A selection procedure is dealing with choosing the best suitable individuals (from a sample of applicants) to the available job. In the selection process it is recommended to begin with what employers are looking for. According to Arnold (2005) this task requires several stages and the first is a job analysis, which is the base of an effective selection process because it helps to choose the right selection method. The second part in the selection procedure is to attract applicants and to choose the appropriate selection method/s. It is also essential to understand the requirements for the method's effectiveness (reliability and validity of predictor test) and the fairness towards applicants (Cook, 2004). The next stage in the selection process is the actual use of that chosen method (or various methods) followed by the selection of a successful candidate. The remaining step is the monitoring of the job performance to provide validity data (Arnold, 2005). This essay is concerned in describing the main aspects of the job analysis and several job analysis techniques will be shortly examined. The reliability, validity and some selection methods that are of interest particularly in UK and USA will be briefly acknowledged. Current selection practice mostly used in USA were according to Rynes, Orlitsky & Bretz (1997) reference checks, then structured interviews, followed by drug tests and unstructured interviews. Moreover, Rynes et al. (1997) found out that graduates were likely to be assessed during a trial work period or by educational achievements. In UK Hodgkinson, Daley & Payne (1995) have discovered that the 'classic trio' is still the most used but is also joined by other methods. Bartram, Lindley, Marshall and Foster (1995) obtained information that small employers carry out interviews where they try to assess the honesty and interests in job of applicants. Many organizations try to select the most suitable individuals by 'classic trio' made up by application forms, interview and letter of reference, but there is the suggestion that references and interviews are inaccurate selection methods (Cook, 2004). Although there are various methods for selection of applicants, the selection methods such as 'classic trio', psychometric testing, work sample tests and assessment centres will be introduced and the fairness towards applicants will be mentioned. These stages mentioned above, especially job analysis, choosing the selection methods and understand the criteria of good selection method, should help when trying to advise a selection committee/the personnel department how to design the selection procedure for new staff.
The selection process starts with a task analysis that defines a job description and a person specification. They represent the picture of the offered position, which can be used for attracting applicants as well, and they also help to prevent from applying of some unsuitable applicants (Cook, 2004). There is a distinction between a task-oriented analysis procedure (concentrating on the work itself) and a person-oriented job analysis procedure that is focused on behavioural and psychological requirements of the job (Sandberg, 2000). Two main problems were found by Cook (2004) in the job description and person specification. Firstly, there were usually stated all of the duties irrespective that their importance or difficulty was not mentioned and secondly, there was a wide use of a 'managementspeak' that does not really explain what to do. Cook (2004, p. 21) noted that "many UK person specifications waste time stating that the applicant must be keen, well motivated and energetic"; this seems to be unnecessary because no employer wants lazy or unmotivated applicants. Job descriptions in USA mostly focus on knowledge, skills and aptitudes. Such competencies are more likely connected to professionals and managers, while competences in UK can be qualified by for example the National Vocational Qualification (NVQ) (Arnold, 2005).
In the job analysis it is necessary to collect information at the first place and there are various forms how to collect them. Cook (2004) pointed out some of the methods: video recording, records of for example accidents or sales, observation, structured and open-ended questionnaires filled in by workers and supervisors, interviews of individuals or group or diaries which capture day to day activities at person's job. Participation is believed by some psychologists to be the best way how to understand the nature of that particular job (Arnold, 2005). Secondly, researchers need to make a sense of what they have collected and this can be done subjectively (based on researchers' impressions), rationally (provided by committee and through consultation in order to be understood and accepted by the organisation) or statistically (Cook, 2004). The statistical analysis has two types: factor (this analysis groups the tasks and identifies the job) and cluster analysis (groups people according to characteristics) (Krzystofiak, Newman and Anderson (1979). Similarities between jobs, provided by such statistical analysis, help in planning training, sickness cover or staff succession (Arnold, 2005).
One of the oldest types of job analysis techniques was produced by Flaganan (1954); it is known as the Critical Incident Technique (CIT) and was used in Second World War to find out failures during military pilot trainings. Flanagan stated that CIT identifies the incidents of job behaviour that results in effective and ineffective performance (good and bad practices). Repertory grid technique (RGT) is popular in UK and is based on worker-oriented information (Kelly, 1955). Kelly (1955) explained the principle of RGT in which workers are asked to imagine a good, an average and a poor worker and then they are invited to explain which two differ from the third and how. Job-oriented Functional Job Analysis (FJA), introduced by Fine and Wiley (1974), describes the nature of various jobs as well as their complexity. Job components inventory was developed by Banks, Jackson, Stafford and Warr (1983) for jobs that require limited skills. McCormic, Jeanneret and Mecham (1972) were interested in the worker-oriented Position Analysis Questionnaire (PAQ) that seems to be the most widely used structured interview with workers and supervisors. Here, the focus concentrates on aspects of worker behaviour and then trained job analysts create their own view about the job. The information collected in PAQ consists of 200 elements divided into six main areas (information input, mental processes, and work output, relationship with others, job context and other aspects). Information about importance to the job, amount of training required, etc. are given there as well (McCormic et al., 1972). McCormic et al. (1972) found an acceptable level of reliability for the PAQ. Yet, there are some doubts about validity. A helpful database is the Dictionary of Occupational Titles (DOT)/O*NET. They present details of thousands of jobs and the ratings of their complexity (DOT) (Cook, 2004). Peterson et al. (2001) examined database O*NET that recently replaced DOT and they mentioned that O*NET includes details such as worker requirements and characteristics, experience requirements, occupational requirements and characteristics and occupation-specific requirements. In general, these techniques help researchers to differentiate between what aspect should be selected from applicant and what aspects can be trained (Cook, 2004).
Recruitment sources play important part in the second stage of personnel selection. Robertson and Smith (2001) have indicated that attracting applicants for the offered position can be in a form of advertisement, job fair or job agencies. They mentioned that applicants can also hear (usually from existing employees) about the offered position; however, this kind of informal recruitment is disliked by some fair employment agencies. Another way called 'walk-ins' is that the applicants can come to ask for available vacancies by themselves. Bartram (2000) concluded that the internet can be used for advertising (job positions but also details of contestants), making applications and then their sifting and even the assessment itself can be performed electronically and much quicker. Bartram (2000) then made a notion that not everyone has an access to the internet that can leads to excluding of some people because of their gender, age or ethnicity differences.
In the second stage of the selection process is also, according to Robertson and Smith (2001), the identification of the selection methods in order to assess whether applicants demonstrate the essential characteristics or not and psychological researchers suggest that the important point is that the good selection method should be reliable and valid. Reliability is an ability of the method to give consistent results over time (Schmidt & Hunter, 1998). Reliability is expressed as a reliability coefficient that evaluates how strongly two measures of data go together, with its perfect correlation value of 1 and no correlation would give the zero value of coefficient (Arnold, 2005). Arnold (2005) pointed out that personnel criteria, such as advancement, promotion, length of service, turnover, punctuality, absence, disciplinary action, accidents and sickness, are easy to collect. However, they may be very unreliable. Validity means the degree to which a method measures what was intended to be measured (Schmidt & Hunter, 1998). The perfect correlation would have the value of one and no correlation would have zero as it was with reliability. A valid selection method should accept good applicants and reject poor ones (Cook, 2004). According to Arnold (2005), an empirical validation studies should be carried out to examine the quality of selection procedure. Unfortunately, as Cook (2004) mentioned, these validation studies are quite time consuming and expensive, which is the reason of many organisations not to conduct them. Validation study in selection is concerned in collecting two sets of data: the predictor (e.g. ratings and scores from interview or psychological tests) and the criterion (work behaviour and performance). Arnold (2005) also stated that the strength of the relationship between predictor and criterion refers to criterion-related validity that predicts productivity. A validity coefficient is the computed correlation between predictor and criterion (Cook, 2004). The reliability and validity of specific selection methods will be mentioned later.
In addition to reliability and validity, it is important to bear in mind the fair employment and equal opportunities. Any discrimination in employment on grounds of race, colour, religion, national origin or gender and disability (in USA also age) is prohibited (Cook, 2004). Candidates' views and reactions on selection process are also essential because applicants with negative view will be likely to refuse the job offer, or refuse organisation's products (Robertson & Smith, 2001). Ryan and Ployhart (2000) stated that candidates prefer more interviews, work samples and assessment centres, but they do not like for example personality tests. In general, they favour more job-related approaches that are found to be more fair as well (Cook, 2004). Unfair selection procedures can be due to selection errors that can arise from e.g. sampling error (variation in results arising from small sample around 50-150 people), poor measurement of both predictor or/and criterion or range restriction of scores (Arnold, 2005). How this unfairness can be reduced will be stated below with selection methods.
The classic trio is made up of the application form, interview and references. The application form usually serves as a first filter for choosing applicants in the selection process. There is a suggestion that this process (called also sifting) is not always done effectively (Robertson & Smith, 2001). Machwirth, Schuler and Moser (1996) found out that some managers, who are responsible for choosing applicants, differ in what they say and what they do. To better specify, managers claim that they choose applicants based on their previous positions and abilities. In fact, when applications appear badly written or untidy, managers tend to reject them without further consideration. Davison and Burke (2000) have noticed that the less job information was given in simulated personnel selection, the greater was the discrimination against females. To prevent gender bias, the Commission for Racial Equality in UK suggested that an employment decisions such as application sifting should be done by at least two people (Cook 2004). Another improvement of sifting is training and experience ratings used in USA in order to assess applicants' experiences and trainings rather than let the sifters decide themselves (McDaniel, Schmidt & Hunter, 1988).
Dany and Torchy (1995) have found in the Price-Waterhouse-Cranfiel survey that 80-100% of European employers use method of interview. Arnold (2005) mentioned that panel interviews (consist of several interviewers) were favourite in UK. Conway, Jako and Goodman (1995) also stated that panel interviews are more reliable. Conway et al. also have found the unstructured interview very unreliable and subjective. They also suggested that the limitation of interviews seems to be because of the inconsistency of the applicant's behaviour and the reliability can be greater when the previous job analysis is made and when interviewers are trained. Huffcutt, Conway, Roth and Stone (2001) mentioned the odd use of the interview for the accessing of personality and mental ability, when in fact; tests of both of them are also available and probably more accurate. Nevertheless, the interview is a good method for assessing social skills (Cook, 2004). Cook (2004) pointed out that interviews have a feature of versatility because they are used to assess variety of abilities, skills, attitudes, personality, etc. and this versatility can be one of the reasons for poor validity. Some other reasons for poor validity, as Arnold (2005) mentioned, can be due to difference between characteristics of interviewers and their motivation. Arnold also noted that factors influencing interview can be also the interviewee's impression management and various biases such as stereotyping (the idea of an ideal worker) or implicit personality theories (tendency to believe when somebody is friendly because he or she is smiling for example). Wiesner and Cronshaw (1988) argued that the validity can be improved by the previous job analysis that is found to be necessary in structured interviews. Structured interviews include structured questions and structured judgements of interviewer by rating scales etc. The mean validity was found twice as high (0.62) than unstructured interview (0.31) and according to Conway et al. (1995) structure interview has 0.53 value of reliability. In addition, structured interviews are perceived more fair (Cook, 2004). The assistance of two or more interviewers also improved interviews (Wiesner & Cronshaw, 1988). Huffcutt and Woehr (1999) have discovered that trainings of the interviewers and taking notes had a significant improvement of validity. The next way how to find out some information about applicants is to ask, for example, previous employers or teachers in a form of 'references'. References are usually the final part of the selection process and do not seem to be perfect in finding effective employees (Arnold, 2005).
Another widely used selection method is, for instance, psychometric testing such as cognitive ability tests. Cognitive ability tests are perhaps the best predictors of job performance and they assess e.g. intelligence (intelligence test used also in US army in the First World War in 1917), verbal skills, numerical ability, spatial perception, etc., but these tests do not appear to be really popular by the public (Arnold, 2005). Hunter and Hunter (1984) pointed out that the validity of cognitive ability test is around 0.5 for most types of work. Schmidt and Hunter (1998) mentioned that the most broad but also controversial test seems to be the General Mental Ability (GMA) test that assesses how individuals are good at understanding and using information. The validity of GMA confirmed Bertua, Anderson and Salgado (2005) in their meta-analysis (statistical procedure) of validity coefficient around 0.5-0.6 and stated cognitive ability tests as predictors of job performance as well. However, these cognitive ability tests are expensive and there is some debate about the biases and adverse impact (when the method results in fewer women or ethnic minorities, but not as a result of discrimination) of these cognitive ability tests, found especially in some ethnic minorities in USA (Roth, Bevier, Bobko, Switzer & Tyler, 2001). Second psychometric test is personality test; however it does have just moderate use and low criterion-related validity of 0.23 (Salgado, Anderson, Moscoso, Bertua & Fruyt, 2003). There can be found an improvement in validity, when personality tests are based on job analysis (Tett, Jackson and Rothstein, 1991).
Both Work sample tests and Assessment Centres are seen as relatively valid selection methods. Work sample tests believe in the behavioural consistency, which basically means that 'past behaviour is the best predictor of future behaviour' (Cook, 2004, p. 192). Hunter and Hunter (1984) determined 0.54 criterion-related validity of work sample test. They stated that the method is based on the instructions given to candidates and then candidates have some specific time to make that particular task. Nevertheless, the work sample tests should assess only individuals with previous job skills; they are job specific and also quite expensive (Robertson & Smith, 2001). Regarding candidates, work-sample tests seem to be quite accepted by contestants. The principle of the assessment centres (AC) is a multi-method assessment of individual or group and ACs have become quite popular (Robertson & Smith, 2001). Cook (2004) discussed that the important aspects are the dimension of performance and the assessment method matrix. He said that when dimensions of work performance are stated, at least two different assessment methods are used (exercises such as business simulations, role play, presentations, interviews, psychological tests and many others are used to test individuals). The reliability of AC was argued to be a complex issue but the validity was found to be around 0.4. Furthermore, ACs were regarded as a fair method without creating adverse impact (Robertson & Smith, 2001). For the improvement of validity it helps when assessors are psychologist rather than managers and when more assessments are used (Gaugler, Rosenthal, Thornton & Benson, 1987).
The main assignment of the selection procedure is to choose the best suitable applicants and reject the bad ones. When advising a selection committee or the personnel department do design a selection procedure, it is essential to state the stages of selection process and then to choose the most appropriate. The task analysis is the first stage that should represent the job description and a person specification. The job analysis improves the quality of selection because it gives interviewers information about what the job involves. Due to this knowledge it is possible to focus on applicants' skills, abilities and knowledge in a fairer selection system. Moreover, since researchers know what they are looking for, it will be easier to justify the choice of selection method in case of any legal dispute. Job analysis methods include for example Critical Incident Technique, Repertory Grid Technique, Functional Job Analysis, Job Components Inventory or structured and perhaps the most used Position Analysis Questionnaire. The helpful O*NET database included various details about thousands of jobs. The recruitment sources were found to be important in the attracting applicants as well as the role of the internet in recent years. It is important do bear in mind that despite of wide range of selection techniques, not all of them are equally appropriate in some selection procedures and in the most of them, there is no guarantee that candidates will be honest and truthful. In that case the reliability and validity of selection methods and their fairness should be considered. Nevertheless, even if the validation studies are recommended, due to time consuming and their cost are rarely conducted by employers. Various meta-analytic studies have examined the results from many separate studies and allowed us to get clearer picture about validity of selection methods. The 'classic trio' was found quite inaccurate because of their often unfair sifting; however, structured interviews achieved relatively high validity and use. The problem with fair employment was discovered mainly in psychometric testing; even when they have quite high criterion-related validity, the adverse impact (not deliberate discrimination) often occurs. Work sample tests and assessment centres with structured interviews were more preferable and viewed as fair by contestants. In general, the improvement of selection methods can be obtained by the use of more trained interviewers and previous job analysis and more validation studies. The adverse impact and various selection errors should be eliminated as much as possible. Yet, the question of cost is also important.