Human Resources Management

Practices and Turnover Intentions of

University Employees in Uganda

A PhD Defense Presentation

By
Godfrey Tumwesigye

Supervisors
Prof JCS Musaazi Prof. Joseph Oonyu
1
 Outline of the Presentation
1. Background and Statement of the Problem
2. Research Questions and Hypotheses
3. Conceptual framework
4. Methodology
5. Findings
6. Implications and Future Research

2
 Introduction
• HRM practices :a set of internally consistent
practices designed to attract, motivate and retaion
employees who can make optimum contribution to
the achievement of its goals (Delery & Doty,
1996).
• RBV: Human resources are the main sources of
competitive advantage (Guest 2011)

3
 Introduction, Cont’d
• Application of favourable HRM practices is widely
believed to:
– Affect employee attitudes and behaviours and
– Lead to sustainable competitive advantage (Arthur, 1994,
Huselid, 1995).
• Ugandan universities are losing their senior
employees through voluntary turnover (Mamdani,
2007; Tettey, 2009; Kasozi, 2007; OAG, 2015;
2016).
• In some universities up to 50% of the established
positions are vacant (OAG, 2015; 2016)

www.Quant.KU.edu 4
 Introduction
• HRM practices may be the necessary intervention which may
control voluntary turnover of competent employees (Metcalf
2005; Daly & Dee, 2006; Pare & Trembley, 2007).
• In spite of the high voluntary turnover in universities, little
attention has been given to examining the antecedents that
compel university employees to leave.
• Most research on HRM practices has been in USA, Europe
and Asia, leaving Sub-Saharan Africa under-researched.
• We cannot assume that relationship found in western
countries are generalizable to other contexts (Bahwar &
Debrah, 2008; Takeuchi et al 2009, etc)

5
 Introduction, cont’d
• Moreover, the few studies in sub-Saharan
Africa have not studied the mediation
mechanisms through which HRM practices
may influence turnover intentions.

• Studies in other countries (e.g. Europe, USA,

Asia) show that Job satisfaction, organizational
commitment and OCB mediate between
HRMP and TOI

www.Quant.KU.edu 6
 Statement of the Problem
• Despite large volume of research on relationship between
HRM practices and turnover/turnover intentions, there are
theoretical gaps in researchers’ understanding of this
relationship, esp. in sub-Saharan African countries
(universities).
• We do not fully understand the mechanisms through which
HRM influences turnover because even in the developed
countries, most research focuses on:
– (1) direct effects of HRM practices on JS, OC, OCB and TOI and
– (2) direct of JS, OC and OCB on turnover and TOI

7
 Problem Statement, cont’d
• Sufficient simultaneous research has not been undertaken,
especially in sub-Saharan Africa.
• Much as Ugandan universities implement HRM practices,
it is not clearly understood, how these practices influence
turnover intentions.
• The mediating effects of employee attitudes (job
satisfaction and organizational commitment) and
behaviours (OCB) is unresearched.
• If these relationships are not examined, universities could
continue losing their most valuable assets (HR) through
turnover.
8
 Purpose of the Study
• To determine whether, in the Ugandan context, HRM
practices predict turnover intentions of University
employees; and
• To assess the mediating effects of job satisfaction,
organisational commitment and OCB on the relationship
between HRM practices and Turnover intentions.

9
 Objectives of the study
1. To examine the relationship between HRM practices and turnover
intentions of university employees.
2. To examine the relationship between HRM practices and job
satisfaction among university employees.
3. To determine the relationship between HRM practices and
organisational commitment among university employees.
4. To examine the relationship between HRM practices and
organisational citizenship behavior among university employees.
5. To establish whether (a) job satisfaction, (b) organizational
commitment and (c) OCB are significant predictors of turnover
intentions
6. To determine the extent to which (a) job sat, (b) org commitment
and (c) OCB mediate the HRM-TOI relationship.
10
 Theoretical Basis of the Study

• Social exchange Theory (Blau,

1964; Gouldner, 1960)
• Theory of Reasoned Action (Ajzen
& Fishbien, 1975)
• Black Box Theory (Pauuwe, 2004;
Takeuchi, 2003; Ramsay, et al.
2000)
11
 Conceptual framework
2.1.

Independent Variable Mediating Variables Dependent Variable:

HRM Practices Job Satisfaction

1. Recruitment and Selection
2.Jojjj
Training and Development
3. Rewards and Recognition Organisational Citizenship
4. Performance Appraisal Behaviours
5. Promotion Opportunities Turnover
6. Job security
Intentions
7. Information Sharing
8. Employee Participation
Organizational Commitment

Figure 2.1: The Conceptual Framework

12
 Conceptual Framework

OC

H5
H3
JS
H2 H6
HRP
H1
TOI
H4

H7
OCB

H8: HRP->JS->TOI
H9: HRP->OC->TOI
H10: HRP->OCB->TOI

13
 Methodology
1. Research design: cross-sectional survey.
2. Six universities (3 public and 3 private) out of
36.
3. Mixed methods of data collection:
a. quantitative data using a questionnaire
b. qualitative data using interview of key
informants (2 per university)

crmda.KU.edu 14
 Methodology, cont’d
Measurement of variables:
4. All variables were measured on a seven point scale (1 = SD
… 7 = SA) using previously validated instruments.
a. HRM practices – 43 items (Delery and Doty, 1996; Pare
&Tremblay, etc.
b. Job satisfaction - 3 items (Camman et al 1983; Takeuchi et
al 2009).
c. Org. Commitment: 17 Items measuring three dimensions
(Meyer and Allen, 1997)
d. OCB- 16 items (Williams & Anderson, 1991; De Lara,
2008).
e. TOI – 4 items (three by Mobley et al, 1978; one self
developed

15
 Methodology, cont’d
5. Main Data source: employees; because what
determines employee attitude is how employees
perceive HRM practices (Gould-Williams &
Mohammed, 2010).
6. Data Analysis:
a. Reliability and validity: Cronbach alpha (SPSS, > 0.70)
(Cronbach, 1959); AVE (> 0.5} & Composite Construct Reliability
(CCR/CR, (>0.70))
c. Normality : Skewness and Kurtosis

16
 Methodology, cont’d
d. Exploratory factor analysis using principal component
analysis (SPSS v. 19.0)
e. Structural equation modelling (SEM) using AMOS v.
20.0
–Confirmatory factor analysis
–Path analysis
8. Why SEM?
• It allows measurement of multiple variables and multiple
relationships simultaneously.
• Includes both observed and latent variables and represented in a
model.
• Superior measures of reliability (AVE, CCR)
• Good for model building and theory testing
• Can Test for mediation effects
17
 Methodology, cont’d
9. Estimation method: Maximum Likelihood
Estimation (MLE)
– MLE is flexible and robust against violations of
multivariate normality (Hair, et al, 2010; Kline, 2005).
10. Why AMOS software?
» Readily available at cost of UShs 200,000/=
» The researcher found it easy to learn and use.
» Graphic features allow non-statisticians to build models and
interpret text output

18
 Factor Analysis
1. The purpose of factor analysis is to reduce the
number of variables in the study. (Gerbing and
Hamilton, 1996; Kline, 2005).

2. With 83 observed variables it would be difficult to

assess relationships and test the hypotheses.

3. FA combines related observed variables into one

latent variable.

19
 Exploratory factor analysis (EFA)
• EFA was conducted using SPSS V. 19
• Extraction method: Principal components; based on Eigen
values greater that 1.0.
• Rotation method: Varimax with Kaiser normalization.
• Variables with factor loadings greater than 4.0 were
retained.
• Cross loaded factors with loading difference of at least 0.2
were retained (Hair et al 1998)

20
 A Note on EFA of OCB

• Two dimensions had been expected for OCB

(OCBI and OCBO) Anderson and Williams
(1991).
• However EFA extracted three dimensions:
they were labeled:
– OCB Co-workers, equivalent to OCBI
– OCBO
– OCB Students (Somech & Bolger, 2002; Somech
& Ron 2006)

21
 Summary of EFA

Summary of the Results of Exploratory Factor Analysis

Scale No. of Items No of % of variance
extracted explained
dimensions
Human Resource Management Practices 33 8 68.157
Job satisfaction 3 1 89.929
Organisational commitments 13 3 74.429
Organisational citizenship behaviour 13 3 63.521
Turnover intentions 4 1 97.966
Total 66 17

Note:
1. Total number of observed variables was reduced from 83 to 66;
2. As proposed, there are 17 latent variables.

22
 Confirmatory Factor Analysis
• Output of EFA were subjected to CFA using
AMOS.
• CFA is a more rigorous technique compared to
EFA (Hair, et al 1998; Tabachnick & Fidell,
1996; Kline, 2005).
• Process:
1. Congeneric Models
2. First order Factor analysis
3. Second order factors analysis
4. Overall measurement model
www.Quant.KU.edu 23
 Congeneric one-factor models
• Purpose: To determine unidimensionality of a
latent construct (Anderson and Gerbing, 1988;
Hair et al, 1998).
• To confirm latent constructs extracted by EFA.
• All 17 latent variables were subjected to one-
factor congeneric modelling and found to have
good fit.

www.Quant.KU.edu 24
 Model fit Criteria
• Chi-square (χ2) and associated degrees of
freedom (df): p> 0.05. i.e. not significant.
– However, this fit index is sensitive to model
complexity, sample size, non-normality.
– Will be reported but not relied on for assessing
model fit.

www.Quant.KU.edu 25
 Model Fit Criteria
• Normed Chi-square (χ2/df): should be < 5.0; less that 3.0 is
preferable.
• Root Mean square Residual (RMR): <0.05 shows good
fit, <0.08 shows acceptable fit.
– However this statistic is sensitive to sample size and model
complexity.
• Root mean square error of approximation (RMSEA):
<0.05 shows good fit, <0.08 shows acceptable fit.

• Standardized root mean square residual (SRMR): <0.05

shows good fit, <0.08 shows acceptable fit.

26
 Model Fit Criteria
• Goodness of fit index(GFI): Values greater that 0.90
indicate good fit.
• However this statistic is sensitive to sample size
and model complexity.
• Adjusted Goodness of fit index(AGFI): Values
greater that 0.90 indicate good fit.
• However this statistic is sensitive to sample size
and model complexity.
• Tucker-Lewis Index (TLI): Values > 0.90 indicate
good fit.
• Comparative Fit Index (CFI): Values > 0.90
indicate good fit.
www.Quant.KU.edu 27
 Model Fit Criteria
• Given the sample size of this study (N=722) and model
complexity (17 latent constructs), fit indices that are
sensitive to sample size (χ2, RMR, GFI and AGF)
will not be relied on. However, χ2 will be reported.

• The study relies on χ2/df, TLI, CFI, RMSEA, &

SRMR to determine model fit.

• Previous researchers recommend that at least three fit

indices should be reported (Byrne, 2010; Hair et al
2006) .

www.Quant.KU.edu 28
 Congeneric Model of Training

MFI: χ2 (4 N=722) = 7.68; χ2 /df = 1.920; SRMR =0.0106;

RMSEA = 0.054;TLI = .994; CFI = 0.998;).

29
 Congeneric Model of Recruitment

(χ2 (5 N=722) = 11.296, p= 0.046, χ2/df =2.259;

SRMR = .0126; TLI=0.993; CFI=0.997; RMSEA =.041).

30
 Other Congeneric Models
• Information sharing (4 items): χ2 (1, N=722) = 2.656, p=
0.103, χ2/df =2.656; SRMR = .0068; TLI=0.993;
CFI=0.999; RMSEA =.048). Factor loadings ranged low of
0.65 to a high of 0.86
• Employment security (4 items): χ2(1 N=722) = 1.346, p=
0.246, χ2/df = 1.346, TLI = 0.999, CFI =1.000; SRMR =
0.0047; RMSEA = 0.022; All items loaded highly on this
factor, the lowest factor loading being 0.76 and the highest
being 0.93.
• Etc. up to 17 latent constructs as identified by EFA. Note:
Some items were deleted during assessment of congeneric
models.
31
First order CFA for HRM
Practices

32
 Assessment of convergent validity (HRM)
CR AVE IS RS TD RW PO ES PD PA
IS 0.878 0.643 0.802
RS 0.888 0.615 0.350 0.785
TD 0.871 0.575 0.482 0.403 0.758
RW 0.805 0.509 0.518 0.405 0.608 0.713
PO 0.828 0.547 0.468 0.458 0.529 0.499 0.740
ES 0.891 0.673 0.249 0.159 0.200 0.197 0.417 0.820
PD 0.813 0.605 0.546 0.319 0.418 0.348 0.420 0.266 0.778
PA 0.763 0.525 0.449 0.406 0.559 0.521 0.564 0.238 0.370 0.725

1. All CR values are greater than 0.70, thus good reliability (Byrne, 2010)
2. All AVE are greater that 0.50 (Hair et al, 2006).
3. All diagonal values (square roots of AVE) are greater than inter-construct
correlations (Kline, 2005)
4. Thus no reliability/validity issues.
5. However it is noted that the correlations of WL with other constructs are
weak.
33
Modified 2nd order CFA for
HRMP

www.Quant.KU.edu 34
 Model Fit Indices for 2nd Order HRM models
Model χ2 df χ2 /df p-value SRMR RMSEA TLI CFI
Re-specified Second 1480.19 487 3.039 0.000 0.0519 0.053 0.915 0.921
Order model of HRM
practices

35
 CFA for OC and OCB
• Through the same process as described for
HRM practices, congeneric, first order and
second order CFA models were assessed and
validity established for:
• Organisational commitment
• OCB
There were no reliability or validity issues (All
CR greater than 0.70 and all AVE greater that
0.50)

www.Quant.KU.edu 36
 1st Order Model for Org. Commitment

χ2 (df=62, N=722) = 236.131; χ2/df = 3.809; p=0.000; RMSEA

=0.062; SRMR = 0.045; TLI = .967; CFI=0.973. All items loaded
highly on their respective factors with loadings ranging from 0.60 to
0.90.

www.Quant.KU.edu 37
 Model Fit indices for overall
measurement model
Model Fit index Value
χ2 (df,) 4097.793 (1832) p = 0.000
Χ2/df 2.237.
RMSEA 0.041
SRMR
TLI 0.920
CFI 0.927

All the inter-correlations were < 0.85 (Kline 2005).

All CR >0.70 (Anderson and Gerbing, 1988; Hair et al, 1998)
All AVE >0.50. Kline (2005).
Square roots of AVE > inter-correlations among latent variables.
Thus, convergent validity and discriminant validity was achieved for all the latent
variables in this study.

38
 Structural Model Analysis and Hypothesis
Testing
• A structural model is the model that specifies how
latent variables relate to each other (Arbuckle,
2005)
• It determines which variables have direct or
indirect effects on the values of other latent
variables (Kline, 2005).
• After achieving construct validity and reliability,
the next step was to specify and test the
hypothesised model.
39
 Training

Information Sharing NS

-0.12**
0.21
Rewards
-0.16**

Appraisal Turnover
NS
intentions

Participation NS

NS
Selection

-0.14**

Job Security
-0.18**

Promotion

Figure 4.1: Structural model of relationships between individual HRM practices and
turnover intentions
40
 RS

TD

PO
0.22

PA -0.47***

TOI
HRMP
PD

ES

IS

RW

RS = Recruitment and selection; TD = Training and Development; PO = Promotion Opportunities; PA =

Performances Appraisal; PD = Participatory Decision Making; ES = Job Security; IS = Information Sharing;
RW = Rewards and Recognition

Figure 4.2: Structural model of relationships between a system of HRM practices and
turnover intentions

41
 Promotions

Information Sharing .125*

.253***
Participation 0.45
.018(NS)

Appraisal
.009(NS)
Job
Satisfaction
Selection .283***

-.054(NS)
Training

.145**

Rewards
.167***

Job Security

Figure 4.3: Structural model of the direct relationship between individual HRM
practices and job satisfaction

www.Quant.KU.edu 42
 RS

TD

P
JS1
O
0.46

P 0.68***
A JS2
HRMP JS
PD
JS3

ES

IS

RW

Figure 4.4: Structural model testing the relationship between a system of HRM practices
on job satisfaction

43
 Selection

Training .13**

-.07 (NS)
Rewards 0.45
..25***

Job Security
.16***
Organisational
commitment
Promotions .15***

.21***
Information

.01(NS)

Participation
.08 (NS)

Appraisal

Figure 4.5: Structural model of the relationships between individual HRM practices and
overall organisational commitment

www.Quant.KU.edu 44
 RS

TD
AC

P
O 0.44

P 0.66***
OC NC
A
HRMP
PD

ES
CC

IS

R
W

HRMP = Human Resource Management Practices; RS = Recruitment and selection; TD = Training and
Development; PO = Promotion Opportunities; PA = Performances Appraisal; PD = Participatory Decision
Making; ES = Job Security; IS = Information Sharing; RW = Rewards and Recognition; OC = Organisational
Commitment; AC = Affective commitment; NC = Normative Commitment; CC = Continuance Commitment

Figure 4.6: Relationship between a system of HRM practices and organisational

commitment

45
 Selection

Training .17**

.07
Rewards 0.45
NS

Job Security
NS OCB

Promotions NS

.32***
Information

NS

Participation
NS

Appraisal

Figure 4.7: SEM model relating individual HRM practices and OCB

46
 RS

TD

OCBO
PO
0.17

PA 0.41***
OCBI
OCB
HRMP
PD

OCBS
ES

IS

RW

HRMP = HR Management Practices; RS = Recruitment and selection; TD = Training and Development; PO

= Promotion Opportunities; PA = Performances Appraisal; PD = Participatory Decision Making; ES = Job
Security; IS = Information Sharing; RW = Rewards and Recognition; OCB = Organisational Citizenship
Behaviour; OCBO = OCB Organisation; OCBI = OCB Individual; OCBS = OCB Students

Figure 4.8: SEM model relating a combined system of HRM practices and OCB

www.Quant.KU.edu 47
 TOI1
0.33
JS1

-.58***
TOI2

JS2 JS TOI
TOI3

JS3
TOI4

JS= Job satisfaction; TOI = Turnover Intentions

Figure 4.9: Structural Model of the direct relationship between job satisfaction and
Turnover intentions

www.Quant.KU.edu 48
 Affective -0.55*** 0.50

Commitment

-0.17*** Turnover
intentions

Normative
Commitment -0.03
(NS)

Continuance
Commitment

Figure 4.10: Structural Model of direct relationships between organisational

commitment dimensions and turnover intentions

www.Quant.KU.edu 49
 AC
0.56

-0.75***
NC TOI
OC

CC

AC = Affective Commitment; NC = Normative Commitment; CC = Continuance Commitment; OC = Organisational

Commitment; TOI = Turnover Intentions

Figure 4.11: Structural model Relating overall organisational commitment and turnover
intentions

50
 TOI1
OCBO 0.50
TOI2
-0.32***

OCBCW OCB TOI

TOI3

OCBS
TOI4

OCBO = OCB organisation; OCBI = OCB Individual; OCBS = OCB Students; OCB = Organisational
citizenship behaviour; TOI = Turnover Intentions

Figure 4.12: Structural model of the direct relationship between OCB and turnover
intentions

www.Quant.KU.edu 51
 .66

Organisational
Commitment

-.78***
.58
.77***
Job
.55
Satisfaction -.18***
.76***
System of HRM Turnover
Practices -.21** Intention
s
.45*** .26 -.03 (ns)

OCB

* p< .05; ** p<.01; *** p<.001. OCB = organisational citizenship behaviour; ns = not significant

Figure 4.13: The Full Structural Equation Model

χ2 =1732.549; df = 725 p = 0.000; χ2/df = 2.390; RMSEA = 0.044; SRMR =

0.0539; TLI = 0.945; CFI = 0.948
www.Quant.KU.edu 52
Table 4.12: Maximum Likelihood Estimates for the Full Model

Standardised
Relationship Regression S.E. C.R. P
Weight
OCB  HRMP System .440 .070 8.395 ***
OC  HRMP System .770 .111 11.800 ***
JS  HRMP System .758 .137 12.338 ***
Turnover Intention  HRMP System -.214 .207 -2.370 **
Turnover Intention  JS -.177 .054 -3.360 ***
Turnover Intention  OCB -.027 .064 -.723 .469
Turnover Intention  OC -.784 .097 -10.941 ***

OCB=Organisational citizenship behaviour; OC = Organisational Commitment; JS = Job Satisfaction;

HRMP = HRM Practices
www.Quant.KU.edu 53
 SOBEL TEST

• Sobel test was used to test for significance of

mediation.
• A P-value of less that 0.05 indicates that the
mediation effect is significant.

www.Quant.KU.edu 54
 Sobel test of Mediation

Table 4.13: Summary of the results on mediation hypotheses (Sobel Test)

SN Hypothesised Relationship z-score p-value Mediation Resu

H8 HRMPJSTOI 3.148** <.01 Supported
H9 HRMPOCTOI 7.100** < .001 Supported
H10 HRMPOCBTOI .460(ns) > .1 Not supported
*** p < 0.001, ** p < 0.01, ns. p < 0.1. HRMP =HRM practices; OC= organisational commitment; TOI =
turnover intentions; JS= job satisfaction; OCB = Organisational citizenship behaviour; ns = Not Significant

www.Quant.KU.edu 55
Table 4.14: Summary of all the results of hypothesis testing

Hypothesis Hypothesised relationship Result

H1 HRM practices will be negatively related to turnover Supported
intentions. (partially)
H2 HRM practices will be positively related to job satisfaction. Supported
(partially)
H3 HRM practices will be positively related to Supported
organisational commitment (partially)
H4 HRM practices will be positively related to organisational Supported
citizenship behaviour (partially)
H5 Job satisfaction will be negatively associated with turnover Supported
intentions.
H6 Organisational commitment will be negatively associated Supported
with turnover intentions.
H7 Organisational citizenship behaviour will be negatively Supported
related to turnover intentions.
H8 Job satisfaction will mediate the relationship between HRM Supported
practices and turnover intentions.
H9 The relationship between HRM practices and turnover Supported
intentions will be mediated by organisational commitment.
H10 Organisational citizenship behaviour will mediate the Not Supported
relationship between HRM practices and turnover
intentions

www.Quant.KU.edu 56
Way forward

THANK YOU

57