Selecting for fit: a direct test of Schneider’s selection proposition

This paper reports an empirical test of Schneider’s (1987) selection proposition that organizations select people who share the organization’s values. The values of 412 applicants to nine utility companies in the United Kingdom were captured and their fit to (1) the values of the organization as viewed by members of the Top Team (P–OV fit), (2) the values of the recruiting departments as viewed by employees within those departments (P– DV fit), and (3) the values of people working in those recruiting departments (P–P fit) was calculated. The results show that selection outcomes are positively associated with P–P fit, negatively associated with P–DV fit, and not associated at all with P–OV fit. Selection effects for fit were small and only present when there was some form of face-to-face contact between applicants and selectors. At the end of the paper, a conclusion discusses the implications of these findings for the fit literature in general and Schneider’s selection proposition in particular. Some managerial implications are also considered. Attraction–Selection–Attrition (ASA) theory (Schneider, 1983a, 1983b, 1985, 1987; Schneider, Goldstein & Smith, 1995; Schneider, Smith & Goldstein, 2000; Schneider, Smith, Taylor & Fleenor, 1998) is one of the main theoretic foundations of the person–organization fit literature. It posits that as organizations mature they are increasingly occupied by people who are similar to each other. This homogeneity happens as a result of three phases of the ASA cycle. Namely, organizations attract people to them who share their values. Organizations select those people who share their values. And finally, there is attrition from those people who find they do not share the organization’s values (i.e. they chose to leave). Schneider (1987) argues that the homogeneity that results from the ASA cycle is potentially very dangerous for organizations as they will become increasingly ingrown and resistant to change. A number of studies have tested the homogeneity hypothesis and produced evidence to demonstrate that the employees in organizations become similar over time (e.g. Denton, 1999; Jackson, Brett, Sessa, Cooper, Julin & Peyronnin, 1991; Jordan, Herriot & Chalmers, 1991; Schneider et al, 1998). There are a much larger number of studies that have explored the attrition phase of the cycle and demonstrated its effect in producing similarity between 1 All correspondence concerning this article should be addressed to Jon Billsberry, Centre for Human Resource and Change Management, Open University Business School, The Open University, Milton Keynes MK8 0DA, United Kingdom; +44 (0) 1908 655888; Fax +44 (0) 1908 655898; electronic mail may be sent to [email protected]. 2 The author would like to thank Kevin Daniels, Peter Herriot, Tim Clark, Jo Silvester, and Chris Coupland for their helpful comments. Billsberry Schneider’s Selection Proposition 2 people (e.g. Boxx, Odom & Dunn 1991; Bretz & Judge, 1994; Chatman, 1991; O’Reilly, Chatman & Caldwell, 1991; Ostroff & Rothausen, 1997; Posner, 1992; Posner, Kouzes & Schmidt, 1985; Van Vianen, 2000; Vancouver & Schmitt, 1991). Much less work has been carried out to test Schneider’s attraction and selection propositions, although there are, of course, many indirectly associated studies of each phase of the cycle which convey strong intuitive support for the framework. The present study makes a contribution to the literature by reporting a direct test of Schneider’s selection proposition that organizations select people who share their values. The ‘similar-to-me’ effect Schneider’s selection proposition is built on the premise that similarity leads to attraction. This attraction leads to decisions by organizational representatives to offer jobs to people who have similar values to those of the recruiting organizations. There are many research findings supporting the idea that people select people like themselves. An example of this research is the study conducted by Prewett-Livingston, Feild, Veres and Lewis (1996). These researchers looked at promotion interviews of police officers in an American metropolitan police department. The researchers were able to monitor the effects of race (Black and White) on ratings in situational interviews. They found that interviewees who were similar to interviewers with respect to their race received higher ratings. Both Black and White raters gave higher ratings to candidates of their own race. This study adds support to the ‘similar-tome’ effect (Rand & Wexley, 1975) which has been demonstrated in many similar studies (e.g. Krainger & Ford, 1985; Lin, Dobbins & Farh 1992; Peters & Terborg, 1975). The ‘similar-to-me’ effect is not just limited to race. Laboratory experiments have shown that it also applies to gender (e.g. Binning, Goldstein, Garcia & Scattaregia, 1988; Gallois, Callan & Palmer, 1992; Wiley & Eskilson, 1985), but field studies (e.g. Graves & Powell, 1996) have produced mixed results. In a field study by Graves and Powell, the researchers found that male recruiters were not affected by sex similarity (or were able to suppress its influence), but that female recruiters sometimes are. Graves and Powell (1996) also found that female recruiters reported better interview experiences with female applicants and evaluated them more favorably. The similar-to-me effect is not confined to race and gender effects. Orpen (1984) examined the interview decisions of interviewers for sales positions in four large South African insurance companies. He found that the interviewers’ personal liking, actual similarity and perceived similarity to the interviewees were all directly related to the selection decision. Findings of this sort merely confirm what is already known: selection is not a perfect science and individuals make ‘similar-to-me’ assessments (Herriot, 1989a). As a result, legislation has emerged over the past forty years to protect those groups of people who are disadvantaged and who have been able to influence legislators about the needs of their case. Hence, most countries now have laws to protect genders from unfair discrimination. Many countries have laws to protect races and religions. The trend is towards legislation to protect other groups disadvantaged during selection such as the disabled and the elderly (Hogarth, 1992). With such weight of evidence pointing towards the similarity effect in selection, Schneider’s selection proposition has a strong intuitive appeal. Selecting for P–O fit P–O fit research after Schneider’s original proposition provides support that organizations want to select people who hold the values of the organization. For example, Rynes and Billsberry Schneider’s Selection Proposition 3 Gerhart (1990) examined the recruiting decisions of managers interviewing MBA graduates from an Ivy League business school for positions in mainly financial and general management. They separated out three different ways in which interviewers form their impressions of P–O fit. The first way is congruence between the values of the applicant and the recruiter. In other words, the recruiter makes the selection decision on fit based on a ‘similar-to-me’ judgment. The researchers termed this form of fit ‘idiosyncratic fit’ (Rynes & Gerhart, 1990, p. 16) because it is individual to each recruiter. They did not regard this form of fit as a form of P–O fit because it was about fit to an individual rather than fit to an organization. The second type of fit assessment that recruiters make is to compare applicants to some form of agreed notion of the characteristics needed to fit a particular organization. This is a ‘similar to us’ judgment and they termed this form of fit, ‘firm-specific employability’ (Rynes & Gerhart, 1990, p. 15). The third form of fit that recruiters might make is to compare applicants to a non-firm-specific form of work suitability. In other words, some people might be better fitted to work in all organizations than others are; this is a form of general employability (Rynes & Gerhart, 1990, p. 15). The researchers found that interviewers of different organizations made different selection decisions regarding firmspecific employability, i.e. every organization has a different set of requirements when assessing candidates for this form of employability, and that the interviewers were more stringent in their assessments of firm-specific employability than of general employability. In other words, interviewers seem to be more concerned to assess whether candidates are suited to the organization than they are to assess whether candidates are more broadly suited for the type of work. Bowen, Ledford and Nathan (1991) reviewed the recruitment and selection processes of three ‘high involvement’ organizations. The three organizations they concentrated on, AFG, Sun Microsystems and Toyota, all sought to recruit “self-motivated, committed people” (Bowen et al, 1991, p. 37) who share the values of the organization. Although these organizations may not be typical, it is interesting to note the length they are prepared to go to in order to recruit ‘whole people’ in the organization’s image. Toyota, for example, screens 50,000 applications for 3,000 jobs and “each employee hired invests at least eighteen hours in a selection process that includes a general knowledge exam, a test of attitudes toward work, an interpersonal skills assessment centre, a manufacturing exercise designed to provide a realistic job preview of assembly work, an extensive personal interview, and a physical exam” (Bowen et al, 1991, p. 36). The authors conclude by suggesting that the recruitment of individuals who fit the organization’s culture is a vital supplement to recruitment on grounds of person–job fit because it helps organizations create a distinctive culture which is maintained by people sharing the organization’s values and goals. In an organizational environment characterized by rapid and regular change, transition and development, the authors argue that recruiting ‘whole people’ who fit the overall organization, rather than those who fit a fixed set of task demands, is the only solution. Anecdotal evidence for selectors’ desire to select for fit and to seek homogeneity as an outcome of the recruitment process comes from Judge and Ferris (1992). In their review of selection in the 1970s and 1980s, these researchers captured references in the literature when organizational recruiters have expressed a desire to recruit for fit. Amongst the companies mentioned are Sears Roebuck, General Motors, and Hewlett-Packard. These corporations employed very different methods to make these fit assessments of applicants. Sears Roebuck used height as an indicator as an important staffing criterion, which seems inappropriate today. General Motors looked at employees’ interpersonal behavior as a guide to their fit and suitability for promotion. Hewlett-Packard relied on interviews to assess fit. These three studies demonstrate that some organizational recruiters are keen to select applicants who are ‘similar to us’: i.e. they want to select people who share the values of the Billsberry Schneider’s Selection Proposition 4 organization. However, whilst these organizations have the intention to select people in this way, these studies do not explore Schneider’s selection proposition directly; i.e. whether or not they actually manage to select for fit. Tests of Schneider’s selection proposition There have been very few studies of the selection phase of Schneider’s ASA framework. As mentioned earlier, most P–O fit studies relevant to Schneider’s framework have focused on the outcomes of P–O fit and examined people in employment, rather than people looking for employment. Two studies were found that investigated Schneider’s selection proposition, namely Adkins, Russell and Werbel (1994) and Cable and Judge (1997). These two studies follow the leads of Schneider (1987) and Chatman (1989) and use value congruence as the currency to explore the fit between applicants and the recruiting organization. The study by Adkins et al (1994) explored Rynes and Gerhart’s (1990) finding that firm-specific employability (i.e. P–O fit) is more important to recruiters than general employability or idiosyncratic fit. Rynes and Gerhart’s earlier study had not delineated the components or currency of this firm-specific employability; it had just found the effect. The study of Adkins et al (1994) used Chatman’s (1989, 1991) assertion that value congruence is central to fit to explore the nature of firm-specific employability. They investigated whether or not the congruence between applicants’ work values and those of the organization contribute to recruiters’ judgments of P–O fit. The researchers studied the interview decisions of corporate recruiters during the ‘milk round’ using the Comparative Emphasis Scale (CES: Ravlin and Meglino, 1987a, 1987b, 1989). Recruiters completed the CES twice; once for their own personal work values and once for their perceptions of the organizations’ work values. After each interview, the recruiters were asked to rate each applicant on P–O fit and general employability. Adkins et al (1994) replicated Rynes and Gerhart’s (1990) finding that it is possible to distinguish between firm-specific and general employability. These researchers also found that the recruiters’ assessment of the fit of applicants to the organization is significantly correlated with the recruiters’ own values, albeit weakly (r=.11, p<.05). Their results also suggest that congruence between applicants and the organization, as judged by the interviewer, did not influence recruiting organizations’ selection decisions, which is contrary to Schneider’s selection proposition. However, it is not a refutation of Schneider’s selection proposition for two reasons. First, the CES has its limitations as a P–O fit instrument as it captures just an abbreviated form of fit (Meglino, Ravlin & Adkins, 1989). Second, this study was concerned with recruiters’ perceptions of applicants’ fit. Hence, it does not measure whether or not applicants actually fit with organizations’ values, but rather measures selectors’ perceptions of that fit, which might be incorrect. The most extensive examination of P–O fit during selection was carried out by Cable and Judge (1997). These researchers considered the P–O fit assessments of interviewers during selection interviews. Building on the work of Schneider (1987), Rynes and Gerhart (1990), Bowen et al (1991) and Adkins et al (1994), they hypothesized that (1) interviewers’ perceptions of the P–O fit of their interviewees would be associated with their actual P–O fit, (2) these perceptions would positively affect their subjective assessments of P–O fit, (3) these perceptions would also positively affect their hiring recommendations, and (4) these hiring recommendations would influence their organizations’ hiring decisions. These four sequential hypothesizes produce a model that they tested this model by looking at the decisions of 38 interviewers recruiting on the college ‘milk round’ in an American university. Their main measure was the Organizational Culture Profile (OCP; O’Reilly et al, 1991), from which they removed items that were “too similar” (Cable & Judge, 1997, p. 550), which reduced the OCP from 54 to 40 items. Interviewers were asked to complete the ‘card sort’ as a ‘paper and Billsberry Schneider’s Selection Proposition 5 pencil’ test, ranking the 40 items in order of “most characteristic of my organization” to “least characteristic of my organization”. In addition, interviewers were asked to assess every person they interviewed using the same tool, but with the prompt being “to what degree is this a characteristic of the applicant I interviewed?” Interviewees were also asked to complete the same tool, but the prompt was changed to “how characteristic is this attribute of me?” In addition, the researchers asked the interviewers for their subjective assessment of every candidate they interviewed with a single item “to what degree does this applicant match or fit your organization and the current employees in your organization?” The researchers also had items for interviewers to report their impressions of the physical attractiveness and their ‘liking’ of every applicant. The results supported all four hypothesizes and the model. In addition, the researchers found that interviewees who were personally liked by the interviewers were more likely to be recommended for hire than less-liked interviewees were. Cable and Judge’s study is interesting because their findings support Schneider’s selection proposition. These findings suggest that interviewers base their P–O fit evaluations on the congruence between their perceptions of applicants’ values and their perception of the organization’s values. In addition, these P–O fit evaluations are significantly related to their selection decisions and those of their organizations, although their subjective P–O fit assessments (i.e. their gut feel about the applicants) are more influential than their assessment of actual value congruence (i.e. a calculation of fit based on the interviewee’s self-reported values and interviewers’ perceptions of the values of their organization). This is evidence that adds weight to the proposition that values are an important component of P–O fit. Despite the support these findings offer to Schneider’s selection proposition, this study is not a direct test of the proposition. Briefly stated, Schneider’s proposition says that organizations select people who share the values of the organization. In the Cable and Judge (1997) study, the source of the organizational values is the interviewers. These are not ‘checked’ or agreed with other organizational representatives. This is important because it is generally accepted that measures of the organizational values must capture communal agreement, rather than the views of individuals (Chatman, 1989, 1991; Rynes & Gerhart, 1990). It is interesting to understand how interviewers’ perceptions of their organizations’ values influences their assessments of P–O fit because of the central role of interviewers in selection decisions (Dipboye, 1992), but the interviewers’ assessment of their organizations’ values might be at variance with how other people in the organization view their organizations’ values. In effect, it is a form of what Rynes and Gerhart (1990) call ‘idiosyncratic fit’. Even though corporate interviewers tend to be organizations’ ‘great and good’ (Dipboye, 1992; Schneider 1987), these are still idiosyncratic perceptions of their organizations’ values. A second reason why the Cable and Judge (1997) study cannot be considered a direct test of Schneider’s selection proposition is that it only considers one variable in the selection decision. As mentioned above, interviewers’ decisions are important, but they are not the only factor. Other factors might include the decisions of other interviewers, applicants’ own decisions about whether to continue their interest in the position, other selection tests and filters some of which involve other individuals and others that do not, people conducting job analyses, and the impact of trade unions and other bodies. All of these factors might influence decisions about which applicants organizations select. Hence, the Cable and Judge (1997) study offers insights about a central process in the making of the selection decision and these insights are in-line with Schneider’s proposition, but it does not address the main proposition directly. A more direct test of Schneider’s selection proposition would be to measure the congruence between the values of applicants and organizations, thereby removing the surrogate role that interviewers’ values play. From this review, the following hypothesis would form a direct test of Schneider’s Billsberry Schneider’s Selection Proposition 6 selection proposition: The value congruence between applicants and organizational members will predict applicants’ performance in the selection process. Measurement of values Previous researchers have used the OCP to assess hypotheses associated with Schneider’s selection proposition (e.g. Chatman, 1991). However, the original card sort is impractical when there are a large number of remote and geographically dispersed respondents (Block, 1978; Kerlinger, 1986; Nunnally, 1978). To combat the impracticalities of card sorts in such circumstances, Cable and Judge (1996) transformed the OCP into a reduced paper and pencil sort that they argue replicates the ranking process of the Q-sort without the need to use cards. Unfortunately for the present study, their revised version of the OCP has been criticized by Barber and Wesson (1998) who examined the way people completed the instrument using concurrent verbal protocol analysis. They built and compared two questionnaires based on the items on the OCP. The first of these replicated Cable and Judge’s (1996, 1997) tool in which respondents are asked to place the items in order of their desirability. The second of these questionnaires employed a Likert-scale and each respondent was asked to indicate the desirability of each of the items. The researchers found that the paper and pencil version of the OCP card sort (1) failed to replicate the cognitive processes of the original card sort, (2) contained items the respondents did not understand, (3) forced respondents to guess, and (4) caused respondents to ask for clarification about the instructions. The Likert-scaled version, on the other hand, presented none of these problems. Barber and Wesson’s (1998) conclusion contained some strong views on the appropriateness of the paper and pencil sort as a substitute to the card sort. ‘These results lead us to conclude that the construct validity of the component parts of the OCP may be compromised by use of a paper and pencil Q-sort, and that the rating version presents far fewer concerns’ (Barber & Wesson, 1998, p. 98). ‘[W]e believe that the behavioral consequences of the frustration experienced by the Q-sort participants are probably understated. Furthermore, it did appear that the more demanding Q-sort task generated substantially different thought processes than did the rating task. [... A] pencil and paper sorting approach has significant drawbacks and no significant advantages relative to a rating approach. We strongly encourage adoption of a rating format of the OCP when actual card sorts cannot be used’ (Barber & Wesson, 1998, p. 99). When the OCP was being developed in the late-1980s, it was generally agreed that a person’s values were hierarchically organized according to their salience to the individual. Since this time, the values literature has moved on significantly and there is now much less unanimity about whether values are hierarchically organized or independently held (Meglino & Ravlin, 1998). Indeed, a main thrust of this literature is now suggesting that values are held independently of each other (Stackman, Pinder & Connor, 2000). Although the values literature is now divided between the two schools of thought, the general thrust of recent research points towards a view that values are held independently of each other. Accordingly, there is theoretical support to justify Barber and Wesson’s (1998) suggestion of a paper and pencil rating version of the OCP (i.e. a version that asks respondents to rate each value independently of others). This is the approach taken in the present study. However, it must be noted that the accepted view of values within the P–O fit literature strongly holds that values are hierarchically structured. In their study of the OCP, Barber and Wesson (1998) noted the difficulty that respondents had understanding the items in the set. Whilst most items were understood by Billsberry Schneider’s Selection Proposition 7 respondents, several items stood out. Initial trials of the OCP items at the sites chosen for the present study supported this view with many people having considerable trouble with their abstract nature. The OCP authors assert that the 54 items are “a comprehensive set of values that could be used to characterize both individuals and organizations” (O’Reilly et al., 1991, p. 495). The items that emerged are short value statements, mostly two or three word statements. Predominantly they are phrased as instrumental values, although some terminal values appear (Stackman et al., 2000). By phrasing the items in a conceptual way, the authors allowed respondents to interpret each value in ways specific to themselves. This idiographic interpretation creates concerns both about the commensurability of individuals’ responses to other respondents and creates potential difficulties for respondents’ understanding each of the values, as highlighted by Barber and Wesson’s (1998) findings. For example, take the OCP value “flexibility”. Does it mean that (1) individuals are flexible? (2) cultures are flexible? (3) managers are flexible? (4) staff are flexible? (5) values are flexible? or something else such as managerial rhetoric for “exploitation”? (Sisson, 1994). Resolving the issue of the appropriate phrasing of values is not isolated to the OCP. Schwartz (1992), for example, has categorized values at a conceptual or universal level extending the work of Rokeach (1973). Although his focus is on universal values, he acknowledges that these are too abstract to be used at the operational level (i.e. in questionnaires, card sorts etc.). In a later paper (Schwartz, 1994) he recommends that values be expressed in terms relating to behaviors suited to the specific environment in which the research instrument is being used. This is important, he argues, because not all universal values are suited to every situation and phrasing them in context-specific terms improves construct validity. Schwab (1980), who also conceptualized values at the conceptual and operational levels, supports this view.