You are on page 1of 12

A Model of Voluntary Turnover

in State Government
SALLY COLEMAN SELDEN
AND DONALD P. MOYNIHAN
SYRACUSE UNIVERSITY

While many researchers have examined turnover in private sector firms, only a few studies have focused on quit rates
in the public sector organization This article develops and tests a conceptual model of voluntary turnover in state
government. The most significant finding is that state employees are less likely to quit in states with on-site childcare
State governments that allow greater internal opportunity for movement are also more successful at retaining employ-
ees. In addition, state employees who are represented by unions and are better paid are less likely to
.
quit

the fall of 1997, the turnover rate in the in the public sector, although the under-
In
U.S. rose to its highest level in over a de-
over

standing of voluntary turnover at the federal


cade (Kepner-Tregoe Business Issues Research level of government is undergirded by a few
Group, 1999), with the average turnover rate exceptional studies (Kellough & Osuna, 1995;
in &dquo;established&dquo; U.S companies running about Lewis, 1991; Lewis & Park, 1989; National
6 percent (Kepner-Tregoe Business Issues Re- Commission on the Public Service, 1990; U.S.
search Group, 1999, p. 2) and in state govern- General Accounting Office; 1990; U.S. Merit
ment about 8 percent (Government Perfor- Protections Board, 1989). Federal analyses
mance Project, 1998). Huge numbers ofwork- have examined the individual characteristics
ers are moving between jobs in the public, not- of those who leave (Lewis, 1991; Lewis & Park,
for-profit, and private sectors on a regular ba- 1989) and the organizational characteristics of
sis (Burgess, 1998). State governments increas- agencies losing employees (Kellough & Osuna,
ingly find themselves in competition for highly 1995).
talented workers (Ingraham, Selden & Drawing on both the existing literatures
Moynihan, 2000). Because of growing turn- in the private and public sectors, this study
over and competition for skilled labor wit- proposes and tests a model of voluntary turn-
nessed this decade, state government officials over in state government.’ Of particular im-

needs to better understand the factors associ- portance is the examination of the impact of
ated with turnover and what can be done to state human resource management practices on
lead people to stay. state quit rates. First, the article clarifies the
Scholars have devoted substantial atten- concept of turnover and examines the broad
tion to turnover in the private sector. Several themes that emerge from the literature. Sec-
scholars have undertaken meta-analyses of ond, we draw from the existing research and
employee turnover (Cohen, 1993; Cotton & develop a model predicting voluntary turnover
Tuttle, 1986; Hom, Caranikas-Walker, Prussia in state government. Third, we test the model
& Griffeth, 1992; Tett & Meyer, 1993). In using data collected from the 1998 Govern-
contrast, a paucity of research exists on turn- ment Performance Project (GPP), the National

63
Association of State Personnel Executives the employer terminates the employment con-
(NASPE), the Book of States, the Bureau of tract. The employee exerts the decision to leave

Labor Statistics, and the Census Bureau. Then the organization, a decision that presumably
we discuss the findings and implications of this the organization would wish to prevent in most
research. instances. To reduce the loss of talented em-
ployees, an employer must understand the rea-
TURNOVER IN THE PUBLIC SECTOR sons associated with voluntary turnover. This

allows an employer to make an informed choice


Defining Voluntary Turnover about what resources, strategies, and actions
Most studies of turnover rest on the notion could be used to prevent undesired separations
that lower turnover is indicative of a strong from occurring.
and more effective organization (Price, 1977,
pp. 92-119).~ Turnover is associated with high Different Strategies to Study Voluntary
organizational costs that are created when an Turnover in Government
employee departs-including the cost of re- A review of the literature reveals that scholars
cruiting and training a new employee, as well have taken several approaches to study volun-
as the impact on production as the new em- tary turnover in the public sector. One strat-
ployee climbs the learning curve (Cascio, 1982; egy is to look at total government employment
Kellough & Osuna, 1995). For example, re- and to compare voluntary turnover between
search by the Saratoga Institute (Kepner-Tregoe the public and private sectors (Utgoff, 1983).
Business Issues Research Group, 1999, p.3) A second approach is to examine voluntary
suggests that the replacement cost of an em- turnover within the federal government over a

ployee is between one and two times the specified period of time. In this genre of re-
individual’s salary. Kepner-Tregoe Business search, some studies focused on why individu-
Issues Research Group (1999, p.3) estimates als voluntarily leave the federal government,
that it costs approximately $134,000 to replace examining age, experience, education, salary,
a human resource manager in the automotive grade, and gender (Lewis, 1991; Lewis & Park,
industry. 1989), while another study considered deter-
Most research has been careful to dis- minants of quit rates across federal agencies
tinguish voluntary turnover from involuntary (Kellough & Osuna, 1995). Only a few stud-
and total turnover (see for example, Shaw, ies have looked at state government (Kim,

Delery, Jenkins & Gupta, 1998). Total turn- 1999; Smith, 1969). Smith, for example, ex-
over
typically includes voluntary turnover amined 1969 quits within a single civil service
(when employees quit); involuntary turnover system to determine if the opportunity struc-
(when employees are fired or laid off); and re- ture of jobs was related to quit rates.

tirements (when employees leave after meet- These studies have shown that volun-
ing specific service requirements for retire- tary turnover is a function of environmental,
ment). Shaw et al. (1998, p. 512) contend organizational, and individual factors
that voluntary turnover and involuntary turn- (Kellough & Osuna, 1995 ; Lewis,1991; Lewis
over are fundamentally distinct phenomena & Park, 1989; Smith 1969; Utgoff, 1983).
and are explained by different factors. With Environmental variables refer to factors exter-
voluntary turnover, the employee rather than nal to the organization, such as level of unem-

64
ployment and geographical region. Organiza- tant factor when an individual compares his
tional variables refer to structural characteris- or her current employer with prospective em-
tics of the organization (e.g., size, unioniza- ployers (Huselid, 1995; Shaw, Delery, Jenkins
tion, and occupational characteristics). Indi- & Gupta, 1998). Figure 1 shows the frame-
vidual variables are the personal characteris- work for this study.
tics of the individual employee that are hypoth-
esized as making a person more or less likely DETERMINANTS OF TURNOVER
to leave (e.g., gender, age or experience). This AT THESTATE LEVEL
research builds on these studies by examining
the impact of environmental factors (unem- State Unemployment
Environmental variables that impact voluntary
ployment and region) and organizational fac-
tors (size and unionization) on voluntary sepa- turnover decisions are primarily economic, and

ration in state government.3 Unlike previous relate to the level of unemployment and the
public sector research on turnover, this study subsequent demand and reward for labor.
also examines the relationship between selected When unemployment increases, voluntary
human resource management policies and turnover is expected to decline as workers will

be reluctant to leave a job in a weak labor mar-


practices (family-friendly policies, pay, and
ket (March & Simon, 1958). March and
training) and quit rates. Mobley (1977) sug-
Simon (1958, p. 100) wrote, &dquo;under nearly all
gests that human resource management prac-
tices matter because they influence an conditions the most accurate single predictor
&dquo;
of labor turnover is the state of the economy.&dquo;
employee’s assessment of his or her current situ-
ation, which may influence a person’s choice Many studies have confirmed this relationship
to seek other employment opportunities. In (Cotton & Tuttle, 1986; Park, Ofori-Dankwa
addition, an organization’s human resource & Bishop, 1994; Price, 1977; Zax, 1988).

management practices may also be an impor- Thus, this study proposes the following:

FIGURE 1 - MODEL OF VOLUNTARY TURNOVER


OF STATE GOVERNMENT EMPLOYEES

65
Hypothesis 1. States with higher rates of un- in the turnover research is organizational size
employment will have less voluntary turnover. (Kellough & Osuna, 1995; March & Simon,
1958; Morissette, Picot & Pyper, 1992; Smith,
Unionization
1979; Utgoff, 1983). Utgoff (1983) suggests
that larger firms will have lower turnover be-
The percent of the workforce that is union-
cause of employees’ enhanced opportunities for
ized is often purported to be an important pre-
internal advancement. Kellough and Osuna
dictor of turnover (Cotton & Tuttle, 1986;
Edwards, 1995; Groothuis, 1994; Huselid, (1995) find that larger federal agencies have
lower quit rates. However, Huselid’s (1995)
1995; Kellough & Osuna, 1995; Park, Ofori-
research shows the opposite; he found that
Dankwa & Bishop, 1994). If unions are suc- in
cessful in expressing worker preferences and
turnover was higher larger corporations.
Though dated, Price’s (1977) review of turn-
enhancing the workplace environment through over studies demonstrates inconsistent results
better working conditions and wages, workers
between size and turnover. Park, et al. (1994)
may be less likely to seek alternative employ-
ment. In addition, because the presence of suggest that the relationship between organi-
zational size and turnover is rather complex.
unions is often associated with an employment
system that supports seniority, employees have
They posit that size is very often a proxy for
additional incentives to remain in their present employee advancement opportunities where it
has a demonstrated effect.
job (Mitchell, 1989). While several studies Because previous research suggests that
support this relationship (Blau & Kahn, 1981;
size is a proxy for opportunity for advancement
Cotton & Tuttle, 1986; Huselid, 1995), a re-
or transfer (Kirschenbaum & Mano-Negrin,
cent study of federal agencies by Kellough and

Osuna (1995) suggests the contrary. Federal 1999; Morissette, Picot & Pyper, 1992; Park,
Ofori-Dankwa & Bishop, 1994; Smith,1979),
agencies with higher percentages of employees an alternative measure is the number of classi-
who were more unionized had higher quit rates
fication or job titles existing within an organi-
(Kellough & Osuna, 1995). However, as zation. Morissette, et al. (1992) point out that
Kellough and Osuna (1995, pp. 63-64) ob- an organization with more classification titles
serve, &dquo;federal government employee unions are

notoriously weak and are forced to operate provides greater internal movement opportu-
within the context of an extremely limited nities, which reduces turnover by increasing
the &dquo;protected&dquo; area for advancement or inter-
scope of bargaining.&dquo; The strength of unions
nal job alternatives. With respect to state gov-
in the bargaining process differs across states.
ernment, the number of classification or job
Unions have considerable bargaining power in
titles is positively and significantly correlated
some states, such as New York and Wisconsin,
with two commonly used measures of size-
and are almost irrelevant in other states, such
state employment and state population.4 How-
as Alabama and Georgia. Thus, we posit that
ever, unlike alternative measures of size, such
Hypothesis 2. States with more employees cov- as number of workers, it has the advantage of
ered by collective bargaining agreements will providing a more accurate portrayal of inter-
have less voluntary turnover. nal opportunities for employee advancement
and transfer..
Size and Internal Opportunity Structure
Another important variable typically included

66
Hypothesis3. The greater the number of classi- Mathews, 1999). Ezra and Deckman (1996)
fication titles in a state, the lower the level of found that balancing work and family was an
voluntary turnover. important determinant of job satisfaction,
which is often associated with turnover. Durst’s
(1999, p. 29) research showed that approxi-
Pay
Scholars purport that pay is an important de- mately 52 percent of local, state, and federal
terminant of turnover (Blau & Kahn, 1981; agencies surveyed agreed or strongly agreed that
Kim, 1999; Leonard; 1987; Lewis, 1991; Park family friendly programs decreased turnover.
Ofori-Dankwa & Bishop, 1994; Powell, States that employ more family-friendly poli-
,

cies are expected to be more successful at re-


Montgomery & Cosgrove, 1994; Shaw, Delery,
Jenkins & Gupta, 1998; Utgoff, 1983). A taining staff
number of studies confirm the importance of Hypothesis 5. States with more family friendly
pay in reducing quit rates (Blau & Kahn, policies will have lower voluntary turnover.
1981; Park, Ofori-Dankwa & Bishop, 1994;
Shaw, Delery, Jenkins & Gupta, 1998). Re- Training
cent research on federal employment is not as Finally, like Huselid (1995) and Shaw et al.,
clear, however (Lewis, 1991; Lewis & Park, (1998), this model includes training opportu-
1989). Despite the relative decline in federal nities.5 More training opportunities are likely
wages compared to private sector wages, Lewis to enhance an employee’s professional devel-
( 1991 ) found that voluntary turnover has been opment, making the individual more likely to
relatively stable. Nevertheless, we contend that stay (Huselid, 1995; Shaw, Delery, Jenkins &
state governments with higher pay rates are Gupta, 1998). Shaw et al. (1998) found that
likely to be more successful in retaining em- training was a significant predictor of invol-
ployees, controlling for the state’s median wage. untary turnover, but not of voluntary turnover.
firms offering more
Hypothesis 4. States with higher average sala- Contraryhad expectations,
to

ries will have less voluntary turnover. training significantly higher discharge rates
(Shaw, Delery, Jenkins & Gupta, 1998). Al-
though the coefficient for training was nega-
Family-Friendly Policies tive when predicting quit rates, it was not sta-
This study also considers how family-friendly tistically
significant (Shaw, Delery, Jenkins &
policies such as flextime (Muchinsky & Mor- Gupta, 1998).
row, 1980), on-site childcare, and job sharing
6. States that offer more compre-
influence quit rates. Providing a set of policies Hypothesis
hensive training will have lower quit rates.
that enhances worker flexibility and makes em-
ployment more feasible for workers with fami- Scholars generally agree that the vari-
lies is likely to increase retention of workers ables identified above affect voluntary turnover.
(Durst, 1999; Ezra & Deckman, 1996; They also report that performance and quit
Newman & Mathews, 1999). The use of poli- rates will have a u-shaped relationship, with
cies such as on-site childcare, flextime, and job high performers more likely to leave for better
sharing assist employees, particularly women, employment opportunities, average perform-
face the dual demands of work and family life ers more likely to stay in the organization, and
(Ezra & Deckman, 1996; Newman & poor performers more likely to leave due to

67
actual or perceived threat of firing (Jackofsky, the Bureau of Labor Statistics, and the
1984; Park, Ofori-Dankwa & Bishop, 1994; Bureauof the Census. We had missing infor-
Williams & Livingstone, 1994). Because our mation for six states, therefore, the analysis is
research focuses on state governments as em- limited to 44 states.6 Three states, Idaho,
ployers, we cannot distinguish whether indi- North Dakota, and Georgia, provided 1996
viduals who left state government were high voluntary turnover statistics rather than 1997
or low performers. data. We contacted officials in each of these
states and learned that voluntary turnover is
DATA AND METHODS relatively consistent across years. As result, we
The model and the several hypotheses are included these three states in our final analy-
tested using data from the following sources: sis. Table 1 provides voluntary turnover rates
the 1998 Government Performance Project, for state governments in 1997.
the National Association of State Personnel We test the model with ordinary least
Executives (NASPE, 1999), the Book of States, squares (OLS) regression, estimating the effects

TABLE 1. 1997 VOLUNTARY TURNOVER RATE FOR STATE GOVERNMENT

68
of each independent variable on voluntary sion equation predicting voluntary turnover in
turnover while holding all other independent state governments. The independent variables

variables constant.7 Table 2 describes how the together explain 54.6 percent of variation in
dependent and independent variables are state quit rates. The equation achieves statis-

operationalized, and reports means and stan- tical significance at the .001 level. The find-
dard deviations for each. We include regional ings of the analysis are largely as expected, al-
controls in our model. We also control for though with an intriguing exception. We find
variation in state income by including the state a positive and statistically significant relation-

median income level for 1997. Due to the ship between unemployment and voluntary
fact that virtually all states employ flextime and turnover. The analysis suggests that states with

job sharing, thereby creating very little varia- higher unemployment rates have higher quit
tion (see Table 2), these variables are dropped rates. This relationship proved robust in the

from the proposed index of family-friendly analysis despite a number of efforts to test its
policies. The remaining family-friendly reliability.8 This finding defies previous stud-
policy-on-site childcare-is retained as a ies on turnover and the apparently
single variable. commonsense assertion that employees are

more cautious to quit their jobs in times of

FINDINGS AND DISCUSSION high unemployment (Cotton & Tuttle, 1986;


Table 3 reports the results of the OLS regres- Park, Ofori-Dankwa & Bishop, 1994; Price,

TABLE 2. MEANS AND STANDARD DEVIATIONS OF DEPENDENT AND INDEPENDENT VARIABLES

69
1977; Zax, 1998). hold for the public sector. All of the previous
One possible explanation of this find- studies confirming this relationship looked at
ing is the cross-sectional nature of our data. private sector firms. Moreover, none of the
As we noted the common sense assertion is that research on turnover in the public sector has
people are reluctant to quit jobs when there is tested this relationship (Kellough & Osuna,
high unemployment. For the period identi- 1995; Lewis, 1991; Lewis & Park, 1989).
fied, however, unemployment varied (with a Public employment systems are associated with
maximum of 7.9% in Alaska, followed by relatively rigid pay systems and high union
6.4% in New York and a minimum of 2.5% presence, offering stable employment and com-
in North Dakota) but overall this was a time pensation with little reference to the supply
of lowered unemployment throughout the and demand of labor (Ingraham, Selden &
United States, with national unemployment Moynihan, 2000).
rate of 4.9%. Given the low national unem- We find support for our hypothesis that
ployment, it may not be plausible to expect the presence of unions, represented by the per-
people to be cautious about leaving their job centage of employees covered by collective bar-
since they may perceive that alternative em- gaining agreements, improves the retention of
ployment is easily securable. employees. Partly this is because unions allow
Another possible explanation is that the employees to gain higher wages-unionization
negative relationship between unemployment is positively correlated with average pay-but
and turnover found in the literature does not increased unionization matters above and be-

TABLE 3. PREDICTING 1997 VOLUNTARY TURNOVER IN STATE GOVERNMENT

Re=.546 ; Adjusted l~=.408; Fvalue =3.965&dquo;&dquo;


*Significant at.10; &dquo;Significant at .05, &dquo;&dquo;Sigmficant at .001; ’Significant at the .004 level but not in the predicted direction.

70
yond average salary. Our findings differ from to quit rates, it is not statistically
significant.
Kellough and Osuna (1995) research that The most plausible explanation is that a clear
showed a significant and positive relationship unidirectional relationship does not exist. On
between unionization and quit rates in federal one hand, states with few
training opportuni-
agencies, but they are consistent with studies ties may be unable to sustain employee satis-
of private firms (see e.g. Blau & Kahn, 1981; faction with regard to professional develop-
Edwards, 1995). However, federal unions are ment, leading to an increased rate of quits.
often weaker than many unions representing Alternatively, states that spend a great deal of
state employees. As mentioned earlier, the role time enhancing the skills of their employees
of unions varies across states. For example, may see that investment lost as many of these
right-to-work states do not provide for collec- employees sell these more marketable skills else-
tive bargaining and may find it easier to nego- where.
tiate certain workplace policies than highly The final variable to show a significant
unionized state governments. However, in relationship to voluntary turnover is on-site
states with a high union presence, the analysis childcare. The coefficient is negative, as ex-
suggests employees are more likely to have a pected, statistically significant at the .05 level.
stronger voice and may garner higher wages The presence of on-site childcare is associated
and other benefits because they are formally with a 2.84 percentage point decrease in vol-
involved in determining these policies.’ untary turnover. States with on-site childcare
Classification titles are negatively related are better able to retain employees. This find-

to turnover, as predicted. Like previous stud-


ing is consistent with Durst’s (1999) work.
ies using size as a proxy of opportunity, we find Further research in this area would benefit from
that states with more advancement and trans- analyzing the nature of the relationship be-
fer possibilities have fewer quits (see Kellough tween childcare options and the rate of turn-

& Osuna, 1995). We also found support for over by gender, and indeed, recruitment levels

two of the three hypotheses relating to HRM by gender (Ezra & Deckman, 1996).
practices. States that pay higher average wages
experience fewer quits. The coefficient on this CONCLUSION
variable is negative and significant at the .05 This study extends our understanding of envi-
level. This finding is consistent with previous ronmental, organizational, and human re-
research on turnover patterns in the private sec- source management factors associated with

tor (Blau & Kahn, 1981; Park, Ofori-Dankwa voluntary turnover in state government. The
& Bishop, 1994; Shaw, Delery, Jenkins & variable unemployment rate is particularly
Gupta, 1998) and in the state of California important. However, understanding the un-
(Kim, 1999). State government employees are expected relationship between unemployment
less likely to quit their jobs when they are well and voluntary turnover in state government
paid, regardless of how their pay compares in demands more research. A careful study of
wages and salaries offered by other employers voluntary quits and the demand and supply of
in the state. This result suggests that prob- labor in the public sector has not been under-
lems of voluntary turnover will be more diffi- taken. The counterintuitive nature of these
cult in states that do pay well.
not results suggests that, as yet, the field has a weak
Although training is positively related understanding of how public employment sys-

71
tems reflect taken-for-granted assumptions formance&dquo; human resource management prac-
about how the labor market works. tices such as performance-related-pay (see for
These state government results provide example, Huselid, 1994) this research suggests
practical direction for state governments with that providing for basic needs of employees to
respect to opportunity structure. In recent juggle the demands of family and career has a
years, many states have focused on reducing large and positive influence on retention. In
the number of classification and job titles in terms of practice, the impact in providing on-

an effort to produce a simpler, streamlined site childcare on turnover provides a clear


employment structure. A related trend has policy suggestion to employers. For public
been broadbanding of titles-collapsing a sector organizations that are experiencing prob-

number of titles into larger bands that offer lems in retaining employees, offering on-site
managers discretion in allocating pay. Since childcare may provide a manageable and ef-
this research suggests that classification titles fective solution.
may have a salutary effect from the employee’s While this study adds to our under-
perspective, states should contemplate carefully standing of voluntary turnover in the public
the potential effect of reducing titles on em- sector, additional research is needed to explore

ployees’ perceptions of advancement opportu- other forces influencing state government quits.
nity. At a minimum, more research is needed to
Further, as state governments struggle understand the way in which public employ-
with turnover, pay will remain an essential tool ment systems interact with the labor market.

for retaining talent. Interviews with state offi- Future studies should explore the dynamics and
cials suggest that turnover problems are often patterns of voluntary turnover in government
concentrated in particular employee areas, such over time and from the perspective of the in-

as Information Technology (Barrett & Greene, dividual employee. The possibility of realiz-
1999). If states determine that turnover prob- ing the impact of human resource practices-
lems are localized to particular areas, they may factors that can be influenced by public man-
target pay increases or bonuses specifically to agers-provides considerable incentive for such
this group of employees. Some states, such as efforts.
Virginia, are considering such policies.
Perhaps the most important finding of AUTHOR’S NOTE
*We would like to thank J Edward Kellough and the re-
this study, in terms of contribution to research
mewers for their comments and suggestions
and practice, is the impact of the on-site
childcare variable. This research indicates that
NOTES
providing on-site childcare has a substantial not present an exhaustive review of re-
’This article does
impact on the percentage of persons leaving search private firms. Rather, it draws on studies
on turnover in

Previous that examine turnover at an organizational or macro level, rather


state employment voluntarily. turn-
than at an individual or micro perspective (such as Dickter,
over research has done little to examine the
Roznowski & Harrison, 1996, Tang, Kim & Tang, 2000; Tett
impact of family-friendly policies, and on-site & Meyer, 1993; Vandenberg & Nelson, 1999).
childcare represents a new factor to be included More recent research has distinguished between
2 turnover

in the future study of voluntary turnover in where poor performers leave(which is beneficial to the organi-
both private and public organizations. While zation) and where high performers quit (Park, Ofon-Dankwa
& Bishop, 1994; Williams & Livingstone, 1994)
many other studies have considered &dquo;high-per- Since our
3 unit of analysis is state government, we do not

72
model individual factors These factors have been examined in .05 level) and positively related to voluntary turnover (with a

a number ofstudies, including Blau and Kahn (1981); Buchko .291 relationship, significant at .055).
(1992); Cotton and Tuttle (1986); Lewis (1991); Lewis and
Park (1989), Muchinsky and Morrow (1980), Smith (1979),
REFERENCES
and Zax (1988).
Barrett, K. & Greene, R (1999, February). Grading the states:
The correlation between number of classification is .548
4
A management report card. Governing, 12,
17-90.
and 4.55 for total state employee and state population, respec-
Blau, F.D. & Kahn, L.M. (1981). Race and sex differences in
tively. Both are significant at the.001level. quits by young workers. Industrial and Labor Relations
We use an index assessing the quality of state government
5
Review
, 563-577.
34
training opportunities taken from the 1998 Government Per-
formance Project.
Burgess, S (1998). Analyzing firms, jobs, and turnover
The analysis excludes Arizona, California, Colorado, Con-
6
Monthly Labor Review, 121
, 55-58.
Cascio, WF (1982). Costing human resources: The financial
necticut, West Virginia and Wyoming
impact of behavior in organizations
. Boston: Kent
To determine whether ordinary least squares was the ap-
7
Cotton, J.L & Tuttle, J M (1986). Employee turnover: A
propnate estimator, we examined our data for heteroscedasticity,
meta-analysis and review with implications for research.
multicollinearity, and influential data (Fox, 1991).An analysis Academy of Management Review 11 , 55-70
of the residuals suggests that the errors follow a largely random
Cohen, A. (1993). Organizational commitment and turn-
pattern (Fox, 1991). We examined the bivariate correlations over: A meta-analysis Academy of Management Journal
and the square root of the variance inflation factor (VIF) to
36,
1140-1157.
detect multicollinearity (Fox, 1991, p. 11). None of the square
Dickter, D.N , Roznowski, M. & Harrison, D.A. (1996). Tem-
roots of the VIFs exceeded 1.7. To detect potential outhers, we
poral tempering An event history analysis of voluntary
examined studentized residuals To assess the importance of
turnover ,Journal of Applied Psychology 81 705-716
outliers, we reran model excluding states with studentized re-
Durst, S. (1999). The effect of family-friendly programs on
siduals of +/-2 (Arizona, Kentucky and New Mexico). We found
that the exclusion of the three states slightly improved the
public organizations Review of Public Personnel Adminis-
tration 19, 19-33
model’s overall explanatory power. However, the direction and
Edwards, PK (1995) Human resource management, union
significance of the relationships between the independent vari- voice, and use of discipline. An analysis of WIRS3 Indus-
ables and the dependent variable were consistent with our origi-
trial Relations
26, Journal 204-221
nal model. Thus, we did not exclude Arizona, Kentucky and
Ezra, M. & Deckman, M (1996). Balancing work and family
New Mexico from our final analysis.
Because the relationship between unemployment and turn-
8
responsibilities Flextime and childcare in the federal gov-
ernment Public Administration Review174-76 56,
over was not as expected, we explored a number of different

models to determine its robustness We examined the relation-


. Newbury Park: Sage
Fox, J (1991). Regression diagnostics
Publications.
ship between turnover and unemployment over a three-year Government Performance Project (1998). State government
period 1995, 1996 and 1997 In all three years the relation- survey of human resource management. Syracuse, NY: Alan
ship remained positive and significant, in both bivariate and K Campbell Institute of Public Affairs, Syracuse Univer-
multivariate analyses Then, we included lagged measures (up
sity.
to five years) of unemployment in the model estimating 1997
Groothuis, P A. (1994) Turnover: The implication of estab-
voluntary turnover Next, we estimated a number of alterna- lishment size and unionization. Quarterly Journal of Busi-
tive models using other factors that might be linked to turn-
ness and ,
Economics 33 41-53
over and/or unemployment. In addition, we identified outly-
Hom, P.W., Caranikas-Walker, F., Prussia, G.E.,& Griffeth,
ing states which exhibited high unemployment (over 5%) and RW (1992) A meta-analytical structural equations analy-
high turnover (over 10%) and included these as a dummy vari- sis of a model of employee turnover Journal of Applied
able in the
regression model The states were Alaska, Idaho,
Psychology 77, 890-910.
Louisiana, Mississippi, New Mexico, Tennessee and Texas The
Huselid, M.A (1995). The impact of human resources man-
inclusion of this dummy variable did not impact the result
agement practices on turnover, productivity, and corpo-
Finally, we included the share of the labor force comprised of rate financial performance Academy of Management Jour-
state employees as a control and the Gross State Product growth
nal
, 635-672
38
Neither of these factors mitigated the original relationship be-
tween unemployment and turnover. We found the relation-
Ingraham, PW, Selden, S.C. & Moynihan, D P. (2000)
People and performance: Challenges for the future public
ship between unemployment and quit rates extremely robust service-the report from the Wye River conference. Public
ight-to-work states are negatively correlated to average
R9 Administration ,
Review 60 52-58.
state employee pay (with a -.356 correlation, significant at the Jackofsky, E.F. (1984). Turnover and job performance. An

73
integrated process model. Academy of Management Re- Park, H.Y, Ofon-Dankwa, J. & Bishop, D.R. (1994) Orga-
9,
view 74-83. nizational and environmental determinants of functional
Kellough, J.E. & Osuna, W (1995). Cross-agency compari- and dysfunctional turnover: Practical and research impli-
sons of quit rates in the federal service: Another look at cations. Human Relations 47, 353-66.

the evidence. Review of Public Personnel Administration Powell, I., Montgomery, M. & Cosgrove, J. (1994). Com-
15,
58-68. pensation structure and establishment quit and fire rates.

Kepner-Tregoe Business Issues Research Group (1999). Avoid- Industrial Relations 33,
229-248.
ing the brain drain: What companies are doing to lock in Price J.L (1977). . The study of turnover Ames, IA. The Iowa
their talent
. Princeton, NJ: Kepner-Tregoe. State University Press.
Kim, M. (1999). Where the grass is greener: Voluntary turn- Shaw, J.D , Delery, J., Jenkins Jr G.D., & Gupta, N. (1998).
over and wage premiums. Industrial Relations 38, 584- An organization-level analysis of voluntary and involun-
592. tary turnover. Academy of Management Journal 41, 511-
Kirschenbaum, Alan & Mano-Negrin, R. (1999). Underlying 525.
labor market dimensions of "opportunities": The case of Smith, C.B. (1979). Influence of internal opportunity struc-
employee turnover. Human Relations 52
, 1233-1244. ture and sex of worker on turnover patterns. Administra-
Leonard, J.S. (1987). Carrots and sticks: Pay, supervision, tive Science Quarterly 24, 362-379
and turnover. Journal of Labor Economics 5, 136-154 Tang, T.L., Kim, J.K., & Tang, D.S. (2000). Does attitude
Lewis, G.B. (1991). Turnover and the quiet crisis in the fed- toward money moderate the relationship between intrin-
eral civil service. Public Administration Review 51, 145- sic job satisfaction and voluntary turnover? Human Rela-

155. tions , 213-235.


53
Lewis, G.B. & Park, K. (1989). Turnover rates in federal Tett, R.P. & Meyer, J.P (1993) Job satisfaction, organiza-
white-collar employment: Are women more likely to quit tional commitment, turnover intention, and turnover. Path
than men? American Review of Public Administration 19, analyses based on meta-analytic findings. Personnel Psy-
13-28 chology 46, 259-293.
March, J.G & Simon, H.A (1958). Organizations
. New U S General Accounting Office (1990). Recruitment and
York: Wiley. retention. Inadequate federal pay cited as primary problem

Mitchell, J.B. (1982). Human resource management: An eco- by agency officials Washington, D.C.: U SGeneral Ac-
nomic approach
. Boston: PWS-Publishing counting Office
Mobley, WH. (1977). Intermediate linkages in the relation- U.S Merit Systems Protections Board. (1989). Who is leaving
ship between job satisfaction and employee turnover. Jour- the federal government? Washington, D.C: Merit Systems
nal of Applied Psychology 62, 237-240. Protection Board
Morissette, R., Picot, G. & Pyper, W (1992). Workers on the Utgoff, K.C. (1983) Compensation levels and quit rates in
move: Quits. Perspectives on Labor and Income 4, 8-17. the public sector. ,
The Journal394-
of Human Resources 28
Muchinsky, P.M. & Morrow, P.C. (1980). A multidisciplinary 406
model of voluntary employee turnover Journal of Voca- Vandenberg, R. J. & Nelson, J.B. (1999). Disaggregating the
uonal Behavior 17, 263-290. motives underlying turnover intention: When do inten-

Newman, M. & Matthews, K. (1999). Family-friendly work- tions predict behavior? Human Relations 52, 1313-1323.

barriers to effective implementation. Re-


place policies: Williams, C.R. & Livingstone, L P. (1994). Another look at
of Public Personnel Administration 34-48.
view 19, the relationship between performance and voluntary turn-
National Commission on the Public Service. (1990). Leader- Academy of Management Journal 37 269-298.
over. ,

ship for America: Rebuilding .


the public service Lexington, Zax, Jeffery S. (1988). Quits and race. The Journal of Human
MA.: Lexington Books Resources 27
, 469-493.

74

You might also like