Turnover Intention and Turnover Behavior

Abstract

As the retirement wave of Baby Boomers approaches, retaining newly hired, mid-career, and retirement-eligible employees will be nearly as crucial as hiring top-quality new people. Using two large data sets on federal employees, we focus on whether human resource management (HRM) practices can affect turnover intention and whether they matter equally at all stages of the federal career. First, however, we test how well turnover intention predicts behavior using a 1% sample of the Central Personnel Data File (CPDF) and the 2005 Merit Principles Survey (MPS). Although turnover intention and behavior are correlated, they respond differently enough to demographic factors to suggest the need for caution in extrapolating the apparent impact of HRM practices from turnover intention to behavior.

Keywords

aging workforce issues employee attitudes behavior motivation federal government HRM grievance issues performance and performance appraisal turnover/organizational mobility workplace environment/culture

Introduction

As the federal government prepares for a retirement wave over the next decade (Lewis & Cho, 2011; Tobias, 2001), its focus on improving recruitment and hiring (Federal Times, 2009a, 2009b) should broaden to include retention of highly qualified new hires and retirement-eligible employees, the two groups with the highest turnover rates. One quarter of federal employees hired between 2006 and 2008 left within 2 years, and two thirds of those who resigned from federal service had less than 5 years of federal experience (Partnership for Public Service & Booz Allen Hamilton, 2010). Only 9% of those eligible to retire are actually retiring each year, but that is more than five times the rate of resignation for those who are not eligible to retire (Partnership for Public Service & Booz Allen Hamilton, 2010). Recruitment, selection, and training of new employees represent a heavy investment in human capital. Successful retention can lower that investment, increase its returns, and help smooth the transfer of institutional knowledge from older to younger generations.

“High commitment” human resource management (HRM) practices may offer federal managers steps they can take to lower turnover. Using the 2005 Merit Principles Survey (MPS), we examine how strongly HRM practices affect turnover intentions and whether those practices influence newly hired, mid-career, and retirement-eligible employees differently. First, however, we assess how well turnover intention predicts turnover. We examine whether age, federal experience, pay, education, race, and gender influence turnover intention and behavior in the same way, by running logit models on both the 2005 MPS and a 1% sample of federal personnel records for 1999-2009. We run models for the full workforce and separately for “new hires” (employees with fewer than 8 years of federal service), “mid-career employees” (those with 8 to 28 years), and the “retirement-eligible” (those with 29 or more years of federal service).

Our analysis shows that the probability of leaving federal service drops rapidly with age and federal experience, is lower for those with high pay relative to their education and experience, is higher for better educated employees, and is about the same across genders and races. Patterns differ enough for turnover intention that we should be cautious in extrapolating finding from turnover intention to behavior. Analysis of the 2005 MPS, however, suggests that merit-based rewards and fair performance appraisals are critical for retaining new employees, but only intrinsic motivation importantly postpones retirement among highly seasoned workers.

Retaining Employees during the Retirement Wave

Turnover matters, because recruiting, selecting, hiring, and training new employees is expensive. We have no good cost estimates for the federal service, but

[R]esearchers have reported that the financial cost to replace a private sector employee who leaves generally can run from 50 percent to 200 percent of the employee’s annual salary depending on the individual’s role, seniority, specialization, performance level and training received while on the job. (Partnership for Public Service & Booz Allen Hamilton 2010, p. 1)

Agency productivity also falls while a position sits open, waiting to be filled by a new hire. Meier and Hicklin (2008) demonstrate that high teacher turnover decreases school performance, and the special demands of their jobs mean that some federal employees may not reach full productivity for 5 years (Partnership for Public Service & Booz Allen Hamilton, 2010). Retention of top-quality employees also preserves institutional memory, which makes it easier to achieve organizational goals (Moynihan & Pandey, 2008), and passing that knowledge to new employees takes time. Because Baby Boomers make up a much larger share of the federal service than of the private sector workforce¹ (Lewis & Cho, 2011), the federal government may not have enough middle-level career bureaucrats to pass this institutional knowledge to new hires, and organizational performance could suffer (U.S. Government Accountability Office, 2009).

Turnover, HRM Practices, and Intrinsic Motivation

Federal managers can probably have their most important impact on retention during the retirement wave through HRM practices that make employees feel valued and fairly treated (e.g., Arthur, 1994; Gould-Williams, 2004; Haines, Jalette, & Larose, 2010; Huselid, 1995; Kim, 2005). Such HRM practices signal a commitment to employee development and well-being, which can increase loyalty and decrease the desire to quit (Gould-Williams, 2004). Arthur (1994) classifies these as “commitment” human resource policies and contrasts them with “control” HRM policies, which focus on reducing labor costs. “Commitment” systems seek to enhance desired employees’ attitudes and behaviors by strengthening their psychological linkages to their organizations (Arthur, 1994). They treat employees as human capital to be developed rather than as costs, and they establish fair and merit-based HRM practices to maintain and develop those valuable assets.

Five “high commitment” HRM practices are particularly important: making sure that hiring, training, performance appraisal, rewards, and grievance systems are fair, effective, and merit-based (Arthur, 1994; Gould-Williams, 2004; Huselid, 1995; Macky & Boxall, 2007). Selecting qualified people is critical to creating a competent workforce, and merit-based hiring is a fundamental principle of federal hiring, which should bolster perceptions of organizational justice and decrease turnover. Employees who receive training express lower intentions of quitting (Kim, 2005) and higher levels of job satisfaction and organizational commitment (Gould-Williams & Davies, 2005). Fair and accurate performance appraisal systems reduce turnover intention (Konovsky & Cropanzano, 1991), whereas manipulation and political abuse of appraisals increases it (Poon, 2004). Allocating both tangible and intangible rewards based on merit can increase employee motivation (e.g., Adams, 1965; Vroom, 1964). Grievance systems can suppress turnover intention because employees can “voice” their concerns through the system rather than “exit” the workplace. A fair grievance system will reduce turnover intentions by enhancing employees’ perceptions of procedural and distributive justice (Colquitt, Conlon, Wesson, Porter & Ng, 2001; Haines et al., 2010).

Because motivation is a core driver of employee attitudes and behaviors, meaningful work that promotes intrinsic motivation can discourage turnover. Intrinsic motivation, doing an activity for the enjoyment of the activity itself, has a larger impact than any kind of external motivation and has a major impact on decreasing turnover intention (Amabile, 1993; Deci & Ryan, 2004). Keaveny and Nelson (1993) demonstrate that intrinsic motivation buffers stress from work while decreasing turnover intention by enhancing work satisfaction. Richer, Blanchard, and Vallerand (2002) show that intrinsic motivation lowers turnover intention and turnover behavior by decreasing emotional exhaustion and increasing work satisfaction.

Managers influence turnover within a broader framework of individual motivation. Previous research shows that public sector employees are less likely to plan to leave their jobs when they have high levels of job involvement and intrinsic motivation (Bertelli, 2007), when they are satisfied with their opportunities for advancement (Kim, 2005), when their employer has good diversity policies (Moynihan & Landuyt, 2008), and when they have high levels of trust in their supervisors (Gould-Williams & Davies, 2005). These findings, like most empirical work on public sector turnover, rely on survey data regarding employees’ intentions. Managers, however, are more interested in turnover than turnover intention, and practices that decrease intentions may not have the same impact on turnover.

Turnover Intention, Turnover Behavior, and Demographic Factors

Linking employees’ perceptions of their work environments, their turnover intentions, and their actual decisions on whether to leave federal service would require a longitudinal research design that surveyed individuals, then tracked them for years—a costly effort that would raise ethical issues (Dalton, Johnson, & Daily, 1999). Thus, some organizational-level studies use actual turnover rates, but most management-oriented, individual-level research employs turnover intention. This vastly simplifies the research task but requires that turnover intention be a good proxy for actual turnover. A review of five meta-analyses examining the correlation between turnover intention and actual turnover finds correlations from .31 to .52 (Dalton et al., 1999). That is, turnover intention accounts for 9% to 25% of turnover. Most of these studies examine intention and behavior in private firms; we create the first estimates for the federal sector.

Economists focus more on actual turnover. They argue that employees change jobs when they expect the rewards to exceed the costs. Potential rewards can be both financial (e.g., higher pay and better benefits) and psychological (e.g., more satisfying work, better job security, and a more pleasant work environment). Costs can also be financial (e.g., moving to another city or losing pension benefits) and psychological (e.g., adapting to a new work environment and disrupting one’s life and those of one’s family).

This helps explain why young workers are more likely than older, more experienced employees to change jobs both in the general economy (e.g., Farber, 1999) and in the public sector (Lewis & Park, 1989; Moynihan & Landuyt, 2008). Younger workers are less likely to have found a good person–job fit or a good person–organization fit, making them more willing to seek other jobs that better use their skills and match their interests and values (DelCampo, 2006; O’Reilly, Chatman, & Caldwell, 1991). New employees tend to have lower salaries, increasing the chances they can find higher paying jobs with other employers. They have invested less in their pension plans (e.g., Ippolito, 1987), investments they would probably sacrifice in a job change. They are also less likely to be married or have children, making geographic moves less disruptive. Moynihan and Landuyt (2008, p. 122) call this the life cycle stability hypothesis: “Older and more settled employees with familial obligations are less likely to quit.”

More importantly, according to labor economists, younger employees’ “human capital” (the skills and knowledge that make them valuable to employers) is primarily “general” rather than “firm specific” (e.g., Becker, 1962; Farber, 1999; Oi, 1962). Labor economists argue that both workers and employers implicitly invest in employees’ skills through formal job training and informal, on-the-job skill development (e.g., Becker, 1962; Farber, 1999; Oi, 1962). In theory, workers accept lower initial pay in jobs where they can develop valuable skills, because they know their pay will grow faster in these jobs; and employers pay workers more than their actual productivity at the beginning of their careers as an investment in their future productivity (Oi, 1962). Education is a form of “general” human capital that develops skills that increase productivity in many types of jobs. Much job training (both formal and informal) is “firm specific”: workers learn the policies and procedures of a specific organization and develop relationships with specific coworkers and supervisors, which increases their productivity in their current agency but would have little value in another organization. Workers’ skills tend to become more firm specific as their careers progress, making turnover more costly for both them and their employers (Hashimoto, 1981; Parsons, 1972), making them less likely to leave.

Employees whose pay is high relative to their education and other qualifications should be less likely to seek or find higher paying jobs outside the federal government. Seen from another perspective, among federal employees at the same salary levels, those with more education (and more general human capital) should have better chances of finding higher paying jobs in the private sector and higher probabilities of leaving federal service, but those with more experience (and more firm-specific human capital) should be less likely to do so.

Because of societal expectations that women have primary responsibility for raising children and maintaining a home, women have traditionally been expected to have weaker attachments to the labor market and higher probabilities of quitting their jobs than men. With the dramatic rise in women’s labor force participation (Stier & Lewin-Epstein, & Braun, 2001) and somewhat more equal sharing of household responsibilities, that argument has weakened. In addition, men (especially White men) in the federal service may have better chances of finding higher paying jobs in the private sector. Although research by labor economists typically finds that federal workers earn more than comparable workers in the private sector (e.g., Asher & Popkin, 1984; Hundley, 1991; Moore & Raisian, 1991; Smith, 1976), the federal pay advantage is much larger at lower levels of the federal hierarchy (Fogel & Lewin, 1974, Hundley, 1991) and for women and minorities (e.g., Asher & Popkin, 1984). Lewis and Park (1989) found no gender differences in turnover chances among comparable White federal employees in the 1980s, and Moynihan and Landuyt (2008) found that women were less likely than comparable men to plan to quit in Texas state government, using more recent data.

Expectations for race differences are also mixed. Minorities may have higher turnover due to negative experiences in the workplace (Sackett & Dubois, 1991; Tsui, Egan, & O’Reilly, 1992), but they may also be less likely to leave public sector jobs because racial pay disparities are smaller and due process protections are stronger in the public than in the private sector (Bernhardt & Dresser, 2002; Llorens, Wenger, & Kellough, 2008).

In sum, we expect the probability that an employee will leave federal service will decline at a decreasing rate with age and work experience until the employee approaches retirement eligibility; then it will start to rise at an increasing rate. Holding other characteristics constant, turnover probabilities should be lower for higher paid employees and higher for more educated workers. Turnover rates are unlikely to vary much with either gender or race/ethnicity. If turnover intention is a good proxy for turnover behavior, the same patterns should hold for workers’ plans to leave federal employment. The dramatic differences in quit rates at different stages in the life cycle (e.g., Farber, 1999; Moynihan & Landuyt, 2008) suggest that HRM practices and motivation may have varying impacts at different life stages. Both are likely to have more impact among new hires and the retirement-eligible, when quitting or retiring is most common.

Data and Method

Data

Central personnel data file (CPDF)

The U.S. Office of Personnel Management (OPM) maintains the CPDF as the government’s central personnel records. Every April, OPM draws a 1% random sample of the CPDF for study purposes. We restrict our sample to full-time, white-collar workers, but we combine data for 1999 through 2007 to boost the sample size.² Because OPM selects the sample based on the final three digits of employees’ social security numbers, employees in the sample one April remain in the sample in all other Aprils in which they are federal employees, allowing us to identify actual turnover, but with some error. We code employees as leaving federal service if they are in the CPDF sample 1 year and gone for at least the next 2 years.³ The vast majority of these employees quit, but some were fired, took extended leave, went on disability, or died. Thus, our measure includes both voluntary and involuntary turnover.

MPS 2005

In summer and fall 2005, the U.S. Merit Systems Protection Board (MSPB) invited 74,000 federal employees to participate in the on-line MPS 2005 (U.S. MSPB, 2007) and 50% (36,926) responded. We restrict the sample to white-collar employees and weight all our analyses. We measure turnover intention by combining two questions. Employees who responded they were “somewhat” or “very likely” to “leave [their] agency in the next 12 months” and that they would be “retiring from federal service” or “resigning from federal service” are coded as 1. Others are coded as 0.

Method

We begin by assessing how well turnover intentions match turnover behavior. Although employees should be more likely to consider leaving than to leave, the two should be strongly correlated. We calculate the percentage leaving federal service and the percentage saying they are at least somewhat likely to leave federal service at each age and experience level, and in each agency. We calculate correlation coefficients within each grouping.

Because all these variables simultaneously influence one’s probability of leaving federal service (or intending to do so), we also run logit analyses for both dependent variables. All models include federal experience, age, education, salary, gender, and race/ethnicity. We measure age and federal service in years and years-squared to capture their curvilinear effects—both turnover behavior and intention should drop rapidly early in the career, plateau in the mid-career, and rise among the retirement-eligible.⁴ We use four dummy variables for education: some college,⁵ bachelor’s degree, master’s degree, and doctorate or professional degree; high school or less is the reference group. We include dummy variables for gender, race, and ethnicity, with White males as the reference group. We measure pay as the natural logarithm of annual salary, on the assumption that percentage (rather than dollar) increases in salary have reasonably constant impact. Because the CPDF model combines several years of data, we include dummy variables for each year, with 2005-2006 as the reference year (to be comparable to the MPS); these variables automatically correct salaries for inflation.

We compare the logit coefficients in the two models to see whether the same variables have the same effects on turnover behavior and intention. We also calculate the predicted probabilities of leaving and of intending to leave for each individual in each data set (restricting the sample to 2005 employees in the CPDF) based on the two models.⁶ We then calculate the correlation between the two probabilities to assess how well turnover intention matches turnover behavior.

To explore what management can do to retain federal employees; we then add the high commitment HRM practices and intrinsic motivation variables to the logit analysis of turnover intention in the 2005 MPS. Table 1 provides details on how we measure these variables. For variables based on multiple survey items, we calculated the means of those items to create the indices. For each multiitem variable, all items loaded on a single factor in a principal component factor analysis; Cronbach’s alpha was greater than .70 for all except intrinsic motivation (.66).

Table 1

Measurements of Variables of MPS 2005

Turnover intention	How likely is it that you will leave your agency in the next 12 months?
	Very likely/somewhat likely/neither likely nor unlikely/somewhat unlikely/very unlikely/don’t know/can’t judge
	If you leave your present job, would you be: Retiring from federal service? /resigning from federal service?/
	moving to another job within the federal service?/not sure
	Note: Mean: 0.14; Std. Dev.: 0.35. The dependent variable is coded 1 for employees who positively responded to the first question (“likely” or “very likely) and answered “retiring from federal service” or “resigning from federal service?” to the second question. By this coding, we filter respondents who are likely to leave the federal service whether they want to find the job outside the federal government or retire
HRM practices
Merit-based hiring	To what extent do you think your supervisor will exercise each of the following authorities in a fair and effective manner?
	—Rating the qualifications of applicants for jobs
	—Selecting people for vacancies or promotions based on their qualifications
	Note: Cronbach’s alpha: .91; Mean: 3.96; Std. Dev.: 1.13
Training opportunities	I am given a real opportunity to improve my skills in my organization
	I receive the training I need to perform my job
	Note: Cronbach’s alpha: .79; Mean: 3.71; Std. Dev.: 0.95
Merit-based rewards	If your team performs well how likely is it that you will receive a cash award or pay increase?
	My organization takes steps to ensure that employees are appropriately paid and rewarded
	If I perform well, it is likely I will receive a cash award or pay increase
	Note: Cronbach’s alpha: .83; Mean: 3.14; Std. Dev.: 1.05
Fair and accurate performance appraisals	In my work unit, performance ratings accurately reflect job performance The standards used to appraise my performance are appropriate
Objective measures are used to evaluate my performance
Note: Cronbach’s alpha: .79; Mean: 3.49; Std. Dev.:0.93	Fair grievance system
I trust third party investigative or adjudicatory agencies (such as the OSC, EEOC, FLRA, MSPB) to respond appropriately to complaints
I believe that the current employment grievance system, if I had occasion to use it, would be fair
I believe that the current employment appeals system, if I had occasion to use it, would be fair
	Note: Cronbach’s alpha: .89; Mean: 3.27; Std. Dev.:0.83
Motivational factor
Intrinsic motivation	My agency’s mission is important to me
	The work I do is meaningful to me
	How important are each of the following in motivating you to do a good job?
	-Desire to help my work unit meet its goals
	-Personal pride or satisfaction in my work
	Note: Cronbach’s alpha: .66; Mean: 4.50; Std. Dev.:0.47

To test whether the effects of the independent variables change over the life cycle, we run the analyses separately for three levels of federal experience: less than 8 years of federal service (when turnover is high but dropping rapidly), 8 to 28 years of federal experience (when turnover is low and relatively stable), and 29 or more years of service (when most employees are eligible to retire and turnover rises again). We are particularly interested in varying effects of HRM practices and intrinsic motivation.

Findings

Descriptive Statistics

Turnover is strongly related to both federal experience and age. One quarter of employees in their first April in the federal service are gone by the next April, 16.5% of those in their second April are gone by their third, and 12.4% of those in their third April are gone by their fourth (Figure 1). By their ninth April, the turnover rate drops to about 5%. It stays below 5% annually throughout employees’ second decade of federal service and below 6% for most of their third decade. It then jumps to nearly 10% for those in their 30th April. Once employees complete 30 years of federal service, turnover is more than 10% annually, rising to about 18% in the 35th year. Turnover also drops quickly with age among employees in their 20s, plateaus among those in their 30s and 40s, and jumps as employees hit age 55 (Figure 2). Exit rates drop from 24% at age 23 to 10% at age 31, hit a minimum of 3.6% at age 46, stay below 5% through age 53, then more than double. Annual turnover (predominantly retirement) stays above 10% from age 54 on and is typically above 15% annually in the 60s.

Figure 1.

Turnover by level of experience

Figure 2.

Turnover by age

Employees are more likely to consider leaving the federal service than to do so. Overall, 7.1% of federal employees exited annually, but 10.2% of MPS 2005 respondents said they were at least “somewhat likely” to look for a job outside the federal government or to retire within the next year. Surprisingly, overestimating one’s likelihood of leaving is especially common late in the career. As shown in Figure 1, federal employees in their first 7 years of service were more likely to leave federal service than to say they were likely to do so. From 8 through 28 years of service, turnover and turnover intention are reasonably similar, but with intention consistently and increasingly higher. Both turnover and turnover intention jump as employees complete 30 years of service, but plans for retirement exceed actual retirement 2-to-1. Between 1999 and 2007, 12.8% of employees with more than 28 years of federal service retired annually, but 27.9% of similarly experienced MPS 2005 respondents thought it was at least “somewhat likely” they would retire or resign within the next year (Figure 1). Among those with 8 to 28 years of service, only 4.7% left annually but 6.5% gave serious consideration to leaving. For those with fewer than 8 years of service, however, 9.1% left annually but only 6.4% expressed a serious intention to leave. The pattern is similar for age (Figure 2): turnover exceeds turnover intention through about age 35, the two are quite similar until age 55, then much higher percentages seriously considered retiring than did so.

Part of the reason might be related to some problems with the representativeness of the MPS 2005. White-collar federal employees who responded to MPS 2005 are older, more educated, and more experienced than the federal civil service generally. Employees less than 30 and with fewer than 8 years of experience are especially underrepresented. This may help explain why more new hires quit than said they were thinking about quitting: those planning to leave federal service may have been less likely to fill out the survey.

How Well Do Turnover Intention and Behavior Match?

Explaining turnover intention is useful primarily if it predicts turnover behavior. Figures 1 and 2 show fairly substantial differences between intention and behavior, but when we use the particular age as the unit of analysis, the correlation between the two is .70. The patterns are more distinct for experience levels; with the year of service as the unit of analysis, the correlation is .38. The final two columns of Table 2 show turnover behavior and intention: though the rates differ by race and sex, when the race–sex combinations are the unit of analyses, the correlation between intention and turnover is .80. Both correlations are significant at the 0.01 level. The results give us some confidence of using turnover intention as a proxy of actual turnover. As both turnover behavior and intention are measured with error, the true correlation is probably stronger.

Table 2

Turnover and Turnover Intention by Agency (Percentage)

Agency	Turnover behavior	Turnover intention
Commerce	10.46	11.53
Defense	10.34	13.83
Homeland security	9.81	12.05
Treasury	9.49	13.82
Interior	9.12	14.74
Veterans affairs	9.09	17.15
Agriculture	9.04	13.55
Justice	8.41	11.40
Air force	8.20	12.78
Army	7.87	13.95
Health and human services	7.64	15.57
State	7.31	17.57
Office of personnel management	7.12	13.30
Education	6.91	14.95
Housing and urban development	6.90	17.00
Labor	6.67	19.30
Navy	6.64	11.69
General services administration	5.92	14.33
Energy	5.90	17.62
Federal deposit insurance corp.	5.80	12.83
Social security administration	5.49	16.51
Transportation	5.08	16.72
Environmental protection agency	4.48	16.37
NASA	3.93	13.71
Mean	7.40	14.68
Standard deviation	1.83	2.19

However, using the 24 agencies that were in both data sets as the units of analysis, we found a negative correlation of –.16. Table 2 provides the list of agencies in the order of actual turnover. Department of Commerce and Department of Defense show the highest turnover of more than 10%, but the average turnover intention was not highest in these agencies. Department of Energy has the highest average turnover intention (17.62%), but its actual turnover (5.9%) was among the lower group of actual turnover. Thus, we should be cautious in extrapolating research results from turnover intention to turnover behavior. Tables 3, 4, and 5 present logit models for turnover using the CPDF and for turnover intention using the 2005 MPS. Turnover decreases at a decreasing rate with both age and federal experience, then increases at an increasing rate with both—consistent with Figures 1 and 2. Race and gender have little impact in either model (Table 3). Asian females are less likely than similar White males both to quit and to intend to quit. Black females and Asian males are also less likely than comparable White males to leave over this period, but this does not show up in turnover intention.

Table 3

Logit Models for Turnover and Turnover Intention, Full Sample

	Turnover behavior	Turnover intention
Natural logarithm of salary	−0.477** (10.30)	0.414* (2.37)
Federal experience	−0.111** (20.00)	−0.056* (2.32)
Federal experience squared	0.003** (21.78)	0.002** (3.77)
Age	−0.185** (16.92)	−0.267** (4.78)
Age squared	0.002** (18.62)	0.003** (6.32)
Some college	−0.113** (3.22)	−0.042 (0.25)
Bachelor’s degree	−0.096** (2.58)	0.052 (0.38)
Master’s degree	0.001 (0.02)	−0.007 (0.04)
Doctorate or equivalent	0.139* (1.98)	−0.256 (1.18)
Asian female	−0.249* (2.54)	−0.822* (2.48)
Black female	−0.214** (4.63)	−0.061 (0.30)
Latina female	−0.212* (2.50)	−0.400 (1.08)
White female	0.011 (0.35)	0.008 (0.07)
Asian male	−0.268** (3.00)	−0.102 (0.28)
Black male	−0.142* (2.24)	0.000 (0.00)
Latino male	−0.084 (1.21)	−0.181 (0.61)
McFadden’s pseudo-R²	.069	.150
Observations	114,086	28,946

Note: Robust z statistics in parentheses. Data for 2000-01 through 2007-08. Models also include dummy variables for year and for Native American males and females.

significant at 5%. **significant at 1%.

Table 4

Logit Models for Turnover, by Length of Federal Service

	0-7 years of service	8-28 years of service	29+ years of service
Natural logarithm of salary	−0.889** (10.79)	−0.388** (5.63)	0.192* (2.15)
Federal experience	−0.362** (9.50)	−0.110** (5.03)	0.077 (0.79)
Federal experience squared	0.032** (6.46)	0.003** (5.03)	−0.001 (0.61)
Age	−0.126** (8.69)	−0.204** (8.65)	1.181** (11.13)
Age squared	0.001** (7.88)	0.003** (11.34)	−0.009** (10.21)
Some college	−0.321** (4.96)	−0.052 (1.02)	−0.088 (1.28)
Bachelor’s degree	−0.261** (4.14)	−0.117* (2.01)	−0.168* (2.18)
Master’s degree	0.066 (0.76)	−0.055 (0.71)	−0.381** (3.69)
Doctorate or equivalent	0.649** (6.35)	−0.267* (2.22)	−0.667** (4.09)
Asian female	−0.020 (0.14)	−0.399** (2.92)	−0.430 (1.64)
Black female	0.057 (0.74)	−0.197** (2.79)	−0.417** (4.72)
Latina female	−0.269* (2.20)	0.021 (0.18)	−0.306 (1.31)
White female	0.209** (3.74)	−0.047 (0.98)	−0.034 (0.53)
Asian male	−0.194 (1.48)	−0.308* (2.10)	−0.212 (1.03)
Black male	−0.005 (0.06)	−0.003 (0.04)	−0.287* (2.51)
Latino male	−0.280* (2.39)	0.041 (0.37)	0.206 (1.24)
McFadden’s pseudo-R²	.056	.065	.059
Observations	28,206	71,355	14,525

Note: Robust z statistics in parentheses. Data for 2000-01 through 2007-08. Models also include dummy variables for year and for Native American males and females.

significant at 5%. **significant at 1%.

Table 5

Logit Analysis for Turnover Intention with Managerial Factors

Variables	0-7 years of service	8-28 years of service	29+ years of service
Training	−0.243 (-1.54)	−0.024 (-0.25)	0.025 (0.23)
Merit-based reward	−0.399** (-2.63)	−0.087 (-1.08)	−0.100 (-1.03)
Performance appraisal	−0.363* (-2.38)	0.105 (0.96)	−0.127 (-1.05)
Selection	0.248* (2.22)	−0.107 (-1.46)	−0.101 (-1.22)
Grievance system	−0.236 (-1.30)	−0.034 (-0.34)	0.089 (0.84)
Intrinsic motivation	−0.516 (-1.85)	−0.716*** (-4.87)	−0.714*** (-3.84)
Constant	5.633 (1.27)	6.358* (2.09)	−42.377*** (-3.84)
Observations	3,385	14,964	4,695
Pseudo R²	.1712	.1773	.0983

Note: Robust z statistics in parentheses. Demographic variables are controlled but omitted in the table.

significant at 5%. **significant at 1%.***significant at 0.1%.

The models differ markedly on earnings and education, however. Higher salaries decrease the probability of leaving the federal service but increase the probability of thinking about it. The effects of education are not statistically significant in the turnover intention model and are odd in the turnover model: those with some college or a bachelor’s degree are less likely to leave than those with high school or less, but those with doctorates or professional degrees are more likely to do so. Only the latter pattern matches our expectations.

Because age and experience have much stronger effects than the other variables in both models, however, the predicted probabilities of leaving and planning to leave are strongly correlated: .78 in the CPDF and .87 in the 2005 MPS.⁷ This strong correlation indicates that studying turnover intention should provide insights into actual turnover. The strikingly different effects of salary, education, race, and sex in the two models, however, suggest the need for caution—as do the differences in Figures 1 and 2. In addition, when the analyses are performed separately for the three stages of the federal career (as described in the next section), the correlation falls to essentially zero for new hires, though it remains .85 or higher for the middle and late career.

Early versus Late Career

Given the odd disparity in the effects of salary in the behavior and intention models, and given the strikingly different turnover rates at different levels of federal experience, we reran the models separately at the three stages of the federal career highlighted in Figure 1. The effects of earnings and education are markedly different across the three stages (Table 4). Salary has a strong negative impact on turnover early in the career, a much weaker negative impact in the mid-career, and a weak positive impact on retirement. High starting salaries and rapid promotions can meaningfully decrease turnover early in the career. A person with the mean characteristics of employees in their first 8 years had a 7.3% probability of leaving within 1 year; a 10% salary boost would lower that probability to 6.7%. Salary has a much smaller impact in the mid-career; for those with the mean characteristics of this group, the predicted probability of quitting is only 3.5%, and raising pay by 10% would only lower that probability to 3.4 percent%. Among those eligible to retire, higher salaries may just mean higher pensions and a more affordable retirement, though the effects are weak. A 10% pay raise at the mean increases the probability of retirement from about 12.0% to 12.2%

The effects of education in the early career resemble the odd pattern of Table 3: holding other characteristics (including salary) constant, employees with doctorates and professional degrees are the most likely to quit, followed by those with high school or less, followed by those with a bachelor’s degree or some college. By the middle career, the effect of education on turnover is negative. Among the retirement-eligible, education has a strongly negative impact on turnover—having a doctorate rather than a high school education essentially cuts the probability of retiring in half (holding the other variables at their means). If better educated employees have more interesting work, this pattern is consistent with extrinsic rewards having the strongest impact on turnover in the early career, but with intrinsic rewards playing an increasingly important role in retention as the career progresses.

Table 5 presents the effects of high commitment HRM practices and intrinsic motivation on turnover intention at different stages in the federal career, using logit models that control all the demographic factors in Tables 3 and 4. Among employees with fewer than 8 years of federal experience, employees’ perceptions that rewards are based on merit and that performance appraisal is fair and accurate reduce their intentions to leave the federal service. One-standard-deviation increases in each of these variables reduce the probability of leaving the federal service by 1.9 and 1.5 percentage points. Other HRM practices and intrinsic motivation are not significant.

HRM practices have little impact on the turnover intentions of mid-career employees (those with 8-28 years of service). Only intrinsic motivation is significant in column 2; a one-standard-deviation increase in intrinsic motivation decreases the probability by 1.8 percentage point. Intrinsic motivation is also the only significant factor among those eligible to retire, for whom a one-standard-deviation increase in belief in agency mission and meaningful work decreases the intention to leave by 7.0 percentage points.

Conclusion

Turnover intention seems to be a reasonable proxy for actual turnover. Turnover intention and behavior are strongly positively correlated using age and experience as units of analysis. Using individuals as the units of analysis, predicted probabilities of leaving and of intending to leave are correlated at .7 or higher, except in the early career, where self-selection bias in the MPS 2005 may be a problem. However, we also found some contrasting results. Turnover intention and behavior are negatively correlated by agency. In addition, one observes discrepancies in the predicted effects of education and salary (and to a lesser extent race and sex). These results argue for the need for caution in applying the implications of the turnover intention analysis to turnover behavior. This is especially true for the early career, where HRM practices appear to matter most.

Turnover rates vary strongly with age and experience. Turnover is concentrated among relatively new employees and those eligible to retire. High turnover may be healthy for both employees and the government in the early career, if it weeds out those whose skills, interests, and motivations don’t match well with the government’s needs. Extrinsic rewards—both their amount and the manner in which they are allocated—matter early in the career. Salary influences whether federal employees quit early in the career, as do perceptions of the fairness and effectiveness of the performance appraisal system and the extent to which the federal government allocates rewards based on merit. Higher entry salaries may both attract more high-quality job applicants and help retain new hires. The quality of management appears to matter far more for new than for experienced employees.

Turnover is low among more experienced employees. Those in their 30s, 40s, and early 50s who have worked for the government for at least 8 years are remarkably unlikely to quit their jobs. In the mid and late career, the importance of extrinsic rewards drops and intrinsic motivation becomes increasingly important. In the CPDF, salary has a strong negative impact on turnover in the early career, but a weak-to-moderate positive impact in later years. Education is weakly related to turnover among new hires, but it has an increasingly negative impact as the career progresses. The singular importance of intrinsic motivation on turnover in the middle and late career in the MPS 2005 implies that better educated employees might be more committed to the federal service because they do more meaningful work and are more likely to believe in the missions of their agencies.

Our finding implies that government may want to focus on its retention efforts for employees who are newly hired and near retirement. Salary, performance appraisals, and fairness in the allocation of extrinsic rewards matter for newly hired workforce. However, those factors matter much less for retaining seasoned workers. The managerial challenge for these employees is to increase the meaningfulness of the work they perform. Enhancing empowerment and employee participation are among the plausible means to achieve that.

Footnotes

The authors presented an earlier version of this article at the Midwest Political Science Association (MPSA) annual meeting, Chicago, April 2010.

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The author(s) received no financial support for the research, authorship, and/or publication of this article.

1.

Baby Boomers have a bigger impact on the federal service than on the rest of the economy, because it expanded while they were entering the labor market and shrank when their turnover rates were lowest. The federal service grew by one fifth between 1965 and 1989 but shrank back to near-1965 levels by 2001 (U.S. Census Bureau, 2009) due to the Clinton Administration’s “reinventing government” efforts. The bulk of federal employees were in their 30s and 40s during the Clinton era, the ages at which few employees leave their jobs (Lewis & Cho, 2011). Hiring dropped dramatically, and between 1998 and 2007, the percentage of federal workers in their 20s dropped by nearly half (from 33% to 18%) whereas the percentage more than 55 rose by one quarter (from 31% to 38%; Lewis & Cho, 2011). With one third of federal employees in 2007 eligible to retire by 2012 (), expansion of the federal service under George W. Bush, and prospects of a more vigorous federal service under Barack Obama, however, hiring needs are increasing rapidly.

2.

Dropping blue-collar workers lowered the sample size by 12%, dropping part-time/temp/interim workers lost 9% of the remaining sample, and missing values cut another 1.4%. Accordingly, our sample is 79% of the full data set.

3.

About 7% of the people who drop out of the sample in 1 year reappear the next; only 1% of those who are gone for 2 years ever reappear, and sometimes that is many years later. Allowing a longer period for employees to reenter the service before identifying an exit would lose more years of current data.

4.

We also tried using 30 dummy variables for age to allow more arbitrary changes in turnover probabilities with age, but that model was not significantly better (at even the .20 level) than the simpler curvilinear model.

5.

The 2005 MPS asked respondents their “current educational level” but did not give an obvious response for those who started college but did not complete a degree. Based on the similarities of their mean salaries, we coded both those with associate’s degrees and those who chose “none of the above” as having some college.

6.

We used the same variable names in both data sets and ran the same models, except that we did not include the year dummy variables in the 2005 MPS analysis. We ran the logit analyses in Stata, used the predict command to generate the expected probability of leaving (or intending to leave) in that data set, then loaded the other data set and ran the the predict command again. This allowed us to generate expected probabilities for both dependent variables for both data sets.

7.

We ran the same models on both the CPDF and the 2005 MPS, then used the Stata predict command to generate both types of probabilities for both data sets.

References

Adams

J. S.

(1965). Inequity in social exchange. In Berkowitz

(Ed.), Advances in experimental social psychology, (Vol. 2, pp. 267–299). New York, NY: Academic Press.

Amabile

T. M.

(1993). Motivational synergy: Toward new conceptualizations of intrinsic and extrinsic motivation in the workplace. Human Resource Management Review, 3, 185-201.

Arthur

J. B.

(1994). Effects of human resource systems on manufacturing performance and turnover. Academy of Management Journal, 37, 670-687.

Asher

Popkin

(1984, October). The effect of gender and race differentials on public-private wage comparisons: A study of postal workers. Industrial and Labor Relations Review, 38, 16-25.

Becker

G. S.

(1962, October). Investment in human capital: A theoretical analysis. Journal of Political Economy, 70, 9-49.

Bernhardt

Dresser

(2002). Why privatizing government services would hurt women workers. (Publication No. B237). Washington, DC: Institute for Women’s Policy Research.

Bertelli

A. M.

(2007). Determinants of bureaucratic turnover intention: Evidence from the Department of the Treasury. Journal of Public Administration Research and Theory, 17(2), 235-258.

Colquitt

J. A.

Conlon

D.E.

Wesson

M.J.

Porter

K.Y.

(2001). Justice at the millennium: A meta-analytic review of 25 years of organizational justice research. Journal of Applied Psychology, 86, 425-445.

Dalton

D. R.

Johnson

J. L.

Daily

C. M.

(1999). On the use of “Intent to…” variables in organizational research: An empirical and cautionary assessment. Human Relations, 52(10), 1337-1350.

10.

Deci

E. L.

Ryan

R. M.

(2004). Handbook of self-determination research. Rochester, NY: University of Rochester Press.

11.

DelCampo

R. G.

(2006). The influence of culture strength on person-organization fit and turnover. International Journal of Management, 23, 465-469.

12.

Farber

H. S.

(1999). Mobility and stability: The dynamics of job change in labor markets. In Ashenfelter

Card

(Eds.), Handbook of labor economics (Vol. 3, pp. 2439-2483). Amsterdam, Netherlands: Elsevier Science B.V.

13.

Federal Times. (2009a, June 24). OPM: Stop using KSAs when hiring. Retrieved from www.federaltimes.com.

14.

Federal Times. (2009b, June 29). OPM: Pay reform by 2011. Retrieved from www.federaltimes.com.

15.

Fogel

Lewin

(1974, April). Wage determination in the public sector. Industrial and Labor Relations Review, 27, 410-431.

16.

Gould-Williams

(2004). The effects of “high commitment” HRM practices on employee attitude: The views of public sector workers. Public Administration, 82(1), 63-81.

17.

Gould-Williams

Davies

(2005). Using social exchange theory to predict the effects of HRM practice on employee outcomes. Public Management Review, 7(1), 1-24.

18.

Haines

III Jalette

Larose

(2010). The influence of human resource management practices on employee voluntary turnover rates in the Canadian non governmental sector. Industrial and Labor Relations Review, 63, 228-246.

19.

Hashimoto

(1981). Firm-specific human capital as a shared investment. American Economic Review, 71, 475-492.

20.

Hundley

(1991, Fall). Public- and private-sector occupational pay structures. Industrial Relations, 30, 417-434.

21.

Huselid

M. A.

(1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38, 635-672.

22.

Ippolito

R. A.

(1987). Why federal workers don’t quit. Journal of Human Resources, 22, 281-299.

23.

Keaveney

S. M.

Nelson

J. E.

(1993). Coping with organizational role stress: intrinsic motivational orientation, perceived role benefits, and psychological withdrawal. Journal of the Academy of Marketing Science, 21(2), 113-124.

24.

Kim

(2005). Factors affecting state government information technology employee turnover intentions. American Review of Public Administration, 35, 137-156.

25.

Konovsky

M. A.

Cropanzano

(1991). Perceived fairness of employee drug testing as a predictor of employee attitudes and job performance. Journal of Applied Psychology, 76, 698-707.

26.

Krueger

A. B.

(1988b, July). The determinants of queues for federal jobs. Industrial and Labor Relations Review, 41, 567-581.

27.

Lewis

G. B.

Cho

Y. J.

(2011). The aging of the state government workforce: Trends and implications. American Review of Public Administration, 41(1), 48-60.

28.

Lewis

G. B.

Park

(1989). Turnover rates in federal white-collar employment: Are women more likely to quit than men? American Review of Public Administration, 19(1), 13-28.

29.

Llorens

J. J.

Wenger

J. B.

Kellough

J. E.

(2008). Choosing public sector employment: The impact of wages on the representation of women and minorities in state bureaucracies. Journal of Public Administration Research and Theory, 18, 397-413.

30.

Macky

Boxall

(2007). The relationship between “high-performance work practices” and employee attitudes: An investigation of additive and interaction effects. International Journal of Human Resource Management, 18, 537-567.

31.

Meier

K. J.

Hicklin

(2008). Employee turnover and organizational performance: Testing a hypothesis from classical public administration. Journal of Public Administration Research and Theory, 18, 573-590.

32.

Moore

W. J.

Raisian

(1991, Winter). Government wage differentials revisited. Journal of Labor Research, 12, 13-33.

33.

Moynihan

D. P.

Landuyt

(2008). Explaining turnover intention in state government: Examining the roles of gender, life cycle, and loyalty. Review of Public Personnel Administration, 28(2), 120-143.

34.

Moynihan

D. P.

Pandey

S. K.

(2008). The ties that bind: Social networks, person-organization value fit, and turnover intention. Journal of Public Administration Research & Theory, 18, 205-227.

35.

W. Y.

(1962, December). Labor as a quasi-fixed factor. Journal of Political Economy, 70, 538-555.

36.

O’Reilly

C. A.

III Chatman

Caldwell

D. F.

(1991). People and organizational culture: A profile comparison approach to assessing person-organization fit. Academy of Management Journal, 34, 487-516.

37.

Parsons

D. O.

(1972). Specific human capital: An application to quit rates and layoff rates. Journal of Political Economy, 80, 1120-1143.

38.

Partnership for Public Service & Booz Allen Hamilton. (2010). Beneath the surface: Understanding attrition at your agency and why it matters. Washington, DC: Partnership for Public Service.

39.

Poon

J. M. L.

(2004). Effects of performance appraisal politics on job satisfaction and turnover intention. Personnel Review, 33, 322-334.

40.

Richer

S. F.

Blanchard

Vallerand

R. J.

(2002). A motivational model of work turnover. Journal of Applied Social Psychology, 32, 2089-2113.

41.

Smith

S. P.

(1976, January). Pay differences between federal government and private sector workers. Industrial and Labor Relations Review, 29, 176-197.

42.

Smith

S. P.

(1983). Are state and local government workers overpaid? In Hirsch

W. Z.

Rufolo

A. M.

(Eds.), The economics of municipal labor markets. Los Angeles, CA: Institute of Industrial Relations, UCLA.

43.

Stier

Lewin-Epstein

Braun

(2001). Welfare regimes, family supportive policies, and women’s employment along the life course. American Journal of Sociology, 106(6), 1731-1760.

44.

Tobias

R. M.

(2001). An aging workforce: A time of opportunity or a time of calamity? The Public Manager, 30(2), 27-30.

45.

Tsui

A. S.

Egan

T. D.

O’Reilly

C. A.

III . (1992). Being different : Relational demography and organizational attachment. Administrative Science Quarterly, 37(4), 549-579.

46.

U.S. Census Bureau. (2009). 2007 census of government employment. Retrieved from http://www.census.gov/govs/apes/index.html

47.

U.S. Government Accountability Office. (2009, February). Older workers: Enhanced communication among federal agencies could improve strategies for hiring and retaining experienced workers (GAO-09-206). Washington, DC: Author.

48.

U.S. Merit Systems Protection Board. (2008, January). Attracting the next generation: A look at federal entry-level new hires. Washington, DC: Author.

49.

Vroom

V. H.

(1964). Work and motivation. New York, NY: Wiley.