E ort , Satisfaction and Outcomes in Organisations

In this paper, an agent-based model of bounded-rational agents, who adapt both their e ort intensity (by the interaction with other employees) and their stay-on-the-job-intention (by the alignment of their personal values with the Human-Resource Management (HRM) practices implemented by the organisation), is proposed. Ouraim is toanalyse: (i) theemergenceofanorganisational cultureand its relationshipwithboth formal organisational structures andemployees’ e ort-behaviours; (ii) the increaseof organisational performance by retaining valuable-performance employees whereas poor-performance employees are dismissed. We have obtained that: (i) Some possible combinations of both employees-e ort behaviours and formal organisational structures can favour the emergence of organisational cultures more than others; (ii) The interaction between employees within matrix structures (balanced or strong) with a democratic team leadership favour the emergence of organisational cultures; (iii) High-e ortmanagers are relevant for the emergence of high-performance organisational cultures; (iv) Turnover (voluntary or involuntary) a ects the emergence of organisational culture negatively. We conclude that the main challenge is "to retain high e ort managers by adapting the set of HRM practices to them".


Introduction
. Recently, employees are o en considered a source of competitive advantage for organisations.To explain this, Guest ( ) argued that there was a need for ( ) a theory on Human Resource Management (HRM), ( ) a theory on performance, and ( ) a theory on how these aspects are linked.Nevertheless, a er reviewing empirical and theoretical research over eight years, Paauwe et al. ( ) observed only modest progress on these three fundamental issues by reviewing empirical and theoretical research over years.In order to drive more longitudinal research, more complex causal models, and multi-level phenomena, Wright et al. ( ) suggested to consider the following concepts (that are o en missing in the analyses of the relationship between HRM and performance): time, cause and individuals.Therefore, in order to consider the variance across individuals, we propose to use an agent-based approach to analyse both its cause and e ect on the relationship between HRM and performance.Although agent-based modelling may bring opportunities and advantages to management and organisations, management seems to hold back to the agent-based model revolution (Secchi et al. ).
. In this paper, we propose an agent-based model (ABM) to link HRM and performance inspired on Guest ( )'s model within formal organisational structure frameworks (see Figure ).Our aim is to analyse: (i) the emergence of an organisational culture and its relationship with both formal organisational structures and employees' e ort-behaviours; (ii) the increase of organisational performance by retaining valuable-performance employees whereas poor-performance employees are dismissed.Organisational structure definitions are as divergent as the approaches to study it.Due to ABMs consist of agents that interact within an environment (Gilbert ), our approach is aligned with Dow ( ) who defined organisational structure as "a relatively stable, either planned or spontaneous, pattern of actions and interactions that organisation members undertake for the purpose of achieving the organisation's goals".Therefore, in this paper, the organisational structure dictates the interactions between individuals.

.
Organisational structure can be categorised as formal (planned) and informal (spontaneous).In this paper, we focus on formal organisational structure.It ranges from functional to projects, with a variety of matrix structures between them (Galbraith ).In a functional organisational structure, the employees are grouped into departments (for instance, engineering, manufacturing and, marketing) and functional areas (electrical and mechanical areas in engineering department), whilst in project structures the employees are grouped into teams.In matrix organisational structure (which can be weak, balance or strong) the employees are grouped into both departments and teams.Salas et al. ( ) summarised down what researchers know about teamwork into "big five" components: team leadership, mutual performance monitoring, backup behaviour, adaptability, and team orientation.In this paper, we focus on the team leadership component because it dictates the formal interactions between team members.

Human Resource Management (HRM) practices
.
The HRM practices (such as job design, human resource planning, recruitment and selection, performance appraisal, promotion, work-family balance programs, etc.) enhance organisational performance.Dyer et al. ( ) stated four levels of HRM outcomes: employee, organisational, financial and market, but we limit to the three levels captured by Guest ( )'s model: financial outcomes (such as profits and return on investment), organisational outcomes (such as productivity and turnover) and employee outcomes.Employee outcomes consist of attitudinal outcomes (such as commitment, job satisfaction and intention to leave) and behavioural outcomes (such as e ort, motivation, cooperation and organisational citizenship). .
In order to model the black boxes of Guest ( )'s model, we have based it on previous ABMs research (Dal Forno and Merlone , ) for the behavioural-outcomes box and on HRM empirical research for the relationship between boxes (Gri eth et al. ; Judge et al. ; Christen et al. ).However, in order to design a model as simple as possible, neither the strategy box nor the financial box have been modelled and not all outcomes of the other boxes have been included in our agent-based model.

Attitudinal outcomes: Satisfaction
. Job satisfaction refers to how content an individual is in a job.It is measured within empirical studies either as an overall (global) feeling about the job, or as a sum of attitudes about various aspects of the job (facet approach) (Spector ).Facets of job satisfaction include rewards (such as pay or fringe benefits), relationship with other people (such as co-workers or supervisor), the nature of the work itself, personal growth, promotion opportunities, recognition, security and facilities, etc.In this paper, we follow the overall approach in order to model job satisfaction.

.
There seems to be a wide consensus among researchers that job satisfaction influences the intention to leave the organisation negatively and the employee's well-being positively (Gri eth et al. ).

Behavioural outcomes: E ort .
Many empirical studies treat e ort as part of job performance.This situation is probably due to the di iculties associated with defining and measuring e ort.From a firm's perspective, e ort and job performance may be di icult to distinguish, and e ort is o en inferred from job performance.From an employee's perspective, job performance and e ort are di erent: e ort is an input to work, and job performance is an output from this e ort (Christen et al. ).Nevertheless, the ABM approach allows to distinguish them.

.
No consensus seems to be exist about the e ort-satisfaction relationship.If e ort is costly for an employee, from a theoretical argument, it should have a negative direct e ect on job satisfaction.This implies that there is a conflict of interest between the employer (who wants the employee to work hard) and the employee (who wants the salary with the minimum possible e ort) (Christen et al. ).However, empirical studies that include e ort as an antecedent tend to find that it has a positive e ect on job satisfaction.Christen et al. ( ) solved this inconsistency between theoretical arguments and empirical findings.They found that e ort has indirect positive e ect on job satisfaction through the mediation of job performance and a direct positive e ect on job performance.When the e ect of job performance is eliminated from the job satisfaction equation, Christen et al. ( ) obtained a negative but insignificant negative e ect of e ort on job satisfaction, being consistent with the assumption that e ort is costly to an agent.Therefore, we do not model a direct e ort-satisfaction relationship. .
In order to model e ort we follow Dal Forno and Merlone ( , )'s ABM.Dal Forno and Merlone ( , ) proposed a multi-agent simulation platform for modeling perfectly rational and bounded rational agents in organisations within an informal organisational structure in order to evaluate the emergence of culture (i.e., an equilibrium where all agents exert the same e ort).In this paper, we consider some bounded rational agent considered by Dal Forno and Merlone ( ), but interacting within di erent formal organisational structures instead of an informal organisational structure.
Performance refers to the degree of achievement of the mission at the work place that builds up an employee (Cascio ).As multiple levels of performance analysis (e.g., individual, group, and organisational) are needed (Wright et al. ), ABM is especially suitable approach to understand the relationship between performance and the other boxes of Guest ( )'s model by considering the variance across individuals. .

Christen et al. (
) obtained that job performance (i.e, performance at individual level) is determined by the employee's e ort and ability.Therefore, we consider that e ort has a direct positive e ect on job performance, but not a direct e ect on job satisfaction.

.
Nevertheless, the performance-satisfaction relationship is a controversial theme.By reviewing empirical research conducted in the th century, Judge et al. ( ) found at least seven di erent ways in which the performance-satisfaction relationship has been specified.Some of these models view the relationship between job satisfaction and job performance as unidirectional (i.e., either job satisfaction causes job performance or vice versa), some of these models view a false relationship or no-relationship, and others view a bidirectional relationship moderated by other variables (such as reward contingency, intrinsic job characteristics, self-esteem, organisational tenure, etc).In order to model job performance, we consider that job satisfaction causes job performance.

.
There seems to be a consensus among researchers that organisational performance (i.e, performance at organisational level) is determined by the HRM practices implemented by the organisation.In this paper, organisational performance emerges from the interactions between individuals within di erent organisational structures.
Many models have been proposed to explain turnover since the original model by March and Simon ( ), where turnover is defined as "the departure of an employee from the formally defined organisation".Watrous et al. ( ) categorise turnover as voluntary or involuntary, as well as functional or dysfunctional.
. Involuntary turnover, which is the process in which the organisation assumes the control over an employee's decision to leave or to stay (and decides to dismiss the employee), is o en labeled as functional turnover because under-performing employees or those who do not fit within the organisational culture are removed and replaced by relatively higher performing new employees.Therefore, involuntary turnover would typically be associated with lower poor performing employees.Dal Forno and Merlone ( )'s demonstrated that, in most situations, by firing under performance employees may increase overall employees' e ort, but it is not guaranteed.In this paper, we follow Dal Forno and Merlone ( )'s agent-based research in order to model involuntary turnover.

.
Voluntary turnover, which is the process in which an individual makes the decision to leave or to stay in the organisation (and decides to leave), is usually dysfunctional because it can be most detrimental to the organisation.Empirical evidence supports the dysfunctional e ects in sense of being associated with decrements in organisational performance (Hausknecht and Trevor ).Research typically focuses on the antecedents of the individual employee's decisions to leave, including employee demographics variables, job attitudes variables (such as job satisfaction and organisational commitment), and the ease of movement (see Gri eth et al. ( ) for a review of turnover research conducted in the th century).Alternative models have been proposed in order to better explain why people leave organisations (Lee et al. ) or why people stay in organisations (Mitchell et al. ).In this context, the term "employee retention", which refers to Human Resource Management (HRM) Practices implemented by companies to prevent valuable employees from leaving the organisation, began to appear as opposite side of the same coin (Holtom et al. ).There is no previous agent-based research related to voluntary turnover.

Organisational outcomes: A shared culture of e ort .
Organisational culture typically is defined as a complex set of values, beliefs, assumptions and symbols that define the way in which a firm conducts its business (Barney ).In this paper, we focus on a shared culture of e ort.It is not imposed by the organisation, but emerges from the interaction between employees.We consider that organisational culture emerges when all agents exert the same e ort (Dal Forno et al. ).In strong organisational cultures, all employees exert the same level of e ort to increase the overall organisational performance, whilst weak cultures may create diversity between the individual and organisational objectives.

.
The paper is organised as follows.In Section the agent-based model is described.A detailed description of the model, using the ODD protocol proposed by Grimm et al. ( ), is available at http://www.eii.uva.es/posada/HURLAB-0.html.In Section the computational experiment settings and some results are reported.Discussion and concluding remarks are pointed out in Sections and .

HURLAB: An Agent-Based Model
. HURLAB (HUman Resource LABoratory) is an agent-based model populated by two types of bounded-rational agents: organisation and employees (see Figure for an overview of the model).In order to distinguish the level of performance analysis (individual, group, and organisational), the boxes of Guest ( )'s model are redistributed according to the type of bounded-rational agent involved (organisation or employee).Lines and signs indicate the direction of e ects, whereas indirect e ects are represented by a dashed line.

.
The rates resulting from the decisions taken, either by the employees or by the organisation, are involuntary turnover (i.e., the percentage of employees who involuntarily leave the organisation) and voluntary turnover (i.e., the percentage of employees who voluntarily leave the organisation).

The organisation .
The goal of the organisation is to achieve the maximum level of average performance with full occupation.Therefore, when an employee leaves the organisation (either voluntary or involuntarily) or retires, they are replaced by other one assuming some organisational replacement costs.The organisation is defined by the following variables: formal organisational structure (which includes the type of configuration and the description of the structure), team (such as team size, team leadership, number of projects per member and number of projects per project-manager), Human-Resource Management (HRM) Practices and outcomes (such as group performance and firm performance) (see Table ).
. Formal organisational structure: type configuration and levels.We define four levels from the general manager level (level ) to lowest level (level ), which define the employee's job status, within the following configuration types: functional, balanced matrix, and strong matrix configurations.The interactions between the employees (represented by solid lines in Figure ) will be determined by both the type of formal organisational structure and the type of team leadership.
. Therefore, in a functional, balanced matrix and strong matrix configurations, department-managers report to (and interacts with) their general-manager, functional-managers report to (and interact with) their departmentmanager and area-managers report to (and interact with) their functional-manager.Workers of each area do not interact with the other workers of their own area due to only formal interactions are considered in this paper.
Only in teams, workers interact between each other, but limited by both their role in the team and the team leadership.
. In a balanced matrix and strong matrix configurations, workers are grouped into teams (represented by dashed lines in Figure ).Individuals working in teams play one of the following two roles: project-manager (who coordinates the project and is represented by a black-person in Figure ) or team-member.Project-managers report to (and interact with) their functional manager in balanced matrix, but to a multi-project-manager in strong matrix.They do not interact between each other.Team-members have to report to (and interact with) their functional manager as well as multiple project-managers (as many managers as projects each employee is involved into) in balanced and strong matrix.
. Team leadership.Project-managers can exhibit a leadership behaviour between two extreme types: egocentered and democratic.In ego-centered leadership (represented by a star-shaped network), the projectmanager interacts with all team members, but members do not interact between each other.In democratic leadership (represented by a whole network), the project-manager and all team-members interact between each other.
. Team size.Research on team size suggests a curvilinear relationship between team size and team e ectiveness.Very small teams (i.e., or people) may lack diversity of perspectives, whereas larger teams (more than .Outcomes: firm performance and group performance.Firm performance is calculated as the average of individuals' job performance.Group performance is calculated as the average of individuals' job performance grouped either into a functional-area or into a team.

Decisions .
In relation to involuntary turnover, the organisation dismisses employees whose e ort is lower than a minimume ort-to-stay level (called "firing-level" by Dal Forno et al. ), with the objective of achieving the maximum level of performance.

.
Besides, related to voluntary turnover, the organisation implements some HRM practices with the objective of both influencing the employees' satisfaction and retaining high-performing employees.This decision does not change over time.
Employee-agents are defined by the following variables: their personal characteristics (such as gender, age and family status), professional characteristics (such as team role, tenure and job status), attitude and behaviour (such as e ort and satisfaction) and outcomes (such as job performance and intention to leave) (see Table ).In this paper, personal and professional characteristics are randomly initialised but, in order to be applied to real organisations, real data can be loaded. .
Employees dynamically interact both between each other and with the organisation, through a set of behavioural rules which determines the e ort intensity (related to involuntary turnover) and the job satisfaction (related to voluntary turnover).Employee's e ort.Employees di er in their e ort intensity.Moreover, the e ort exerted by each employee can vary in terms of intensity and persistence over their professional life, adapting their e ort intensity by considering both their behaviour and their interactions with other employees.
. We consider the following bounded-rational behaviours considered by Dal Forno and Merlone ( ), which were extrapolated from human-subject experiments, but the e ort intensity ranges from to instead of to : Null e ort: this agent exerts always the same, almost null e ort (between and ); Shrinking e ort: this agent halves the e ort provided by its last partner; Replicator: this agent exerts the same e ort as its last partner exerted in the previous interaction; High e ort: this agent exerts always the same high e ort (between and ); E ort comparator: this agent compares its e ort to its previous partner's e ort; it increases its e ort by % if it is lower than its partner's one and vice versa; Averager: it averages its e ort with last partner's e ort.The idea behind is that e ort is contagion.
. With whom employees interact with, is defined by their position in the organisational configuration.Changes in e ort intensity are observed by using a graded chromatic scale, which ranges from blue (low e ort) to red (high e ort).

.
Employee's satisfaction.The employee's satisfaction is estimated by assessing how much the HRM practices implemented by the organisation are aligned with their career goals and personal values.This feeling, which is private, changes over the employee's life and depends on both the personal characteristics (such as age, gender and family status) and the HRM practices implemented by the organisation.To measure it, we propose an endorsement scheme grounded on a methodology proposed by Cohen ( ) to reproduce heuristic reasoning: where j are the possible HRM Practices, b is the employees' sensitivity to the organisation implements HRM practices and ranges from (no sensitivity) to , s j is the relevance of the HRM−j practice to the employee−i and ranges on a -point Likert scale from (not a priority) to (essential), τ (j) is the implementation of the HRM−j practices by the organisation (τ (j) = 1 when the HRM−j practice is implemented and τ (j) = −1 when it is not), and S max (i) is the employee−i's satisfaction when all HRM practices valued by the employee are implemented by the organisation.In other words, employees sort in their minds a set of HRM practices into relevance order according to their personal characteristics.When one of these HRM practices is (is not) implemented by the organisation, employees have a satisfaction (dissatisfaction) feeling whose value depends on both s j and b.
. For instance, when only two HRM practices can be implemented by the organisation, the following situations can take place: the organisation either implements HRP − j or not (no-HRP − j), and the employee−i considers the HRP − j practice either important or not (see Table ).Therefore, the satisfaction ranges from -(maximum dissatisfaction when the organisation does not implement any HRM practices which are important to each employee) to (maximum satisfaction when the organisation implements all HRM practices which are important to each employee).When both HRM practices are important to an employee but only one is implemented by the organisation, the relative importance of the HRM practices for the employee (s 1 versus s 2 ) determines the satisfaction value.This value is when both HRM practices have the same importance for an employee.It is positive (i.e., the employee is satisfied but not completely) when the HRM practice which has been implemented is more important than the other one.It is negative (i.e., the employee is dissatisfied but not completely) when the HRM practice which has not been implemented is more important than the other one.Extending to more HRM practices is elementary. .
The importance of the HRP − j practice for the employee (s j ) depends on employee gender (male/female), family status (family/no family) and age group (less than years/ -years/ -years/more than years)(see Table ).It changes over the employee's life depending on the employee's personal characteristics.Empirical studies on exploring the influence of gender (Gri eth et al. ) found that women give preference to intrinsic values (for example, challenging job, development opportunities, and autonomy).In contrast, men significantly prefer extrinsic values (high earnings and job security).In order to this model be applied to a real organisation, a survey of employees can be conducted to assess how well possible HRM practices would fit with their career goals and personal values.The better they fit, the higher the likelihood of an employee feeling professionally and personally tied to the organisation.

.
How much the employee is satisfied, not only depends on the relative importance of the HRM practices for the employee (s 1 and s 2 ) but also on the employees' sensitivity to the organisation implements HRM practices (b).The e ect of HRM practices implemented (or not) on job satisfaction is higher as the employees' sensitivity increases (see  .perf ormance = ef f ort + d s * satisf action, where d s is the influence of satisfaction on job performance.It ranges from to .When d s = , the job performance is only influenced by e ort.The influence of satisfaction on job performance is higher as d s increases.
In order to simplify our model, we consider that this influence is the same for all employees.It is fixed at the beginning and does not change over the simulation.
. Changes in the job performance are observed by using a graded chromatic scale, which ranges from blue (low performance) to red (high performance).

Employee's decision .
We consider that the employees leave the organisation when they have the intention to leave.

Simulations and Results
. We consider an organisation within a formal organisational structure whose lowest level is populated by employees who are distributed into functional areas (bellowing to departments).The time of simulation ranges from to years.We have evaluated the following four scenarios.For each scenario, we have conducted some experiments on merging of di erent types of e ort behaviour (null e ort/shrinking/ comparator/replicator/averager), organisational formal structure (functional/balanced matrix/strong matrix) and team leadership (ego-centered/democratic). repetitions per combination of parameters are considered to do the statistical analysis.
Baseline scenario: The influence of formal organisational structures .
. Our aim is to calibrate the e ect of the organisational structure on the emergence of organisational cultures (i.e., an equilibrium where all agents exert the same e ort) by comparing our results with Dal Forno and Merlone ( )'s results where formal organisational structures were not created.The main result achieved is that matrix structures (balanced or strong) with a democratic team leadership can lead to the emergence of highperformance organisational cultures in homogeneous averager populations. .
The evolution of the average e ort in a run (for di erent types of e ort behaviours within a functional formal organisational structure) are reported in Figure .We observed that, in the last period, the average e ort converges to in homogeneous shrinking populations, whilst it converges to values lower than in homogeneous comparator populations, fluctuates around in homogeneous replicator populations and converges to in homogeneous averager populations.
. In order to evaluate if there is convergence to the same e ort intensity in several repetitions within di erent formal organisational structures, the mean (and standard deviation) of the average e ort achieved in the last period of repetitions are reported in Table .We have obtained that:  • For homogeneous shrinking populations, there is always convergence to an organisational culture where all the employees exert a null e ort.Our results reproduce Dal Forno and Merlone ( )'s results independently on the formal organisational structure.It can be observed in Table that the average e ort converges to and the standard deviation is very low.This is visible in Figure as an only blue-bar with frequency .
• For homogeneous replicator populations, there is no convergence to an organisational culture.Our results reproduce Dal Forno and Merlone ( )'s results independently on the formal organisational structure.High standard deviations can be observed in Table for all formal organisational structures.We have obtained a Pearson coe icient of variation which ranges from , (within a Balance-Egocentered structure) to , (with a Strong-Democratic organisational structure).This is reflected in Figure by multicolour bars equally distributed.We have obtained the same conclusion by doing , and repetitions.
• For both homogeneous comparator and homogeneous averager populations, there is not a clear convergence to an overall organisational culture.We have obtained a Pearson coe icient of variation which ranges from , (for homogeneous comparator within a Strong-Democratic organisational structure) to , (for homogeneous averager populations with a Strong-Democratic organisational structure).We also observe green bars with frequency around (see Figure ).Therefore, there are reasons for a deeper analysis in order to evaluate either the emergence of subcultures associated to functional areas in one run or the emergence of cultures that, in di erent runs, converge to di erent e ort intensities or not convergence as it happens in Dal Forno and Merlone ( ).
Figure : Mean of e orts distribution within a functional formal organisational structure .
In order to evaluate the emergence of subcultures associated to functional areas, the mean in repetitions of the standard deviations of the e orts exerted in the last period by the employees in each area (sd-area) in homogeneous-averager and homogeneous-comparator populations are reported in Table and Table , respectively.In order to evaluate if these subcultures converge either to the same e ort intensity or not, the standard deviation of the areas' average e ort in the last period of repetitions are reported in Table and Table (last row).We have obtained that: • For homogeneous averager populations, subcultures associated to each area emerge in all formal organisation structures as the mean of the standard deviations of the e orts by area are very low.But note that, in strong matrix structures with either ego-centered or democratic leadership (SE and SD), the highest mean of the standard deviations is achieved by area-, which is the project-managers' area.Moreover, in democratic leadership in both matrix structures (BD and SD), subcultures converge to the same e ort intensity as standard deviations between the means of the average e ort by area is very low ( , and , , respectively).However, this e ort intensity is not the same in all repetitions, that is the reason why the standard deviation of the average e ort in repetitions reported in Table is high.We have also evaluated the influences of time size, obtaining that the e ort distribution moves to higher e ort intensities as the team size decreases (see Figure ).
• For homogeneous comparator populations, high standard deviation values can be observed in Table in all formal organisational structures analysed.Therefore, there is no emergence of subcultures neither convergence to the same e ort in a run.
Our aim is to calibrate the e ect of high-e ort employees within formal organisational structures on the emergence of organisational cultures by comparing our results with Dal Forno and Merlone ( )'s results.The main result achieved is that, in heterogeneous populations, the presence of high-e ort at management levels can lead to the emergence of organisational cultures.

.
Although Dal Forno and Merlone ( ) found di iculties to observe the emergence of organisational cultures in heterogeneous populations, they obtained that the average e ort which results from introducing high-e ort employees in homogeneous populations boosts the average e ort.Moreover, its value depends on the proportion of high-e ort employees introduced, except on null-e ort populations where their influence is rather predictable (a linear combination of null-e ort and high-e ort, weighted by their relative proportions).Adapting Dal Forno and Merlone ( )'s results to the range to (see Figure -le ), the average e ort is , in , %null-, %high populations, around in , %shrinking-, %high populations, and in both , %averager-, % high populations and , %comparator-, %high populations.
. Nevertheless, we have observed emergence of high-performance cultures when, in some heterogeneous populations within a formal organisational structure, high-e ort individual work at supervisor-levels (i.e., areamanagers and department-supervisors).The percentage of high-e ort employees needed for the emergence of organisational cultures drops to the percentage of supervisors ( % in our case: employees at low-level and at management-levels) (see Figure -right).) .
Our aim is to analyse both the entry of new workers and the departure of low-e ort employees.The main result achieved is that involuntary turnover a ects the emergence of organisational cultures negatively.
. The departure of low-employees increases as the firing-level increases.Therefore, involuntary turnover increases (except in homogeneous shrinking populations where the involuntary turnover rate is always near %).The departure of low-e ort employees and the entry of new workers lead to opposing e ects on the emergence of organisational cultures.
. On the one hand, dismissing low-performance employees leads to high average e ort organisational cultures.
In Figure the evolution of the average e ort in a run for di erent types of e ort behaviours (within a functional structure) are reported.In homogeneous shrinking populations, the average e ort in the last period ranges from to , in homogeneous comparator populations from to , and in homogeneous averager populations from to .We observe that the average e ort is higher as the firing-level increases, but when the firing-level is too high the opposite e ect appears.The mean (and standard deviation) of the average e ort achieved in the last period of repetitions within two extreme types of formal organisational structures (functional and Strong-Democratic) are reported in .
On the other hand, the entry of new employees whose e orts are not aligned with their co-workers' e ort, both reduces the emergence of organisational cultures causing variations in the convergence patterns to an e ort intensity.We observe in Figure these variations either between periods (in textitshrinking and textitreplicator populations) or for high firing-levels (in comparator and averager populations).The weight of these variations depends on both the type of e ort behaviour and the firing-level.In Table , both the mean of standard deviation of e orts by each area and the standard deviation of the areas' average e ort are reported for homogeneous averager populations within strong-democratic organisational structure for di erent firing-levels ( , and ).We observe that they increase as firing-level increases (by comparing the results obtained for di erent firing-levels (Table ) with the Baseline scenario results (Table 's last row)).We had the temptation of removing these variations between periods by selecting only workers whose e orts were higher than the firing-level.We discarded it because, however, real organisations select their employees by estimating their ability, but not their e ort.
phenomena of these variables relationships.The model we propose incorporates these variables that are relevant in empirical studies.They were modeled at individual level according to empirical studies (Gri eth et al. ; Judge et al. ; Christen et al. ).Emergent phenomena from the interactions of individuals were observed at organisational level.

. E ort.
The model we propose includes some bounded-rational behaviours considered by Dal Forno and Merlone ( ), which were extrapolated from human-subject experiments: high-e ort, low-e ort, shrinking, averager, replicator, and comparator.These behaviours can be considered more or less realistic according to real life, but the idea behind is: "the e ort exerted is contagious".Although this intuitive idea could be familiar to practitioners, to date there has been little information available about another person's e ects on performance depending on what that other person is doing (Desender et al. ( ) addressed this issue).
. E ort-organisational structure-organisational culture.We have obtained that both some e ort behaviours exerted by the employees and some formal organisational structures can favour the emergence of organisational cultures: homogeneous shrinking populations, homogeneous averager populations, or even homogeneous replicator populations within an appropriate formal organisational structure, but never homogeneous comparator populations.
. In homogeneous shrinking populations, low-performance organisational cultures emerge.However, organisations are interested in high-performance organisational cultures.Some formal organisational configuration could favour the emergence of high-performance organisational cultures.In particular, we have obtained that high-performance organisational cultures and sub-cultures emerge in homogeneous averager populations within matrix structures (balanced or strong) with a democratic team leadership.The emergence of highperformance organisational cultures is also favoured when high-e ort employees at management levels (i.e., functional-areas and department-supervisors) interacts with any of types of employees' e ort behaviours (shrinking, averager, replicator, and even comparator). .

Schein (
) advanced the idea that organisations can have subcultures linked to a di erent management team, either co-existing or conflicting.In our paper, we have achieved that subcultures are associated to some types of e ort behaviours.Co-existing subcultures are associated to both functional areas and some types of e ort behaviours.Nevertheless, conflicting subcultures are associated to both organisational levels and conflicting e ort behaviours.
. Performance-turnover-organisational culture.Turnover (voluntary or involuntary) a ects the emergence of organisational cultures negatively.Therefore, in order to promote high-performance organisational cultures, the main challenge is to retain high-e ort managers.This challenge can be faced by adapting the set of HRM practices to those valued by these high-employees, moreover taking into account the costs associated to turnover.Boushey and Glynn ( ), by reviewing case studies taken from the most-relevant research papers on the costs of employee turnover, demonstrated that it costs businesses about one-fi h of a worker's salary to replace that worker.They found that average costs to replace an employee are % of annual salary for high-turnover low-paying jobs (earning under $ , a year), % of annual salary for mid-range positions (earning $ , to $ , a year) and up to % of annual salary for highly educated executive positions.Although, in order to make a decision, this economic valuation is crucial, it has been omitted in our model.
. Understanding both the degree to which high-performance HRM practices a ect organisational performance and the conditions that moderate this relationship, helps researchers to build contingencies in theory and aids practitioners to justify investments in HRM practices (Combs et al. ).

Summary and Further Research
. This article attempts to show an alternative approach to understand the link between Human Resource Management theory and theory on performance.The HURLAB-is a simulation platform which provides a simple way to study the complex interactions between di erent types of agents interacting within di erent types of formal organisational configurations in organisations with di erent set of HRM practices.
. We have obtained that: (i) Some possible combinations of both employees-e ort behaviours and formal organisational structures can favour the emergence of organisational cultures more than others; (ii) The interaction between employees within matrix structures (balanced or strong) with a democratic team leadership favour the emergence of organisational cultures; (iii) High-e ort managers are relevant for the emergence of highperformance organisational cultures; (iv) Turnover (voluntary or involuntary) a ects the emergence of organisational culture negatively.We conclude that the main challenge is "to retain high e ort managers by adapting the set of HRM practices to them". .
The most obvious extension is the validation of our model.Windrum et al. ( ) reviewed three of the most influential approaches to empirical validation that have been developed in the agent-based literature.The history-friendly approach constrains parameters, interactions, and decision rules in the model in line with the specific, empirically-observable history of a particular industry.The other two approaches employ empirical evidence to identify sub-regions in the potential parameter space and, within these sub-regions, the model is expected to replicate some relevant statistical regularities or stylised facts.

.
Nevertheless, none of these approaches can be used to validate our agent-based model because it has been already designed employing empirical evidence.Therefore, we cannot use the same empirical evidence to validate it.The only way to validate our model is to apply it to real life, in a particular organisation.However, in order to do it, our model has to be extended including the following issues: costs associated to turnover (Boushey et al. ), national cultural di erences (Hofstede et al. ), and a more complex e ort-satisfactionperformance relationship (Judge et al. ).All these works are in process.
. Finally, although we are aware that our model could be rather simple by comparing with real life, we believe that our simulation platform could be of interest to researchers (inspiring empirical studies) and practitioners (allowing to test Human Resource Management decisions both without consequences and before being implemented in a real organisation by characterizing the employees in a virtual environment).

Figure
Figure : Agent-based model proposed based on Guest (

Figure :
Figure : Formal organisational structures (functional, balanced matrix, and strong matrix) and team leadership (egocentered and democratic) Figure ).However, in certain vocational professions where b could even be equal to , although Table : Relevance of HR Practices (F:family; NF:no-F NF F NF F NF F NF F NF F NF F NF F HRP-HRP-HRM practices could be welcome and employees could have a sorted list in mind, they are not the reason of the satisfaction feeling..Changes in satisfaction are observed by using a graded chromatic scale, which ranges from black (low satisfaction) to red (high satisfaction).

Figure :
Figure : Satisfaction when only one HRM practice is implemented

Figure :
Figure : Evolution of the average e ort in a run in homogeneous populations within a functional structure

Figure :
Figure : Team size influence on e orts distribution in homogenous averager populations within strongdemocratic organisational structure

Figure :
Figure : Evolution of the average e ort in a run in homogeneous populations within functional structure and di erent firing-levels

Table :
Variables of the organisation-agent

Table :
Variables of an employee-agent

Table :
Mean and standard deviation of average e ort in homogenous populations

Table : E
orts by areas in homogenous averager populations

Table : E
ort by areas in homogenous comparator populations Table in order to compare a firing-level = with the baseline scenario's results (see Table , first and last columns).

Table :
Mean and standard deviation of average e ort in homogenous populations with firing-level=