assumption of Proposition 3.1 is reduced to the above inequality which imposes
β<2α. |
We deduce from the expression (3.6) that
∂qω(q)=F0μα(1+q)−β(ρ+μβ2α(1+q)). |
This wage structure is compatible with a bounded average salary profile obtained as
Example 2. In the case of power laws
m(q)=m0(1+q)−α,F=F0(1+q)−β,α>1,β>0. | (3.7) |
The second assumption of Proposition 3.1 is reduced to
∂qln(m)=−α1+q,∂qln(F)=−β1+q,ˆv=μ(1+q)α−1,vM=(μ+ρ)(1+q)α−β2, |
which imposes
μα−1<μ+ρα−β2⟺μ(1−β2)≤ρ(α−1)⋅ |
We deduce from the expression (3.6) that
∂qω(q)=F0(1+q)−βμα−1(α−β2)(μ+ρα−β2−μα−1). |
and thus
ω(q)=F0μ(α−1)2(ρ(α−1)−μ(1−β2))(1+q)1−β1−β+C,C∈R. |
The condition that the average salary
Example 3. We now ask another question, an individual agent observes the workers population distribution
We assume an exponential decay (resp. increase) for
m(q)=m0e−αq,ω(q)=ω0eβq,α>0,β∈R. | (3.8) |
From an individual perspective, this means that the optimal evolution speed is given by:
ˆv(q)=μαand∂qln(m)=−α. |
Leveraging (3.6), this leads us to the following search cost structure:
∂qF(q)−2αρμF(q)=−2α2βμ2ω0eβq. |
This means that, up to a constant
F(q)=F0e2αρμq+2α2βω02αρμ−βμ2eβq | (3.9) |
We come back to the two dimensional framework in
m(z,q)=m1(z)m2(q),r=r(z)>0,r′(z)<0. | (3.10) |
We are going to solve the problem with separated variables for the solution
∂qu(z,q)=ˆv(z,q)F(q),ˆv1(z)=u1(z),ˆv2(q)=∂qu2(q)F(q), | (3.11) |
where, to define the
Then, the HJB equation can be solved with
u2(q)r(z)∂zu1(z)+12F(q)(∂qu2)2u1(z)2−ρu1(z)u2(q)+ω(z,q)=ρCu. |
Furthermore, differentiating in
∂qω(z,q)=−F(q)ˆv2(q)r(z)∂zu1(z)−12u1(z)2∂q[F(ˆv2)2]+ρu1(z)F(q)ˆv2(q). | (3.12) |
On the other side, the Fokker-Planck equation (2.9) determines
∂z(rm1)m1+ˆv1(z)∂q(ˆv2(q)m2)m2+μ=0, |
which gives, with a free parameter
∂q(ˆv2(q)m2)=−bm2,bˆv1(z)=μ+∂z(rm1)m1. | (3.13) |
Following Section 3.1, this determines explicitly
ˆv2(q)=bm2(q)∫∞qm2,∂qˆv2(q)=−ˆv2(q)∂qln(m2)−b. | (3.14) |
Therefore, using (3.12), we arrive at a non-trivial relationship between the individual strategy and the wage structure
∂qω(z,q)ˆv2(q)F(q)=ρu1(z)−r(z)∂zu1(z)+bu1(z)2−u1(z)2ˆv2(q)[12∂qln(F(q))−∂qlnm2]. | (3.15) |
It is immediate, by its homogeneity, that the above formula does not depend on
Proposition 3.2. We assume that the population distribution
∂z(rm1)+μm1>0, |
that for some constant
ρu1(z)−r(z)∂zu1(z)+u1(z)2≥b1u1(z)2>0,u1(z)=μ+∂z(rm1)m1, |
and assume that, for all
0≤1m2(q)∫∞qm2<vM(q):=b112∂qln(F)−∂qln(m2). |
Then, there is a cost function
Example 4. Let us consider a working population distribution that follows a power law
m(z,q)=m0(z+1)−α1(q+1)−α2,α1α2>1. | (3.16) |
From the relations (3.13), (3.14), we infer that (assuming
ˆv1=μ+r′(z)−α1r(z)z+1,ˆv2(q)=(q+1)α2−1>0. |
Assuming, for simplicity of the calculations, that
r(z)=r0(z+1)ν>0,μ(z)=r0μ0(z+1)ν−1,0>ν>α1−μ0, |
(e.g hierarchal progression speed is convex decreasing in
∂qω=ˆv2ˆv1F[ρ+ˆv1−r∂zln(ˆv1)−ˆv1ˆv2(β2+α2)1q+1]. |
The condition
1+ρ(z+1)1−ν−r0(ν−1)r0(ν−α1+μ0)>β+2α22(α2−1). |
Assume that the same framework as in the previous subsection except that promotions rate are stochastically depending in the hierarchical level
∂qω(z,q)=−F(q)ˆv2r(z)∂zu1(z)−12u1(z)2∂q[F(ˆv2)2]+ρu1(z)∂qu2(q)−F(q)ˆv2∂2zzu1(z). | (3.17) |
The individual strategy
∂z(r1m1)m1+ˆv1(z)∂q(ˆv2(q)m2)m2+μ=12m1∂2zz(σ2m1), |
which gives again two equalities
∂q(ˆv2(q)m2)=−bm2,bˆv1(z)=μ+∂z(r1m1)m1+12m1∂2zz(σ2m1). |
These determine directly
Example 5. Leveraging the example of the Section 3.2, we assume that
σ(z)=√2σ0 (z+1)ν−α1+12. |
The expression of
ˆv1=r0(ν+α1−μ0+σ0ν(ν+1))(z+1)ν−1>0. |
The new condition indeed becomes
∂qω>0⟺1+ρ.(z+1)1−ν−r0(ν−1)r0.(ν−α1−μ0+σ0.ν(ν+1)>(q+1)−α2β+2α22(α2−1)). |
Note that the case with uncertainty is much more complex from an analytical standpoint than the deterministic case from Section 3.2.
Therefore, introducing uncertainty increases movement speed and decreases the firm size threshold
We now calibrate our framework to real world data and explain what the associated findings are in light of the previously developed framework. Note that the data that are used in this section represent the professional services firms (PSFs in short) in the US (NAICS code 54). The public data used in this paper has been extracted from the US Census Bureau 2015 files https://www.census.gov/programs-surveys/susb.html.
The Figure 1 confirms the well established observation that employees distribution follows a power law in firm size. A linear regression on the log-log plane can be performed and leads to:
m(q)≈m∞qα,α≈1.02 | (4.1) |
which is in accordance with the formulas (3.7) in second example of Section 3.1. Note that
The wage increases with the company size as shown in the Figure 2 and we can try to fit the parameter
With the data of the previous subsection, it is possible to estimate the internal dynamics of the US PSFs sector. Assume that
This estimate and the fat tail distribution (4.1) directly translate into an evolution speed
As per the search costs
−F(q)[(μ+ρ)2(α−1)qμ−2α]+∂qF=−2(α−1)2ω0βqβ−3μ2. |
Therefore, we find the expression
F(q)=F1q(μ+ρ)2(α−1)μe−2αq−2(α−1)2ω0βμ2∫q1q(μ+ρ)2(α−1)μ+β−3e−2αqdx. |
As displayed on the Figure 3, search costs are convex in firm size. Interestingly search costs are higher for small firms and then quickly decrease when entering mid sized firms, where they quickly become negligible.
We have seen how individual workers can leverage publicly available information such as wage structure (see Glassdoor) and employment distribution (records can be maintained by countries). Yet this paper shows that the relationship between labor market structure and individual careers is not a one way street. So let's assume that a firm has a target in terms of structure
Assume that the firm employment across departments is distributed according to a power law (size wise). Assume also that from a hierarchical standpoint managerial ranks are also distributed according to a power law. This means that the span of control of a manager is not fixed but gets lower in the higher ranks. These assumptions can be used to describe a flat organization structure. This can be described by the equation (3.16) used in our 4th example. Assume for example that
As we are looking at an internal labor market, let's assume that the promotion speed
This leads to the following evolution internal evolution speeds:
ˆv(z,q)=v0r0(z+1)(q+1). |
This means that the bigger the employee's department is, the slower the lateral progression is. Additionally the higher the employee is in the pyramid, the less chances there is for a lateral move. This raises an interesting question of trade off between lateral and vertical moves. The ratio
Finally assume that search costs are increasing with
∂qω=F0v0r0(z+1)(q+1)3[ρ+r0√v0z+1−r0(z+1)2(ln(r0√v0)−ln(z+1))−3v0r0(z+1)(q+1)2]. |
As depicted in Figure 4, this means that wages have to change at entry level between department to incentivize the required dynamics, but that the higher the employee gets in the hierarchy, the less difference there will be in terms of wage because of department size (e.g. span of control).
Although interesting, note that this example still suffer some limitations. It indeed assumes that the firm is an isolated labor market (e.g wages are not subject to competition). Additionally we assume that external hiring is limited to the bottom of the hierarchical pyramid (e.g
This paper uses the MFG approach to understand the relationships between individual careers and wage structure. We have shown, in a steady state set up how company can internally structure their wage distribution to incentivize mobility. We have also shown that individuals can estimate their job search costs from simple market information. These two questions range in the field of inverse problems for the MFG system of equations.
From the application point of view, our analysis presents a couple of shortcomings. First of all, one may challenge the steady state assumption that was made throughout the model. We believe that finding non steady analytical solutions to the framework will require additional assumptions. Otherwise numerical simulations will most probably be required. Second, the model relies on the idea that workers are interested in maximizing their income. This is quite a narrow view of what may happen in reality. A natural expansion of the framework would be to enhance the workers objective function with some utility consideration. One may want to add leisure considerations for instance or discuss how income taxation may affect labor dynamics. Finally we have made very strong assumptions regarding the structure of search costs. Even though this was helpful to construct our model, we have not found any formal search costs analytical structure in our literature review. As such, we believe that it may be interesting to run an empirical study to better understand the nature of the search costs. This can then quickly be re-embedded into our framework to yield career planning considerations. One may also wish to relax the assumptions of homogeneity in the labor group under consideration and of the full observability of the wage structure among this group.
From the mathematical point of view, one may wish to determine general conditions on the population distribution to ensure that the inverse problem is solvable. In view of the difficulties encountered by the planning problem, we can expect this question raises considerable difficulties also. One may also wish to find a formalism which determines the full wage structure and avoids the undetermined additive constant.
To simplify, we consider the deterministic case and the general notations
ddsX(s,t,x)=R(X(s,t,x),v(s)),X(t,t,x)=x, |
Jv(⋅)(t,x)=∫∞te−ρ(s−t)F(s,X(s,t,x),v(s))ds. |
We compute, using the notation
∂tJ=−F(t,x,v(t))+ρJ+∫∞te−ρ(s−t)∂F(s,X(s,t,x),v(s))∂XQ(s)ds. |
The last term requires some manipulations. The quantity
ddsQ(s)=∂R(X(s,t,x),v(s))∂XQ(s),∂X(t,t,x)∂s+∂X(t,t,x)∂t=0, |
that is also written
ddsQ(s)=∂R(X(s,t,x),v(s))∂XQ(s),Q(s=t)=−R(x,v(t)). |
But we may also compute
ddsˉQ(s)=∂R(X(s,t,x),v(s))∂XˉQ(s),ˉQ(s=t)=Id. |
This means that
Q(s)=−R(x,v(t))ˉQ(s). |
Back to the formula for
∂tJ=−F(t,x,v(t))+ρJ−R(x,v(t))∫∞te−ρ(s−t)∂F(s,X(s,t,x),v(s))∂XˉQ(s)ds |
and we observe that
∂xJ=∫∞te−ρ(s−t)∂F(s,X(s,t,x),v(s))∂XˉQ(s)ds. |
We conclude that
−∂tJ−R(x,v(t))∂xJ+ρJ=F(t,x,v(t)), | (A.1) |
which indicates that this has to be considered as a backward problem, with
The optimal cost is defined as
u(t,x)=maxv(⋅)Jv(⋅)(t,x). |
It is obtained when choosing
−∂tu=maxv[R(x,v)∂xu−ρu+F(t,x,v(t))]. |
For a rigorous derivation, see [4,5,22,23].
B.P. has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant agreement No 740623).
B.P. and D.S. acknowledge partial funding from the ANR blanche project Kibord ANR-13-BS01-0004 funded by the French Ministry of Research.
All authors declare no conflicts of interest in this paper.
[1] | Lasry JM and Lions PL (2006) Jeux à champ moyen. I. Le cas stationnaire. CR Math 343: 619–625. |
[2] | Lasry JM and Lions PL (2006) Jeux à champ moyen. II. Horizon fini et contrôle optimal. CR Math 343: 679–684. |
[3] |
Lasry JM and Lions PL (2007) Mean field games. Jpn J Math 2: 229–260. doi: 10.1007/s11537-007-0657-8
![]() |
[4] | Besoussan A, Frehse J and Yam P (2013) Mean Field Games and Mean Field Type Control Theory, Springer. |
[5] |
Gomes D and Sade J (2014) Mean field games models a brief survey. Dyn Games Appl 4: 110–154. doi: 10.1007/s13235-013-0099-2
![]() |
[6] | Gueant O, Lasry JM and Lions PL (2011) Mean Field Games and Applications. Springer Berlin Heidelberg, Berlin, Heidelberg, 205–266. |
[7] |
Kräkel M and Schöttner A (2012) Internal labor markets and worker rents. Journal of Economic Behavior and Organization 84: 491–509. doi: 10.1016/j.jebo.2012.08.008
![]() |
[8] |
Dohmen T (2014) Behavioral labor economics: Advances and future directions. Labour Economics 30: 71–85. doi: 10.1016/j.labeco.2014.06.008
![]() |
[9] |
Carmona R, Delarue F and Lacker D (2017) Mean field games of timing and models for bank runs. Appl Math Opt 76: 217–260. doi: 10.1007/s00245-017-9435-z
![]() |
[10] |
Achdou Y, Buera FJ, Larsy JM, et al. (2014) Partial differential equation models in macroeconomics. Philos T R Soc A 372: 20130397–20130397. doi: 10.1098/rsta.2013.0397
![]() |
[11] |
Achdou Y, Giraud PN, Larsy JM, et al. (2016) A long term mathematical model for mining industries. Appl Math Opt 74: 579–618. doi: 10.1007/s00245-016-9390-0
![]() |
[12] | Porretta A (2013) On the planning problem for a class of mean field games. CR Math 351: 457– 462. |
[13] |
Porretta A (2014) On the planning problem for the mean field games system. Dyn Games Appl 4: 231–256. doi: 10.1007/s13235-013-0080-0
![]() |
[14] | Lions PL (2013) Cours au collège de france. Technical report, Collège de France. |
[15] |
Doumic M, Perthame B, Ribes E, et al. (2017) Toward an integrated workforce planning framework using structured equations. Eur J Oper Res 262: 217–230. doi: 10.1016/j.ejor.2017.03.076
![]() |
[16] | Perthame B, Ribes E, Salort D, et al. (2017) A model for cost effcient workforce organizational dynamics and its optimization. ArXiv preprint ArXiv:1707.05056. |
[17] |
Achdou Y, Camilli F and Capuzzo-Dolcetta I (2013) Mean field games: convergence of a finite difference method. SIAM J Numer Anal 51: 2585–2612. doi: 10.1137/120882421
![]() |
[18] |
Achdou Y and Porretta A (2016) Convergence of a finite difference scheme to weak solutions of the system of partial differential equations arising in mean field games. SIAM J Numer Anal 54: 161–186. doi: 10.1137/15M1015455
![]() |
[19] | Wiatrowski WJ (2013) Employment-based health benefits in small and large private establishments. |
[20] | Abowd JM and Kramarz F (2000) Inter-industry and firm-size wage differentials: New evidence from linked employer-employee data. Technical report, Cornell University. |
[21] |
Rogerson R, Shimer R and Wright R (2005) Search-theoretic models of the labor market: A survey. Journal of economic literature 43: 959–988. doi: 10.1257/002205105775362014
![]() |
[22] | Bardi M and Capuzzo-Dolcetta I (1997) Optimal control and viscosity solutions of Hamilton- Jacobi-Bellman equations. Birkhäuser Boston. |
[23] | Fleming WH and Soner HM (1993) Controlled Markov Processes and Viscosity Solutions. Vol 25, Springer |
1. | Edouard Ribes, What is the effect of labor displacement on management consultants?, 2021, 1, 2662-9399, 10.1007/s43546-021-00049-6 | |
2. | Carina Geldhauser, Enrico Valdinoci, Optimizing the Fractional Power in a Model with Stochastic PDE Constraints, 2018, 18, 1536-1365, 649, 10.1515/ans-2018-2031 |