understanding the determinants of employer use of selection methods

22
PERSONNELPSYCHOLOGY 2003.56 UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS STEFFANIE L. WILK, PETER CAPPELLI Management Department Wharton School of the University of Pennsylvania This study uses national probability data from over 3,000 employers to examine why employers differ in their use of employee selection meth- ods. Although the research on employee selection is voluminous, there have been only a handful of studies that look at the employers' selection decisions. In contrast to those other studies,we focuson characteristics of work as predictors of firms' decisions regarding selection practices. Beyond the relationship to the overall extent of selection methods used, we argue that specificwork characteristicswill affect the use of specific types of selection methods. We find, for example, that the greater the skill requirements of a position, the more likely that the establishment will use those types of selection methods that tap into the ability and skills of the applicants, namely, academic achievement and test perfor- mance. Discussion and suggestions for future research are offered. Employer decisions about the selection of employees are central to the operation of organizations and to outcomes that matter to individu- als, organizations, and society. The few studies that have examined em- ployer decisions about selection practices find a wide variance in both the methods and extent of selection use in organizations (BNA, 1983; Terpstra & Rozell, 1997). Those studies have focused on characteristics of the human resources department (Terpstra & Rozell, 1997) and the effect of national culture on firms' choices (Ryan, McFarland, Baron, & Page, 1999) to explain the variance in employer practices. Virtually no research has looked directly at characteristics of the work itself as a pre- dictor of selection practices. Although the staffing literature assumes a link, little is known about the relationshipbetween work characteristics We thank Nancy Rothbard and Paul Sackett for their helpful comments and sugges- tions. The data used in these analyses were collected with the support from the Edu- cation Research and Development Center program, agreement number R117QW11-91, CFDA 84.1170. as administered by the office of Educational Research and Improvement, U.S. Department of Education. The findings and opinions expressed here do not reflect the positions, policies of OERI, the U.S. Department of Education, or the Bureau of the Census. Correspondence and requests for reprints should be addressed to Steffanie L. Wilk, Management Department, Wharton School of the University of Pennsylvania, 3620 Locust Walk, 2000 Steinberg Hall-Dietrich Hall, Philadelphia, PA 19104-6370; [email protected]. COPYRIOHT Q 2003 PERSONNELPSYCH0u)oY. INC 103

Upload: steffanie-l-wilk

Post on 21-Jul-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

PERSONNELPSYCHOLOGY 2003.56

UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

STEFFANIE L. WILK, PETER CAPPELLI Management Department

Wharton School of the University of Pennsylvania

This study uses national probability data from over 3,000 employers to examine why employers differ in their use of employee selection meth- ods. Although the research on employee selection is voluminous, there have been only a handful of studies that look at the employers' selection decisions. In contrast to those other studies, we focus on characteristics of work as predictors of firms' decisions regarding selection practices. Beyond the relationship to the overall extent of selection methods used, we argue that specific work characteristics will affect the use of specific types of selection methods. We find, for example, that the greater the skill requirements of a position, the more likely that the establishment will use those types of selection methods that tap into the ability and skills of the applicants, namely, academic achievement and test perfor- mance. Discussion and suggestions for future research are offered.

Employer decisions about the selection of employees are central to the operation of organizations and to outcomes that matter to individu- als, organizations, and society. The few studies that have examined em- ployer decisions about selection practices find a wide variance in both the methods and extent of selection use in organizations (BNA, 1983; Terpstra & Rozell, 1997). Those studies have focused on characteristics of the human resources department (Terpstra & Rozell, 1997) and the effect of national culture on firms' choices (Ryan, McFarland, Baron, & Page, 1999) to explain the variance in employer practices. Virtually no research has looked directly at characteristics of the work itself as a pre- dictor of selection practices. Although the staffing literature assumes a link, little is known about the relationship between work characteristics

We thank Nancy Rothbard and Paul Sackett for their helpful comments and sugges- tions. The data used in these analyses were collected with the support from the Edu- cation Research and Development Center program, agreement number R117QW11-91, CFDA 84.1170. as administered by the office of Educational Research and Improvement, U.S. Department of Education. The findings and opinions expressed here do not reflect the positions, policies of OERI, the U.S. Department of Education, or the Bureau of the Census.

Correspondence and requests for reprints should be addressed to Steffanie L. Wilk, Management Department, Wharton School of the University of Pennsylvania, 3620 Locust Walk, 2000 Steinberg Hall-Dietrich Hall, Philadelphia, PA 19104-6370; [email protected].

COPYRIOHT Q 2003 PERSONNEL PSYCH0u)oY. INC

103

Page 2: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

104 PERSONNEL PSYCHOLOGY

and selection practices. This study examines the role of work character- istics on a firm’s selection practices using a unique data set on employ- ment practices at the establishment level.

Terpstra and Rozell(l993) found a positive relationship between a greater use of selection methods and organizational performance. This typifies the “more is better” perspective about the use of selection. An- other perspective is that not all selection methods are relevant (e.g., Guion, 1998), and organizations should be selective in what they use, particularly when selection can be costly. The tradeoff between using more selection methods to reduce the risk of selection mistakes and the cost of the methods forces employers to make choices about the methods to use and, subsequently, the information to gather about applicants. Or- ganizations use various methods to collect information about applicants.

Rather than use all methods all the time, organizations make deci- sions about the methods that will provide the most useful information for selection. Some are self-report, for example, resumes and applica- tion blanks, and others are collected by the organization, tests and work samples are examples. Each method provides a mechanism to gather particular types of information about applicants. If an Organization is interested in an applicant’s academic background, using transcripts and teacher references would be two methods that could provide that infor- mation. Likewise, if the organization is interested in an applicant’s work history, resumes, employer references, and the like are likely methods for collecting such information. Information about the ability and skill level of an applicant can be readily collected using tests and work sam- ples designed as part of the selection process. We believe that the na- ture of the work being selected for will have a positive relationship with specific types of selection methods used by firms. In effect, the charac- teristics of work will influence how firms select employees.

Schmitt and colleagues in reviewing the selection literature note that there is a dearth of research on the determinants of selection at the or- ganizational level (Schmitt & Chan, 1998; Schneider & Schmitt, 1986). Although a few studies, like Jackson, Schuler, and Rivero (1989), be- gan to examine factors at the organizational level-mainly industy type and organization size-that affect the choice of some human resource practices, treatment of selection in organization level studies has typi- cally been reduced to a question(s) about formal testing (e.g., Huselid, Jackson, & Schuler, 1997; Becker & Huselid, 1998).

lbo recent studies, however, have focused primarily on the organiza- tional-level selection decision. Terpstra and Rozell(l997) surveyed or- ganizations on their use of different methods of selection (e.g., cognitive tests, structured interviews) and found that the reasons for not using a particular selection practice varied based on resource constraints, legal

Page 3: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAF’PELLI 105

concerns, industry, and the knowledge of the human resources profes- sionals in the firm. Based on these descriptive results, they argue that selection in organizations is not scientifically performed and call for ad- ditional research on selection practices at the organizational level.

Ryan, McFarland, Baron, and Page (1999) examined factors external to the employer, specifically the role of national and cultural differences, to explain variations in employer use of selection practices. They found that some of the differences in selection use by firms in different coun- tries were explained by cultural differences in uncertainty avoidance and power distance. For example, a firm in a culture high in uncertainty avoidance is more likely to use tests and conduct interviews as part of the selection process. However, some findings were counterintuitive, for example, countries high in uncertainty avoidance used fewer selec- tion practices than organizations with low uncertainty avoidance, leading them to call for more research on organizational selection practices.

Although Ryan et al. (1999) looked at factors external to the organi- zation, we examine factors inside the organization as predictors of selec- tion practices. We go beyond Rrpstra and Rozell(l997) to look specif- ically at the role of work characteristics on decisions regarding selec- tion practices. Rather than assuming firms take a “more is better” ap- proach to selection, we examine how differences in the nature of work predict firms’ use of different types of selection methods in addition to their overall use of selection practices.

Hypotheses

Work characteristics tell us something about the content of the work and, thus, about the complexity of the job. Moreover, both the job analysis and utility analysis literatures suggest that work characteristics and job complexity are related to the selection process. Job analysis (e.g., Harvey, 1991), for example, is a mechanism for describing a job’s con- tent and complexity in such a way that the information can subsequently be used to define the required knowledge, skills and abilities, and other characteristics needed by applicants (e.g., Schmitt & Chan, 1998). Job analysis suggests that the more varied and complex the job, the more demands there will be on the selection process to collect relevant candi- date information. If the ideal selection process starts with a job analysis, then the characteristics of the work should be related to the selection decisions of firms.

Likewise, the parameters in a utility model suggest that the job con- tent is important in determining the usefulness of selection practices (see Boudreau, 1991). Utility models such as the Brogden-Cronbach-Gleser (BCG) model contain the Sa, parameter, or the value of the variance

Page 4: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

106 PERSONNEL PSYCHOLOGY

in performance of a worker in the position in question (see Brogden, 1949; Cronbach & Gleser, 1965). SO, is one of the most controversial parameters because of the difficulty in measuring it. Various methods of quantifying S . have included comparing performance ratings to ratings of job content through job analysis (Cascio & Ramos, 1986) and using wages as a proxy (Schmidt & Hunter, 1983). The argument is that jobs that are more demanding increase the need to reduce the variability in performance. That need, in turn, creates pressure to improve employee selection. Thus, an organization’s selection system is in part determined by these work characteristics. Further, selection methods should match the nature of the work such that less demanding work and more demand- ing work should predict the use of different selection practices.

Specifically, we examine the role of work characteristics on em- ployer choices about selection practices using three characteristics re- lated to work content and demands: skill requirements, training, and pay. Skill requirements have a direct relationship to work characteristics and work complexity because they directly capture various components of the work itself. Training and pay are signals of work characteristics and complexity. The amount of formal training required for the job sug- gests the job may or may not be easily mastered and suggests that the firm is willing to invest in it (e.g., Jones et al., 2001). The pay level sug- gests something about the human capital demands and the value of the work to the firm (e.g., Henderson & Fredrickson, 1996).

Where skill requirements are greater, other things equal, we should expect that employers have a greater need for effective employee selec- tion and may engage in more selection activities. This is particularly so where the skills are not easily captured by clear credentials (e.g., certifi- cations), as where teamwork and related skills are more important. For example, Klimoski and Jones (1995) note that teamwork requires an ex- panded set of knowledge, skills, and abilities (KSAs) that includes coor- dination, decision making, and planning with others. Similarly, Stevens and Campion (1994) find a more complex mix of the KSAs necessary for success in team-based work that includes interpersonal (e.g., conflict res- olution, communication) and self-management (e.g., goal setting, per- formance management) KSAs.

Certain work characteristics, for example, job rotation and self- managed teams, help identify situations where skill demands are greater and we predict that they should positively relate to firm’s overall use of selection methods.

Hypothesis la : Skill requirements will be positively related to the extent of use of selection methods used by firms.

Page 5: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 107

Further, although we believe that greater skill requirements should be related to the overall demand for selection information, we argue that greater skill requirements should create a special demand for selection information that taps directly into the skills and abilities of the applicant, such as academic records and tests. These methods will be more useful when skill requirements are greater than, for example, selection meth- ods that focus on a person’s previous work experience. Work experience often captures time-based elements of an applicant’s past (e.g., tenure) rather than data about the content of their work (Quifiones, Ford & %a- chout, 1995). This suggests that the link between work experience and skill requirements is weaker.

Hypothesis Ib: In particular, skill requirements will have a stronger pos- itive relationship with the use of selection methods focused on academic achievement and test performance than on methods that focus on work experience.

Where companies make substantial investments in employees, selec- tion may bring greater benefits by helping to ensure that the employees have attributes that will make such investments worthwhile. An illus- tration of this phenomenon is the positive relationship between selec- tion and training found in the literature. We know, for example, that engaging in more selection is likely to positively influence training per- formance (Hausknecht, Trevor, & Farr, 2002)’ which in turn influences subsequent task performance (Hanish & Hulin, 1994). This is presum- ably due to the fact that a more select group of workers will be more prepared to absorb and use training investments on the job. However, we know little about how firms’ decisions about selection relate to their decisions about training. In one of the few studies on the topic, Barron, Bishop, and Dunkelberg (1985) found that employer search, measured both in number of applicants interviewed and the time spent per inter- view, was positively related to the level of training provided to new hires. Similarly, we expect employers who invest in more employee training to also engage in more selection overall.

Hypothesis 2a: ’Ifaining will be positively related to the extent of use of selection methods used by firms.

Although we believe that training will have a positive effect on all types of selection, we argue that it will be especially true for selection practices that focus on the skills and abilities, rather than work history, of the applicant. First, the literature on the relationship between selection and training has used ability-based measures (Hanisch & Hulin, 1994; Hausknecht, Trevor, & Farr, 2002), and there is precedent for linking

Page 6: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

108 PERSONNEL PSYCHOLOGY

these particular types of selection and training. Second, similar to the arguments for skill requirements, the link between work history and training is weaker in that methods to tap an applicant’s work history may not capture the capabilities or trainability of the candidate. Selection methods that focus primarily on the performance of applicants in other learning situations, such as performance in school and on tests, should be more directly related to the amount of training provided. Further, if an organization is willing to invest a great deal in training a new worker, previous work experience may be less important.

Hypothesis 2b: In particular, training will have a stronger positive relation- ship with the use of selection methods that focus on academic achievement and test performance than on the use of methods that focus on work ex- perience.

An alternative argument might be that training a workforce is a sub- stitute for more intensive selection to identify applicant skills, the “make or buy” decision (e.g., Miles & Snow, 1984). Employers who invest less (more) in training might, under this argument, be expected to invest more (less) in selection to uncover applicants’ existing skills. However, context may make a difference. When skill demands are greater, invest- ment in both selection and training will be more likely to ensure that workers have the requisite characteristics and skills necessary. That is, at lower skill demands, the substitution argument is more likely than at higher skill demands.

Hypothesis 2c: Skill requirements will moderate the relationship between training and selection practices where training serves as a substitute for selection at lower skill requirements and not at higher skill requirements.

Wages influence the recruitment and retention of workers (e.g., High- house, Stierwalk, Bachiochi, Elder, & Fisher, 1999; Rynes & Barber, 1990; Williams & Dreher, 1992) and therefore play a role in the staffing process. Selection, though, is distinct from recruitment and attraction (see Rynes & Barber, 1990) and knowledge about the relationship be- tween wages and selection practices is limited. Wage levels may suggest something about the content of the work (e.g., Jones et al., 2001). Higher wages signals that the job may be more inherently demanding and, there- fore, require greater human capital and higher wages for that additional human capital. Henderson and Fredrickson (1996) found a positive re- lationship between pay and information processing demands of work for CEOs. Likewise, pay is a direct measure of the cost of a poor per- former-with a higher wage indicating greater cost of a poor selection decision. When labor is expensive (wages are high) the cost of an unpro-

Page 7: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 109

ductive employee is greater, other things equal. A $50 per hour worker who is poorly matched to the job costs employers five times as much in opportunity cost as a $10 per hour worker who is poorly matched. Work that is more valuable should presumably encourage the organization to be more careful with selection and engage in more extensive selection practices.

Hypothesis 3a: Wages will have a positive effect on the extent of use of selection practices used by firms.

We also expect that wage levels should have a positive relationship with specific selection measures. Wages are likely to be related to se- lection methods that focus on work experience (e.g., resumes) as they provide information about a candidate’s previous seniority, past wages, and, to some extent, external marketability. Likewise, a firm paying a high wage for a position would likely be interested in candidates’ skills and abilities. Thus, wages should also have a positive relationship with the use of selection methods focused on academic achievement and test performance as well as work experience.

Hypothesis 36: In particular, wages will have a positive relationship with the use of selection methods that focus on candidates’ academic achieve- ment, test performance, and work experience.

Using an establishment-level survey, we explore these hypotheses. Descriptions of data, analyses, and results follow. Interpretations and conclusions from the study are also offered.

Method

Data

The EQWNational Employer Survey. Our data come from establish- ment-level surveys of employment practices conducted by the US. Bu- reau of the Census for the National Center on the Educational Quality of the Workforce. The survey measured how work is actually done in the facility, not the policies that might exist in employee handbooks. The original National Employers Survey (NES I) was administered by the Bureau of the Census as a telephone survey in August and September of 1994 to a nationally representative sample of private establishments. The sampling frame for the survey was the Bureau of the Census’s Stan- dard Statistical Establishment List (SSEL), the most comprehensive and up-to-date listing of establishments in the United States. It examines establishments with more than 20 employees, which may be the most

Page 8: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

110 PERSONNEL PSYCHOLOGY

appropriate for understanding employment practices (those with fewer than 20 employees represent approximately 85% of all establishments in the U.S., but those with more than 20 employees account for approx- imately 75% of all workers; Lynch & Black, 1998). In addition, very small establishments are unlikely to be appropriate for or amenable to studying work practices because their working arrangements are often so informal and variable (changing day to day in response to even rou- tine developments such as absenteeism) as to make it difficult to respond in a meaningful way to questions about work systems. The survey over- sampled establishments in the manufacturing sector and establishments with more than 100 employees. Public sector employers, nonprofit insti- tutions, and corporate headquarters were excluded from the sample.

In administering the NES, the target respondent was the plant man- ager in the manufacturing sector and the local business site manager in the nonmanufacturing sector. Although a target respondent was used for efficiency in data collection, data were collected from records, files, and employees who were most knowledgeable about a particular type of data (e.g., financial, human resources). Gerhart, Wright, and McMahan (2000) suggest that if multiple respondents or sources should ideally be answering a questionnaire about HR practices, this should be made an explicit request of the main respondent. This is exactly what this study does. Not only does this improve the quality of these data but it also mit- igates the single-source problems of more traditional surveys (Huselid & Becker, 2000).

Computer Assisted Telephone Interviewing (CATI) was used to ad- minister each survey, which took approximately 28 minutes to complete. The respondents were likely to have had experience completing estab- lishment surveys, having been contacted by the Census on other occa- sions. The Census notified establishments that someone would be con- tacting them to gather information and a copy of the survey/topics were made available ahead of time. At the time of the survey collection, if data were available that the respondents did not have, they were encour- aged to provide that in a subsequent phone call. If the respondents did not know a piece of information and knew that the information would be difficult to get, they either provided an estimate or did not answer the question. However, they are repeatedly encouraged to give good data or no data at all. For example, several questions that involve dollar amounts or number of hours, figures that may not be easily accessible in some establishments, had the most missing data in our study.

The survey was repeated in August of 1997 (NES II), the data used here, and the logic of the sampling was similar. A total of 5,465 estab- lishments responded to NES I1 for a response rate of 78%.

Page 9: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 111

TABLE 1 Means and Standard Deviations for Selection Information Question

Once you have established a pool of applicants for a(n) (typical production job) opening, what sources of information do you use to evaluate the candidates? Rate the following on how regularly you obtain information from them (5 = always, 3 = regularly, 1 = never)

SD - M - Item - Applicant form 4.57 0.94 Resume 3.52 1.30 References of previous employes 3.84 1.19 References of former teachers or instructors 2.05 1.02 School transcripts 1.95 1.14 Other information on student achievement in 1.64 0.86

Interviews 4.61 0.79 Tests administered as part of the interview 2.66 1.60 Drug/alcohol screens 3.67 1.78

high school (e.g., examples of work)

Work samples 1.99 1.21

Vunu bles

Descriptions of dependent and key independent variables are pro- vided below. Most of the key variables reference a particular job in the firm, namely the key job of the production or front-line workers. This created a common job focus for respondents when they were asked about context and human resource practices and policies (Gerhart, Wright, McMahan, & Snell, 2000). Control variables are described at the end of this section.

Selection practices. These measures refer to selection practices fo- cused on the production or front-line employees (in the service sector, front-line employees are those with customer contact).

The survey asks: “Once you have established a pool of applicants for a (production level job) opening, what sources of information do you use to evaluate the candidates?” The scale, which ranged from 1 = never to 5 = ulwuys, captures regularity of use across 10 selection sources. The sources include application form, resumes, references, interviews, tests, and drug and alcohol screens (See Thble 1 for descriptive statistics on the ~D~. .uzc@ ~ese/ectionpraci?&~ weusd~apuz@ 44mgqpof practices examined in the personnel literature on employee selection. Guion (1998) reviews research that examines different sources of infor- mation about applicants. Our measure captures the major categories of assessment outlined by Guion (1991,1998). For example, OUT measure includes both cognitive and noncognitive tests (e.g., work samples), per- sonal history and record of accomplishment information from resumes, transcripts and other sources of biodata, and interviews.

Page 10: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

112 PERSONNEL PSYCHOLOGY

Several measures were created from these data. First, we created a means-use score comparable to the Ryan, McFarland, Baron, and Page (1999) measure “extent of method use.” This measure is the mean-use score across the 10 sources. A higher score indicates that an estab- lishment uses more information sources more often when selecting new front-line production workers.

In addition, the selection practices (listed in ’Ihble 1) were logically grouped into categories that represent the three types of information collected from employers: academic achievement, test performance, and work experience. Use of applications, interviews, employer refer- ences, and resumes were grouped together for the first factor. Because these capture aspects of an applicant’s past experiences (Guion, 1998), we labeled this factor Work Experience. Teacher references, other high school information, and transcripts were grouped on the second factor, which was labeled Academic Achievement. Tests used as part of the se- lection process and work samples tests were grouped on the third factor, which was labeled Test Performance. We did not include drug and al- cohol screening tests in this factor because they are not designed, as the other tests are, to measure skills.

A confirmatory factor analysis revealed good fit of this 3-factor model to the data with a goodness of fit index (GFI) of 0.96, an adjusted good- ness of fit index (AGFI) of 0.93, a root mean square error of approxi- mation (RMSEA) of 0.08 and a comparative fit index of 0.89. Further, the chi-square of the 3-factor model was 430.39 (df = 24; p < .01). We compared this model to a one-factor model (GFI = 0.89, AGFI = 0.82, RMSEA = 0.14, CFI = 0.67, chi-square = 1213.33, df = 27, p < .01) and the chi-square difference test indicates that the 3-factor model is a superior fit to the data.

Analyses were run using both factor scores and unit-weighted aver- ages of the items for each factor with no substantive change in results. Given that, we used the average of the unit-weighted scores in our anal- ysis for ease of interpretation of results. Therefore, Work Experience, Academic Achievement, and Test Performance measures ranged from 0 to 5.

The independent variables used to test the above hypotheses are presented below:

Skill requirements. Several variables are used to examine different aspects of the skill requirements of the jobs as measures of the work characteristics hypothesis. The first asks directly whether skill require- ments overall have risen in the last 3 years (1 = yes; 0 = no). Ideally, the question should ask about the level of skill requirements rather than their change, although respondents can assess the latter much more eas- ily. And other things equal, rising skill requirements in the past suggest

Page 11: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 113

higher requirements now. We use three variables to measure the ex- tent of interpersonal or cross-functional skill requirements: the percent- age of nonmanagerial and nonsupervisory (e.g., production and front- line) workers who work in self-managed teams, the percentage of such workers involved in regularly scheduled meetings to discuss work-related problems, and the percentage of these workers who are involved in job rotation. Teamwork of various kinds is central to the notion of high performance work, and job rotation may be seen as a proxy for cross- functional skills (Appelbaum & Batt, 1994).

Training. The average number of training hours provided by the employer to production or front-line workers in the last year is used to measure the extent of formal training at the establishment. The natural log of this variable is used in the analyses as the distribution of training hours did not meet OLS assumptions for normality.

Wages. Average hourly pay for production workers is used to mea- sure wages. The natural log of this variable is used in the analyses as the distribution of wages did not meet OLS assumptions for normality.

The control variables used in the analyses are presented below: We included industry dummies at the Zdigit SIC level to control for

differences in jobs and skills across industries. We control for various as- pects of establishment size (total sales, whether the establishment is part of a multiestablishment firm, the number of employees by size category, management levels in the organization, and span of supervisory control) to account for differences across establishments in the resources they could potentially devote to selection and the types of selection in which they may engage (Barber, Wesson, Roberson, & 'Ihylor, 1999). We also control for various characteristics of the employees-the distribution of women and minorities and the percentage of union coverage-because legislation covering protected groups as well as union contracts may af- fect hiring and selection practices independently from the factors under consideration here. We also control for average education levels of em- ployees by occupation in that education may also reflect important dif- ferences in jobs and in potential credential requirements for these jobs.

Attempts at validating data depend on the difficulty in obtaining in- dependent information to compare to the data being used. Many at- tempts at validating data are limited to examining the reliability of the existing data, the extent to which the information is internally consistent. We can go a step further here and wmpare the values for those variables used in these analyses that were also collected in the 1994 NES in order to provide some information about whether the data are reliable over time. Although none of the dependent variables were in the 1994 survey, many of the independent variables were, and the correlations between the 1994 and 1997 values for these variables indicate that the data are

Page 12: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

114 PERSONNEL PSYCHOLOGY

reasonably consistent across the two surveys, suggesting reliability. For example, the correlation for the level of unionization, which on average should be reasonably stable, is 0.90. The demographic variables, which might have changed somewhat with turnover across 3 years, are 0.86 for the percentage of women and 0.82 for the percentage of minorities and the correlation of overall employment levels is 0.70. Average hourly pay, which presumably is adjusted at least every year, is still correlated at 0.74. The skill requirement variables have lower correlations4.36 for self- managed teams, 0.35 for teamwork, and 0.36 for job rotation-no doubt because some firms have increased and some have decreased their use.

Analyses

OLS was used to estimate relationships between the measures of selection and the above independent variables. Hierarchical regression analyses were used to test each of the hypotheses with controls entered in the first step, and the hypothesized variable of interest in the final step. Both additional variance explained as well as covariate effects are reported. For Hypothesis 2c, interaction terms were added in the final step to test for moderator effects.

Results

Table 2 contains means, standard deviations and the intercorrelations between key study variables. The correlations among the four measures of selection use are positive and significant (range: 0.71 to 0.16, p < .01). The lowest correlations were between the three different types of selec- tion, indicating that, although they are related, each measure is captur- ing unique aspects of selection. Most of the work characteristics vari- ables (skill requirements, wages, and training) are significant and posi- tively related to all of the selection measures. The exception is percent- age use of job rotation, which is not significantly related to any of the selection variables.

The results for the main analyses are presented in Table 3.

Work Characteristics and Organizational Selection Practices

Table 3 contains the results for Hypotheses la, lb, 2a, 2b, 3a, and 3b. Significant additional variance was explained by skill requirements for the extent of use of selection methods supporting Hypothesis la. The sig- nificant measures were the percentage of workers involved in job rota- tion (p = 0.07, p < .05), whether or not skills have increased for this position (p = 0.13, p < .Ol), and the percentage of workers involved in work-related meetings (p = 0.09, p < .01). A measure of the use of self-

Page 13: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILKAND CAPPELLI 115

TABLE 2 Means, Standard Deviatwns, and Intercorrelations Among Study Rriables

Variable 1 2 3 4 5 6 7 8 9 1 0 1. Extent of use - 2. Acad. achieve. 70** - 3. Work exper. 71.' 32** - 4.Rstperform. 62** 25** 16'. - 5. %Jobrotation 03 03 -01 03 - 6. %Self-mgdteam 10" 06** 04* 13** 17** - 7. Skillchange 20** 13" 14** 13** 03 12** - 8. %Meetings 18** 11** 15** 10'. 13" 22** 12** - 9. Log(trainhrs) lo** 08** 05' 09** 05' 12.. 04. 11** - 10. (Pay) zo*+ 11" 07** 19.. -lo** 12;. 19** 11*+ 10- - Mean 3.06 1.89 4.14 2.32 20.04 16.09 0.54 57.00 (SD) (0.60) (0.81) (0.72) (1.13) (31.11) (29.86) (0.55) (42.89) - -

Nofe: Decimals have been omitted. * p < .05. **p < .01.

TABLE 3 Hiemrchical Regression Resuh- Work Charactektics,

Hypotheses la, lb, 2a, 2b, 3a, and 3b0

Extent of use Work Acad. ' Itst of selection methods experience achieve. performance

Controls Yes Yes Yes Yes A in RZ O.ll** O.ll** 0.07** 0.06**

% Workers involved in 0.07* 0.03 0.05+ 0.05

% Workers involved in 0.02 -0.02 0.00 O M +

Skill change& 0.13.. 0.11** 0.10** 0.07* % Workers involved in 0.09*' 0.07** 0.06+ 0.05+

A in Rzc 0.03.' 0.02- 0.02** om** 'If a i n i n g 0.10" 0.05 0.08.. 0.06*

A in RZC 0.01 * * 0.00 0.01'. 0.01.

A in Rzc 0.01' 0.01. 0.00 0.01 *

gci&

job rotation

self-managed teams

work meetings

I?!!! 0.10* 0.07' 0.00 0.09*

F 5.75** 4.77** 3.10'. 2.93.. N 1,207 1,207 1,207 1,207 n t d Rz 0.17 0.14 0.10 0.10

Standardized regression coefficients are presented. 'A rise in skills is coded 1, and a skill decline is 0. "Each change in R2 represents the most conservative test, the change in that block after

all the other blocks have been entered. + p < .10 * p < .05 **p < .01.

Page 14: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

116 PERSONNEL PSYCHOLOGY

managed teams was not a significant predictor of extent of selection method use. Hypothesis lb was not supported. Skill requirements were predictive of all types of selection equally (change in R2 = .02, p < .01 for all selection types), including Work Experience, which was not hypothesized.

Support is found for Hypotheses 2a and 2b, investment in training is positively related to the extent of use of selection methods (0 = 0.10, p < .01). Hours of formal training provided to front-line workers ex- plains significant variance for the use of Academic Achievement (p = 0.08, p < .01) and Test Performance (p = 0.06, p < .05) in- formation during selection but not for Work Experience (p = 0.05, p = ns). This supports Hypothesis 2b, namely that training would predict the use of more ability- and skill-related selection practices (Academic Achievement and Test Performance) more than the use of methods that tap Work Experience.

No support was found for Hypothesis 2c, the hypothesized moderator effects of training on skills. The interactions did not explain significant additional variance for either the extent of use of selection methods or for any of the finer-grained measures of different types of selection methods.

Hypothesis 3a was partially supported. Wages explained signifi- cant additional variance for the extent of methods use (p = 0.10, p < .05). It also explained significant additional variance for selection on Work Experience (p = 0.07, p < .05) and Test Performance (p = 0.09, p < .05), but not for Academic Achievement (p = 0.00, p = ns). The greater the wages, the more extensive the selection process and the more the likely information is drawn from work experiences and test perfor- mance, but not from academic achievement. This is contrary to Hypoth- esis 3b, which predicted a positive relationship between wages and d three of the different types of selection.

Discussion and Conclusions

Examining the selection process per se, and not simply the outcomes of selection, is important for several reasons. First, because the selec- tion process is one of the most powerful ways to shape the characteris- tics of organizations, understanding differences in selection practices is a good way to begin understanding the extent of organizational differ- ences. Schneider’s (1987) attraction-selection-attrition (ASA) model relies on selection decisions to explain how organizations become more and more homogenous. Homogeneity can have a negative effect on or- ganizational adaptability and, hence, on performance (Schneider, Gold- stein, & Smith 1995).

Page 15: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 117

Second, there are many different reasons why an organization in- vests in selection. This is a first attempt to evaluate the role of work characteristics. This research may provide some insight for future stud- ies given that different measures of work characteristics, skill require- ments, training, and pay, matter for both overall and specific types of selection practices.

Third, selection is often implicitly treated as a “more is better” prac- tice. This study, in contrast, helps shift the discussion toward examining the choices employers make about different types of selection practices. We find that that organizations use different types of selection methods depending upon the nature of the work being done and related practices such as training and wage levels. However, it is impossible to derive nor- mative inferences from these data. These results provide a description of what organizations are actually doing. When combined with future studies on selection and organizational effectiveness, these findings can improve our understanding of organizations use of selection and the out- comes that result.

Work Characteristics and Selection

Based on extensive theory connecting work characteristics and selec- tion (see Schmitt & Chan, 1998), we hypothesized that three measures of work characteristics, namely skill requirements, level of formal training, and pay, would affect employer selection practices. We found that all of these measures influenced overall selection use. Skill requirements, amount of formal training provided, and the level of pay were all predic- tive of the extent of use of selection methods. Further, skill requirements explained the greatest variance in the extent of use of selection methods. Unlike Ryan et al.’s (1999) study of national culture and selection, we found a strong relationship between context, here work characteristics, and the aggregate measure of selection, extent of method use. We be- lieve that is due to the closer natural link between work characteristics and selection processes as compared to the link between national culture and selection.

Because the type of selection should vary based on the context, we also examined the role of the three work characteristics measures on the use of three distinct selection methods: Work Experience, Academic Achievement, and Test Performance. These analyses provide a more fine-grained examination of the relationship between work characteris- tics and selection.

As work demands increase, as represented in increasing skill require- ments, training, and pay, organizations consistently rely on testing meth- ods in the selection process. This category includes objective and orga-

Page 16: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

118 PERSONNEL PSYCHOLOGY

nizationally provided tests that organizations may find more trustworthy and reliable when the job requires higher skills, provides more training, and demands higher pay. This supports our hypotheses.

We also believed that organizations would rely on academic achieve- ment-based selection methods for all measures of work characteristics as work demands increased. We found support for this for increas- ing skill requirements and formal training, but not for pay. Miller and Rosenbaum (1997) found that employers did not rely as heavily on school-related information during the selection process as on informa- tion that came from interviews and the organization’s own social net- works (e.g., employee referrals). This was due to the fact that organi- zations saw school-related information as less “trustworthy.” For this study, Work Experience and Test Performance have elements that are directly measured by the employer (e.g., interviews, tests) whereas Aca- demic Achievement is either self-reported or requires a link to a social network with schools and teachers that the organization may not have and the schools in their area may not foster. Why this is the case for pay and not for the other measures of work characteristics is puzzling. It may be that when the cost of a poor selection decision is clearly evident to an employer, as is the case when wages are high, they tend to focus on those types of selection methods over which they have more control. Future research should examine the selection decision-making process of orga- nizations to determine the influence of different work characteristics on the cognitive processes of decision makers within firms. School-to-work initiatives should examine the implications of a lack of a relationship be- tween pay and education information.

Finally, we believed that only pay would have a strong link to selec- tion methods focused on work experience. We found a link with pay, but also found a link with skill requirements. We know that work experience can have both time-related and more content-related aspects (Quifiones et al., 1995). We assumed in this case that the more basic time-related information comprised the Work Experience measure, given that it is the most common and more easily collected information. However, if it con- tained more complex information about amount and type of work, then the relationship between skill requirements and work experience makes sense. The lack of a relationship with training was hypothesized and, even in light of this argument, may indicate a substitution effect. That is, when organizations plan to provide more formal training to work- ers, previous work experience may be less important and indications of ability to learn (e.g., performance in school and on tests) may be more important.

Results from the different types of selection methods find that only when skill requirements increase do firms tend to use more of all three

Page 17: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILK AND CAPPELLI 119

types. Greater formal training increases a firm’s use of methods that fo- cus on academics and test performance, but higher pay increases a firm’s use of methods that focus on work experience and test performance. These results are controlling for the other types of selection methods. Thus, organizations tend to focus on different types of selection meth- ods when work demands, measured in three ways, increase.

An implication of these findings is that organizations are linking job and selection in a way to maximize the person-job fit of the incoming worker. As complexity of work increases, firms use more selection meth- ods and use selection methods that capture the applicant’s capability to do the work. Research on person-job fit has found that workers gravi- tate to jobs with complexity levels commensurate with their ability (Wilk, Desmarais, & Sackett, 1995; Wilk & Sackett, 1996). This study exam- ines the process from the organization’s perspective. However, selection should also improve fit between the applicant and other aspects of the work (e.g., personality fit) and organization. For example, selection is used to improve fit between an applicant’s values and the organization’s culture (Cable & Judge, 1997) and this study cannot determine how the firms’ selection decisions influence that.

Limitations

Unlike Terpstra and Rozell(1993), we do not have the data necessary to categorize these selection methods in terms of their validity. For example, we do not know if the interviews conducted by the firms are structured or unstructured. Likewise, we do not have data on the type or content of the tests used by the firms to determine how well constructed they are. Thus, we cannot say anything about the usefulness or the value of the selection choices of these establishments and future research on employer decisions about selection may want to consider this issue. Our models only account for limited variance (maximum R2 = 0.17) in firms’ selection decisions. More research is needed to understand the many determinants, among them work characteristics and validity, of this complex organizational choice.

Gerhart, Wright, McMahan, and Snell(2000) and Huselid and Becker (2000) have recently debated the reliability and validity issues involved in collecting human resource (HR) practices information. We believe these arguments do not directly undermine the reliability of our data. First, Gerhart, Wright, and McMahan (2000) make the point that their critique applies to studies based on large organizations where practices may vary considerably within the organization and not necessarily to smaller, plant-level organizations where variance is likely to be smaller. Smaller groups and more homogenous groups, they argue, are easier to

Page 18: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

120 PERSONNEL PSYCHOLOGY

assess and therefore the data are more likely to be accurate. Our obser- vations are for individual establishments, unlike the corporate and mul- tisite observations in most other studies. Over half (54%) of our sample consisted of establishments with less than 250 employees and 84% had less than 1,OOO employees. This is a much smaller sample than the 40,000 employees or even 4,000 employees that represent the average number of employees from the Gerhart, Wright, McMahan, and Snell(2000) and Huselid and Becker (2000) studies, respectively. As a result, our data do not mask the substantial variations that can occur within site.

Likewise, the Gerhart, Wright, McMahan, and Snell (2000) argu- ments are less applicable to studies that focus on a single group of em- ployees because of the greater homogeneity of the group to which the HR practices being discussed applies. This is the situation in our study. Our study asks the respondents to focus specifically on a particular group of worker: the front-line, production job. Gerhart, Wright and McMa- han (2000) illustrated the improved reliabilities from plant-level data focusing on a specific worker group (see their ‘Iilble 1).

Finally, as we noted in our paper, we had a single contact in each establishment, a problem noted in the debate. The contact was encour- aged to use multiple sources of data (e.g., archival, subject matter ex- perts, etc.) to respond to the questions posed and encouraged to collect data they did not have and provide that in a later interview, if neces- sary. If they did not know a piece of information, and that information would be difficult to get, respondents either provided an estimate or did not answer the question. However, they are repeatedly encouraged to give good data or no data at all. Although we cannot be certain that a contact did not simply estimate when he or she did not have the requi- site data, one variable in this study that would be difficult to have easy access to-the total training hours of the average production, front-line worker-was one that respondents most often skipped, suggesting that they did not simply guess at their responses. This may point to the fact that respondents took seriously the “good data or no data” request.

As an additional check on the single source issue, we statistically an- alyzed our data and checked for a method factor. Per the suggestion of Podsakoff and Organ (1986), we ran the Harman one-factor test. The re- sults of an unrotated factor analysis are examined to see if a single factor emerged illustrating a strong method factor. Our results indicated that 22 factors emerged from a factor analysis of our dependent and inde- pendent variables, with no single factor accounting for the majority of the variance. Further, Doty and Glick (1998) found that although com- mon method variance was prevalent in organizational research, common method bias was not so problematic. They argued specifically that when studies focus on “fairly concrete constructs, the use of multiple methods

Page 19: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WILKAND CAPPELLI 121

may not be essential” (p. 399). This study focuses on concrete constructs (e.g., counts, percentages of employees covered by practices, etc.) rather than more judgment-based, cognitively demanding constructs.

Thus, we do not believe we have a substantial reliability problem. We concur with the arguments in both the HR literature and the method literature that encourage future research to gather data from multiple raters and sources when possible.

We use listwise deletion of observations when information is miss- ing for any of the items used in the analysis in order to ensure that dif- ferences in results across coefficients are not due to differences in the sample being used for each coefficient (as opposed to pairwise deletion where only the missing item itself is excluded). Because we use a great many items, it is quite likely that any given observation will miss at least one item and therefore be deleted from the analysis. As a result, the sample size is reduced. We perform some simple analyses to examine whether the final sample used in these analyses is different from the re- maining cases, which as a whole is representative of the population of all establishments. Comparing the sample used in the analyses with those establishments not included and calculating 2-tailed t-tests of mean dif- ferences suggests these differences: our sample has trivially higher inten- sity of selection efforts (less than 0.20 difference on extent of methods used and less than one more method used [0.67]), is larger in terms of number of employees (4% less in the less than 50 employees category and 5% more in the 250-999 employees category), and generally rank higher on work characteristics. For example, there is a less than 3% dif- ference on percentage use of job rotation and self-managed teams and a less than 6% difference on percentage use of meetings. Our sample has higher pay for production-level workers (by less than 70 cents per hour) but there is no difference in formal training provided. (Complete results are available on request.)

The issue of causation is a potential concern for all of the above analyses given the limitations of cross-sectional data. ’Ib some extent, all of the arguments rely on the assumption that employee selection is designed to support, rather than drive,,the other practices in the orga- nization. It would, for example, be somewhat odd to imagine an em- ployer choosing an approach to employee selection and then adapting their work organization around it. Perhaps a more reasonable assump- tion is that all human resource practices are part of a system, that sen- sible employers think about them simultaneously, and that, in fact, the direction of causation is not straightforward but is closer to recursive, And in that case, more complicated dynamic modeling may be no better than simple OLS models of the kind presented here.

Page 20: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

122 PERSONNEL PSYCHOLOGY

Conclusions

This is, to our knowledge, a first test of how the characteristics of work relate to the actual selection practices of firms in a wide variety of industries. Although the literature on the effects of selection is volu- minous, virtually no research has been done on the determinants of the selection process itself, on the employer’s choices. We find that more demanding work is, for the most part, positively related to greater use of selection methods. Work characteristics that should factor into selection decisions, according to job analysis and utility theory models, do. This holds true for a large sample of firms, spanning a wide variety of indus- tries. However, the findings also reveal that selection is not just about gathering the most data about a candidate, a “more is better” proposi- tion. Depending on the work characteristic, firms rely more heavily on certain types of selection methods rather than all of them.

This study, with more varied measures of selection practices, pro- vides an opportunity for deeper understanding of the types of selection that relate to particular contexts. That is, the types of selection used by firms varies by context (e.g., work characteristics). Researchers study- ing organization selection decisions should be mindful of the role of and relationship between context and types of selection. Further, because this study focused on the entry-level selection decisions, future research should examine the role of context for hiring at different levels in the organization (e.g., managerial).

Likewise, there are implications for practice. Organizations should examine the various characteristics of work and from that consider the information most relevant and useful to that type of work in designing a selection process. For example, if the work has greater cognitive de- mands, then tests and academic achievements may be more useful. The more challenging the work or ambiguous the needed skill set, for ex- ample, like those associated with high performance work practices, the more extensive the use of certain types of selection. We find no evidence of selection serving as a substitute for practices such as training regard- less of job complexity. These results contribute to the growing literature on the relationships between employment practices and the factors that drive them.

REFERENCES

Appelbaum E, Batt R. (1994). The new American workplace: 7kansfoming work systems in the United States. Ithaca, NY: ILR Press.

Barber A, Wesson M, Roberson Q, ‘Iiiylor MS. (1999). A tale of two job markets: Organi- zational size and its effects on hiring practices and job search behavior. PERSONNEL PSYCHOLOM, 52,841-867.

Page 21: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

WIJX AND CAPPELLI 123

Barron J, Bishop J, Dunkelberg W (1985). Employer search The interviewing and hiring of new employees. Review of Economics and Statistics, 51,43-52.

Becker B, Huselid M. (1998). High performance work systems and firm performance: A synthesis of research and managerial implications. In Ferris GR (Ed.), Research in personnel and human resources management, (Vol. 16, pp. 53-101) Stamford, CI? JAI Press.

Boudreau J. (1991). Utility analysis for decisions in Human Resource Management. In Dunnette MD, Hough L (Eds.), Handbook of industrial and organizatwnalpsychol- OD, (Vol. 2, pp. 621-745).

Brogden H. (1949). When testing pays off. P E R S ~ N N E L P S Y ~ H ~ ~ ~ ~ ~ , 3,133-154. Bureau of National Affairs. (1983). Employee selection procedures. Washington, D C

Cable D, Judge T (1997). Interviewers’ perceptions of person-organization fit and orga-

Cascio WF. Ramos RA. (1986). Development and application of a new method for assess-

Author.

nizational selection decisions. Journal of Applied Psychology, 82,546561. . .

ing job performance in behavio&leconomi~~erms. Journal of Applied Psycholo& 71.20-28.

Cronbach L, Gleser G. (1965). Psychological tests andpersonnel decisions. Urbana.: Uni- versity of Illinois Press.

Doty DH, Glick W. (1998). Common methods bias: Does common methods variance really bias results? Organizational Research Methodr, I , 374-406.

Gerhart B, Wright P, McMahan G, Snell S. (2000). Measurement error in research on human resources and firm performance: How much error is there and how does it influence effect size estimates? PERSONNEL PSYCHOLOGY, 53,803-834.

Gerhart B, Wright P, McMahan G. (2000). Measurement error in research on the human resources and firm performance relationship: Further evidence and analysis. PER- SONNELPSYCHOLOGY, 53,8554372.

Guion R. (1991). Personnel assessment, selection and placement. In Dunnette MD, Hough L (Eds.), Handbook of industrial and organizational psychology, (pp. 327- 397), Palo Alto, CA: Consulting Psychologists Press.

Guion R. (1998). Assessment, measurement and prediction for personnel decawns. Mah- wah, NJ: Erlbaum.

Hanisch K, Hulin C. (1994). Tho-stage sequential selection procedures using ability and training performance: Incremental validity of behavioral consistency measures, PERSONNEL PSYCHOLOGY, 47,767-785.

Harvey RJ. (1991). Job analysis. In Dunnette MD, Hough L (Eds.), Handbookof industrial and organizutionalpsychology, (pp. 71-164). Palo Alto, CA. Consulting Psycholo- gists Press.

Hausknecht J, 2evor C, Farr J. (2002). Retaking ability tests in a selection setting: Impli- cations for practice effects, training performance, and turnover. Journal ofApplied

Henderson A, Fredrickson J. (1996). Information-processing demands as determinant of CEO compensation. Academy of Management Journal, 39,575-606.

Highhouse S , Stierwalk S, Bachiochi P, Elder A, Fisher G. (1999). Effects of advertised hu- man resource management practices on attraction of African American applicants.

Huselid M, Becker B. (2000). Comment on “Measurement error in research on human resources and firm performance: How much error is there and how does it influ- ence effect size estimates?” by Gerhart, Wright, McMahan, and Snell. PERSONNEL

Psycholo@, 87,243-254.

PERSONNEL PSYCHOLOGY, 52,425442.

PSYCHOLOGY, 53,835-854.

Page 22: UNDERSTANDING THE DETERMINANTS OF EMPLOYER USE OF SELECTION METHODS

124 PERSONNEL PSYCHOLOGY

Huselid M, Jackson S, Schuler R. (1997). Technical and strategic human resource manage- ment effectiveness as determinants of firm performance. Academy of Manugement

Jackson SE, Schuler RS, Rivero JC. (1989). Organizational characteristics as predictors of personnel practices. PERSONNELPSYCHOLOGY, 42,727-786.

Jones R, Sanchez J, Parmeswaran G, Phelps J, Shoptaugh C, Williams M, White S. (2001). Selection or training? A two-fold test of the validity of job-analytic ratings of trainability. Journal of Business and Pvcholo@, 15,363-389.

Klimoski R, Jones R. (1995). Staffing for effective group decision making: Key issues in matching people and teams. In Guvo R, Salas E and Associates (Eds.), Team effectiveness and decision making in organizations. San Francisco, C A Jossey-Bass.

Lynch L, Black S. (1998). Beyond the incidence of mining: Evidence from the Nufwnul Empbyer Survey. National Center on the Educational Quality of the Workforce, University of Pennsylvania, (working paper).

Miles RE, Snow CC. (1984). Designing strategic human resources systems. Organizational DyMmics, 13,3652.

Miller S , Rosenbaum J. (1997). Hiring in a Hobbesian World: Social infrastructure and employers' use of information. Work and Occuputiom, 24,498-523.

Podsakoff P, Organ D. (1986). Self-reports in organizational research: Problems and prospects. Journal of Manugement, 12,531-544.

Quiiiones MA, Ford XU, Teachout MS. (1995). The relationship between work experi- ence and job performance: A conceptual and meta-analytic review. PERSONNEL PSYCHOLOGY, 48,887-910.

Ryan AM, McFarland L, Baron H, Page R. (1999). An international look at selection practices: Nation and culture as explanations for variability in practice. PERSONNEL

Rynes S, Barber A. (1990). Applicant attraction strategies: An organizational perspective. Academy of Manugement Review, 15, 286-310.

Schmidt F, Hunter J. (1983). Individual differences in productivity: An empirical test of estimates derived from studies of selection procedure utility. Jorrmal of Applied Psychology, 68,407414.

Schmitt N, Chan D. (1998). Personnel selection: A theoretical approach. Thousand Oaks, C A Sage.

Schneider B. (1987). The people make the place. PERSONNELPSYCHOLOGY, 40,437-453. Schneider B, Goldstein H, Smith DB. (1995). The ASA framework An update. PERSON-

Schneider B, Schmitt N. (1986). Stuffing organizafiom. Glenview, IL Scott, Foresman. Stevens MJ, Campion MA. (1994). The knowledge, skill, and ability requirements for

teamwork Implications for human resource management. Journal of Munugement, 20,503-530.

Terpstra D, Rozell E. (1993). The relationship of staffing practices to organizational level measures of performance. PERSONNEL PSYCHOLOGY, 46,2747.

Terpstra D, Rozell E. (1997). Why some potentially effective staffing practices axe seldom used, hblic Personnel Management, 26(4), 483-495.

Wilk SL, Desmarais LB, Sackett PR. (1995). Gravitation to jobs commensurate with ability: Longitudinal and cross-sectional tests. Journal of Applied Pgcholw, 80, 79-85.

Wilk SL, Sackett PR. (1996). Longitudinal analysis of ability job complexity fit and job change. PERSONNELPSYCHOLOGY, 49,937-967.

Williams M, Dreher G. (1992). Compensation systems attributes and applicant pool char- acteristics, Academy of Manugement Journal, 35. 571-595.

Jow~, 40,171-188.

PSYCHOLOGY, 52,359-391.

NELPSYCHOLOOY, 48,747-773.