Forced distribution system

Though many researchers have pointed out several drawbacks in FADS, due to the absence of any suitable alternative, it has been (and continues to be) adopted by many industries over a long period of time. The purpose of this paper Is to point out some serious limitations of this system and propose a simple modification to overcome these limitations. Design/ methodology/approach – FADS determines the relative positions of the employees involved in similar work by comparing them against one another, and based on their performance, the employees receive different grades.Here the authors use the Likelier scaling method to convert these grades into numerical scores, then these scores are used to estimate the average performance of each group of employees, which Is referred to as the group Index. Taking these group Indices Into consideration, the authors propose a modeled performance score of each employee for their final evaluation. Efficiencies of the existing FADS and the propose d modified version are compared using a simple measure of rank correlation known as the Sandal's taut-statistic.Findings – Extensive simulation studies show that the modified algorithm is uniformly better than the existing one over different schemes for allocations of employees to deferent projects, and depending on the allocation scheme, It can lead to substantial Improvement. Relationally/value – This paper can be appraisal system based on a forced distribution and the first that provides a simple but effective solution which can be adopted by the organizations using FADS for performance appraisal.Keywords Performance appraisal, Statistical methods, Human resource management Paper type Research paper 1 Introduction Performance evaluation is regarded as one of the most powerful human resource raciest (Judges and Ferris, 1993; Murphy and Cleveland, 1995, p. 4). It provides a justification for human resource decision such as rewards, career planning, transfers, training , counseling, mentoring, termination, etc.Performance appraisal provides the employer an opportunity to communicate with the employees about the mission, strategy, vision, values and objectives of the organization, and it personalizes organizational strategy into individual performance criteria. It has been observed that employee motivations to perform, to develop capabilities and to improve future reference are influenced by the performance appraisal system (Land et al. , 1978; The authors thank the anonymous reviewer for providing several helpful comments and suggestions that led to substantial improvement of the article.Despite the importance of a performance evaluation system, extensive studies in this field have identified significant shortcomings in its applications that include different types of biases stemming from rating errors, sources of performance information and individual differences (Arrive and Murphy, 1998). Among these various shortcomings, rating bias is the most predominant area of research, which indicates the tendency of the raters to provide lenient or stringent rating (Beret et al. 992; Runes et al. , 2002). This systematic bias leads to lack of discrimination between high and low performance and automatically disrupts the whole essence of performance decades, several researchers have explored different methods to overcome the rating bias and to improve the accuracy of performance evaluation system (Goff et al. , 1996). Studies on performance evaluation are mainly focused on two types of appraisal systems – absolute and relative.In an absolute rating system, individual performance is evaluated against a pre-determined standard, whereas a relative evaluation yester determines the relative positions of different employees by comparing them against one another (Duffy and Webber, 1974). Though there are advantages and disadvantages in both of these systems, some studies have pointed out the superiority of the relative grading system over the absolute one (He-man, 1986; Nathan and Alexander, 1988; Wander and Goofing, 1997).Many renowned organizations including General Electric (GE), Hen, Microsoft, American Express and Goldman Cash have used and some still use relative grading system for performance evaluation in the form of a forced distribution system (FADS) (Grotto, 2005). FADS was developed in an attempt to directly deal with the problems of rater leniency and lack of discrimination while measuring an individual's performance (McCarty, 1988).This system forces the managers to discriminate between high and low performers either by sorting the employees into some pre-determined performance categories based on a pre-defined distribution or by ranking them on the basis of their relative performance (Gurgling et al. , 2004). The first process is also known as the criterion-reference rating, while the second one is known as the norm- reference rating (Visionary and Ones, 2000; Visionary, 2001).The wide use of FAD S as an objective measure of employee performance was mostly popularized by Jack Welch at the beginning of his tenure at GE under the name of â€Å"vitality curve† (Bossily and Charka, 2002; Itchy and Sherman, 2001). Welch introduced this system to develop an objective measure to discriminate between high and low performer so that the culture of â€Å"rewarding doers† can be established, which in turn can be helpful for â€Å"building muscle† of the organization. In GE and many other organizations, FADS is considered as a developmental instrument for achieving a performance-oriented culture.Though FADS has several advantages, many organizations have been observed to perceive this system negatively (Rock et al. , 2007). Many researchers and practitioners have also pointed out that a forced distribution in performance evaluation leads to extreme level of Job dissatisfaction among the employees with high potential to perform (Gray, 2002; Madman, 2006; Prefer and Su tton, 2006). In practice, a relatively low-performing member in a high- performing team can often be better than the best performer in an average performing team.The FADS is used to evaluate the members working in different roofs or teams separately, and the rigidity of this system forces the companies to reprimand all low performers of each and every group. As a result, some high-potential performers may be asked to leave the Job or they may leave the organization voluntarily due to dissatisfaction. The negative consequences of this performance evaluation system have been observed in many organizations. For instance, Ford had a well-publicized unsuccessful experience with a forced ranking support this ranking system at all.Many employees, who had received positive feedback for years, were suddenly categorized as under performers. As a result, dozens of Ford employees and ex-employees sued the company because of this system of evaluation. Vishnu et al. (2006) studied the long-term e ffect of the â€Å"bell curve†, which is a form of forced distribution, on organizational dynamics. According to their views, pressure of the bell curve can facilitate the performance to a certain level, but constant pressure demoralizes the employees. As the company shrinks, the rigid distribution of the bell curve forces the manager to categorize a high performer as a mediocre one.Also, it is often assumed that the employees identified as low reformers, on account of their unsatisfactory performance, are usually replaced every year by fresh talent, who can add up to the output of the organization. Here we may argue that the likelihood of the presence of poor performers amongst these new additions is another possibility, which can adversely affect the system. Moreover, from the financial point of view, replacement of the employees with the fresh talent is also a costly affair. Blame et al. 2009) conducted a study on a student population to understand their reactions towards different types of FADS, and pointed out that â€Å"less trending consequences for low performer† is the most powerful variable in determining the attractions towards different types of FADS. On the basis of this study, we may assume that the organization, where stringency is higher for the low performers, has less chance to get higher number of Job applicants. In order to overcome these limitations of the forced ranking appraisal system and to protect the employees with high potential, Vishnu et al. 2006) proposed to use an evaluation system based on the semi-bell curve, where instead of putting fixed proportions of employees into different groups of performance levels, an organization adjusts these proportions depending on the set of employees it is dealing with. However, the adjustment scheme they proposed was subjective and somewhat ad hoc. Instead of using the semi-bell curve, an organization can use any other curves as well, or it can even use different curves for diffe rent set of workers involved in different projects. It depends entirely on the company policy, and we have no prerogative to decide that.So, in this paper, we do not recommend the use of any particular curve. Here, we repose a modified algorithm for forced ranking performance appraisal that can be used irrespective of the nature of the curve(s) used by the companies for the evaluation of their employees. In the following sections, we assume the bell curve appraisal system for the demonstration of our method. However, the description of our method will make it clear that our modification is not limited to the bell curve system, and it can even be used when a company adopts different types of curves or distributions for evaluation of its employees involved in different projects. . Objectives of the study The main objectives of our study are given below: evaluation of the present appraisal system under different schemes of allocation of workers and identifying the limitations of the fo rced ranking appraisal system; propose a new method for performance appraisal to overcome these limitations; and extensive comparison between the proposed and the existing method of performance appraisal to demonstrate the usefulness of the proposed method. The next three sections address each of these objectives in turn. . Evaluation of the forced ranking performance appraisal system First, we carry out emulation studies to evaluate the performance of the existing forced ranking appraisal system. Let us consider an organization with 3,000 employees. Suppose that there are 30 projects run by the organization, and 100 employees are involved in each of these projects. For the time being, we assume that the potentials of these employees are known and they perform according to their potentials.We generate 3,000 observations from a normal distribution with mean 50 and standard deviation 10 (so that almost observation lie in the (O, 100) range) and consider them as true potentials of diff erent employees. Here we consider two different allocation schemes for assigning the employees to different projects, and we will refer to them as random allocation and extreme allocation, respectively. In random allocation, the employees are randomly assigned to different projects. In extreme allocation, employees having similar potentials are assigned to the same project.For instance, the 100 employees having the highest potentials are assigned to one project, the next 100 employees to another project and so on. In each of these two cases, we evaluate the performance of the employees involved in each project, and the grades are located to them following the bell curve system. Assume that the organization uses six different Grades A (best)-F (worst) for its employees, and in the bell curve system, the proportions of employees to receive these grades are PA h if h 1?F(2), BP h PEP h F(2)?F(1) and PC h PDP h F(1)?O. , where ? ) is the cumulative distribution function of a standard no rmal variant ( Johnson et al. , 1995). Now, one may be curious to know how FADS perform in such situations. In order to investigate this, we compute the correlation coefficient between the potentials of the employees and the grades obtained by them and use it as a measure of efficiency of the appraisal system. Same grade, we use the Sandal's t-statistic (Kendall, 1938) as an appropriate measure of correlation.Note that unlike the product moment correlation coefficient, this statistic is invariant under any monotonically increasing transformation. So, instead of normal distribution, if we generate 3,000 observations from any other distribution, the efficiency measure based on the Sandal's t-statistic will remain the same. In the case of random allocation, the existing method performed quite well, and it led to a rank correlation of 0. 716. But in the case of extreme allocation, it had drastically poor performance.The rank correlation turned out to be 0. 024. So, essentially there was no correlation between the potentials of the employees and the grades obtained by them, and the forced ranking system was as bad as random grading. Note that random allocation of workers to different projects is presumed rare in practice. Usually the employees are assigned to different projects based on their expertise, and also depending on the difficulty level of the project.Therefore, it is not so rare that the employees having higher potentials are assigned to high-end projects, and hose having relatively lower potentials are assigned to low-end projects. In such cases, the existing performance appraisal system will perform poorly, and as a result some of the high (low) potential workers will get lower (higher) grades. Being unsatisfied with the grading, some of these high-potential workers may lose motivation to work hard and some of them may leave the organization for a new Job.This is quite harmful for organizational functioning. In practice, in many cases, the allocation of the workers in an organization is somewhere between the totally random allocation and the extreme allocation. From the above discussion, it is quite transparent that in such cases, the existing appraisal system may only have a moderate performance depending on the extremity of the allocation scheme. This clearly shows the necessity to develop a new method for performance appraisal, which can have satisfactory performance even in the case of extreme allocation.We develop one such method in the next section. 4. Modification to the forced ranking performance appraisal system: a new method for performance appraisal The existing method of performance appraisal simply considers the present year's elating performance of the employees involved in the same project, and the grades are allocated to them only based on their present performance. Because of this relative grading, an employee in a high-performing team can get poor grades in spite of performing better than all other members in an average performing team.In order to overcome this limitation, here we take the previous year's grades of the employees into consideration. Suppose that there are n employees working in a particular project, and Gig is the grade obtained by the tit employee (I h 1, 2, y, n) in the previous year's appraisal. In order to develop a modified appraisal system, we assume that the individual performance of the employees may vary from their previous year's performance, but the average performance of these n employees remains almost the same. For computing this average performance, we follow the Likelier, (1932) scaling method.It is a statistical tool that converts the Grades Gig into numerical scores s(Gig) and makes it possible to calculate the mean score I h for this group of employees. Though here we assume that the performance evaluation is done once in a year, this method can be used when the evaluation cycle has shorter or larger eroticism or even when it is periodic. Let us assume tha t a company uses a total of K Grades AAA, AAA, y, AK in the appraisal system, where AAA and AK denote the highest and the lowest grades, respectively.Also assume that it puts the best performing Pl proportion of employees in AAA, the next up proportion in AAA and so on. Therefore, the worst performing PC proportion (here Pl up pep PC h 1) of employees receive the Grade AK. So, if we assume that the performance of the employees are normally distributed (this is the assumption the companies made when they adopt the bell curve system), the mean of a truncated tankard normal distribution with truncation at and above the (1 ?up)the quintile of the standard normal distribution can be used as the score function for AAA .