Content uploaded by Christopher A. Cooper
Author content
All content in this area was uploaded by Christopher A. Cooper on Oct 06, 2021
Content may be subject to copyright.
The Phoenix Pay System and Intention to Quit the Federal Public Service
Christopher A. Cooper
University of Ottawa
Luc Turgeon
University of Ottawa
Abstract: Since its roll out in 2016, the Phoenix pay system has been mired by an
unending string of problems negatively affecting hundreds of thousands of public
servants costing billions of dollars to fix. But Phoenix’s negative consequences go well
beyond these obvious and immediate effects; more latently, Phoenix is potentially
weakening a core component of good governance: a permanent public service. Drawing
upon psychological contract breach research and using data from the 2017 Public Service
Employee Survey, this article investigates whether employees experiencing problems
with Phoenix are more likely to intend to quit the public service. The results from various
regression models show that the greater the problems an employee has endured from
Phoenix, the more likely they are to express their intention to quit the federal public
service in search of employment elsewhere. More worrying still, this relationship is
particularly pronounced among younger and more educated public servants, thus
signifying a potential loss of future talent for the federal public service.
Keywords: Permanency, Phoenix pay system, Psychological contract breach, Public
service, Turnover, Good governance.
Preprint:
Cooper, Christopher A., Luc Turgeon. 2021. The Phoenix Pay System and Intention to
Quit the Federal Public Service. Canadian Public Administration, 64 (3), 389-420.
https://doi.org/10.1111/capa.12434
Phoenix is a defining moment—a wake-up call—that...needs to
lead to a deeper understanding and correction of the pervasive
cultural problems at play.
- Michael Ferguson, Auditor General of Canada (2011-2019)
Rob Paterson, president of Alterna Savings, had never before seen anything like it. Over
the last few weeks, hundreds of members were arriving at the branches of his credit union
in tears. Having already gone through their savings to pay for their mortgage and other
essentials, these financially and emotionally exhausted members were now asking to cash
in Registered Retirement Savings Plans and stop life-insurance payments. Paterson kept
hearing two commonalities in the stories these members told: they worked for the federal
public service, and they had been burnt by the new Phoenix pay system introduced seven
months before. To this point, Paterson confessed “this has been a testament to public
servants because I don’t know how many other employees would continue to work if they
were not getting paid” (May 2016: A10).
Since its roll out in February 2016, the Phoenix pay system has been mired by an
unending string of problems. Political actors and the media have been attentive to the
destructive consequences Phoenix has had on the lives of thousands of public servants as
well as the enormous waste of taxpayers’ dollars required to fix the problem. Yet the
negative consequences of Phoenix might go well beyond these obvious and immediate
effects. More latently, Phoenix might be leading some employees to leave the federal
public service, and as such, weaken a core pillar of good governance: a permanent public
service. High levels of employee turnover not only hurts organizational performance
(Boyne et al. 2011; Shaw 2011), but within the public service, can weaken the overall
quality of governance (Cornell 2014; Nistotskaya and Cingolani 2016).
Rob Paterson—with more than two decades of executive experience—was baffled
by how many distraught members continued to work for the public service. Paterson’s
astonishment is also justified scientifically. Research studying psychological contract
breaches—an employer’s failure to respect implicit or explicit promises made to an
employee—provides good theoretical reasons to expect that Phoenix is pushing some
bureaucrats to quit the federal public service in search of employment elsewhere (Lemire
and Rouillard 2005). Drawing upon this research, this article uses data from the 2017
Public Service Employee Survey (PSES) to test the relationship between experiencing
problems with Phoenix and an employee’s intention to quit the public service. The results
from several multivariate regression models controlling for various personal and
organizational factors also affecting turnover show that experiencing problems with
Phoenix is associated with a higher intention to quit the federal public service. More
worrying still, this relationship is considerably more pronounced among younger and
more educated employees, who under normal circumstances are already more likely to
leave the public service (Lewis and Cho 2011; Pitts et al. 2011), thus signifying a
potential further loss of future talent for the public service. Given that our empirical
analysis excludes public servants’ departing because of retirement or an end of contract,
we believe that the results point to a latent problem facing Canada’s public service that
has yet to be seriously recognized.
The remainder of this article is organized into five sections. The first section
provides a brief background of the Phoenix pay system. The second section explains why
bureaucratic turnover is such an important issue and can weaken the quality of
governance, before reviewing what we know about the causes of turnover. Drawing upon
psychological contract theory, this section also develops hypotheses between
experiencing problems with Phoenix and turnover intention. The third section describes
the data and research methods. The fourth section presents and interprets the results of
our empirical analysis. The conclusion considers what our findings suggest about the
potential latent affects Phoenix might have on the quality of governance within Canada’s
public service.
Background of Phoenix
In 2018, following public outcry by civil servants, a string of damaging stories in the
media and critical reports by the Auditor General, the Trudeau government signaled its
intention to move away from Phoenix, a new pay system that had been rolled out in 2016.
As a result, a NextGen HR and Pay team was created within the Treasury Board
Secretariat to explore options to replace Phoenix. The team made clear that before
transitioning to a new system, Phoenix would have first to be stabilized (Benay 2018), a
process that would take five years and cost approximately $2.6 billion dollars (Office of
the Parliamentary Budget Officer 2019: 6). But how did we get in such a mess? Why was
Phoenix created in the first place?
In 2007, Public Works and Government Services Canada indicated in its
Departmental Performance Report that its pay system relied on outdated technology and
was too dependent on the knowledge and expertise of employees who were set to soon
retire (Senate 2018: 11). As a result, in 2009, the Government embarked on a new
initiative—Transformation of Pay Administration Initiative (TPAI)—that included two
distinct but overlapping dimensions: pay consolidation and pay modernization (Goss
Gilroy 2018).
Pay consolidation involved moving from a decentralized to a centralized system
of pay. At the time of TPAI’s launch, each department and agency was responsible for
processing the pay of their own employees and had a number of pay advisors responsible
for processing pay, providing advice to employees and correcting errors. The total
number of pay advisors was over 2,000, who were spread over more than 101
departments and agencies (Office of the Auditor General 2017). In 2012, Public Services
and Procurement Canada began the process of centralizing responsibility for pay services
in a new Public Service Pay Centre in Miramichi, New Brunswick. By 2016, around
1,200 pay advisor positions had been eliminated and replaced with 460 new pay advisors
and 90 support staff (Office of the Auditor General 2017).
Pay modernization involved the replacement of existing software, the Regional
Pay System (RPS), viewed as “severely outdated, ineffective and cost-inefficient” (Goss
Gilroy 2018), in part because it required many manual processes by pay advisors. The
software chosen to replace the RPS was a commercial off-the-shelf system that Public
Services and Procurement Canada named Phoenix and IBM was hired to help “design,
implement, integrate and deploy (...)” (Office of the Auditor General 2017).
The Phoenix pay system was launched in two waves: 34 departments and
agencies in February 2016 and another 33 in April 2016 (Senate 2018: 13). Right from
the beginning, Phoenix was plagued by complications, stemming from various sources,
including incomplete file transfers and a lack of experience and training among the newly
hired pay advisors. According to the Office of the Auditor General (2017), an estimated
62% of public servants were paid incorrectly at least once in the 2016-2017 fiscal year.
As a result, the office reported that pay errors associated with Phoenix were worth
approximately $520 million: the government owed 51,000 employees a total of $228
million and 59,000 employees owed the government a total of $295 million.
In his 2018 Spring report, Auditor General Michael Ferguson concluded that
Phoenix was an “incomprehensible failure” (Office of the Auditor General 2018) caused
by an underlying cultural problem within the public service. Echoing concerns voiced by
some Canadian (Savoie 2013) and international scholars (Boräng et al. 2018), Ferguson
specifically blamed an overly “obedient public service” unwilling to provide frank and
fearless advice to governments (Office of the Auditor General 2018), itself caused by a
trend of surging political control over the bureaucracy (Peters and Pierre 2004),
evidenced by a rising influence of ministerial staff and a shortening of senior public
servants’ tenure (Office of the Auditor General 2018).
Unfortunately, while Ferguson saw the massive failure of Phoenix as an outcome
caused by a public service whose “ability to convey hard truths has eroded” (Office of the
Auditor General 2018), the present article argues that Phoenix itself has become an
affect; further weakening Canada’s quality of governance by pushing some public
servants to quit their jobs in light of what they perceive to be a massive failure of their
employer to satisfy an essential obligation of any employee-employer relationship: pay.
Why does Turnover Matter?
Not only is a permanent public service a core characteristic of Weber’s ideal-type
bureaucracy, but it was also a main objective of the UK’s seminal Northcote-Trevelyan
report in 1854, which sought to improve the quality of governance and influenced
administrative reforms in many commonwealth countries during the early 19th century.
For instance, a Royal Commission in Nova Scotia stated:
We place at the forefront of our recommendations in relation to those problems
that fall within the purview of the Provincial Government the recommendation
that a permanent civil service should be established at the earliest opportunity.
The general arguments in favour of such a service are so well known that they
need not be repeated here. (emphasis added, Nova Scotia 1934: 87)
Permanency has since become a primary component of the “public service
bargain” in Westminster countries (Cooper 2020); governments agree to hire public
servants according to merit and give them permanency, and in return, public servants
agree to impersonally provide governments with their objective expertise. Recent
research suggests that permanency is indeed an important factor of good governance
(Cornell 2014; Miller and Whitford 2016). Permanency can lead public servants to
develop unique preferences, foster their expertise and increase their willingness to
provide candid advice, and improve accountability.
Unique preferences. Being elected, politicians tend to make decisions that
prioritize short-term consequences and electoral repercussions (Boräng et al. 2018). In
contrast, because the future employment of permanent public servants does not depend
upon the government in power winning the next election, permanency encourages
bureaucrats to develop preferences that take long-term consequences and scientific
information into consideration more than is the case with popularly elected transient
politicians (Miller and Whitford 2016).
Expertise. Permanency can foster expertise at the individual and organizational
level. At the individual level, the longer an employee holds their position the more likely
they are to develop knowledge and experience of issues, policies and programs (Cornell
2014). Over time employees also accrue interpersonal knowledge of their organization—
knowing who to talk to, and how to talk to them—a key ingredient in getting things done
smoothly and efficiently (Abrams et al. 2003). Permanency also facilitates the
preservation of knowledge within the organization, by allowing information to be passed
from senior to junior employees (Corbett et al. 2018). For example, in his discussion of
the Phoenix fiasco, Auditor General Ferguson lamented the high turnover of senior
bureaucrats in Public Services and Procurement Canada as well as others, asking, “how
could a Deputy Minister understand the issues...in such a short tenure?” (Office of the
Auditor General 2018).
Fearless Advice. Since Hirschman’s (1970) Exit, Voice and Loyalty, research
shows that employees who intend to stay within an organization are more likely to voice
a dissenting opinion to others and speak out against organizational wrongdoing (Whitford
and Lee 2015). This is because employees who are committed to staying have an invested
interest in their organization’s long-term well-being (Cohen 1993). Conversely,
employees who intend to soon depart are less likely to go through the trouble of voicing a
dissenting opinion or reporting wrongdoing, because personally, they have less to gain
from any organizational improvement that their efforts may produce (Burris et al. 2008).
Accountability. Having permanent employees improves organizational
accountability. While the Westminster principle of ministerial responsibility impedes
public servants from being politically accountable to parliament (Savoie 2008),
permanency does increase accountability within the bureaucracy. When public servants
occupy the same job for many years, it is much easier to identify, ask questions and
obtain information from the appropriate people. Moreover, frequent turnover allows
individuals, and even the organization itself, to escape blame by simply dismissing past
wrongs with a “that was then [or them], this is now” mentality (Hood 2011: 70). In sum,
not only does turnover create a need to search for, hire and train new personnel—
estimated to cost up to 200 percent of an employee’s annual salary (Allen et al. 2010)—
but it also negates many of permanency’s benefits.
Theoretical Framework: Causes of Turnover
How might problems with the Phoenix pay system affect turnover in the federal public
service? On the one hand, there are good reasons to think that it might not have much of
an effect. Public sector employees tend to be more stable within their jobs than private
sector employees (Wang et al. 2012). One reason for this is because public sector
employees are generally more averse to risk than individuals working in the private
sector, thus reducing their openness to quitting their job in search of something new
(Buurman et al. 2012). Although a contested matter (Llorens and Stazyk 2011), and one
that actually varies over time and across job types, conventional wisdom claims that
many public sector jobs are better remunerated than equivalent positions in the private
sector (Palacios and Li 2020), and research shows that satisfaction with pay reduces
turnover intention (Lee and Whitford 2008). Public servants who are satisfied with their
remuneration (problems with Phoenix notwithstanding) may judge that they are
ultimately better off in their present position and decide to stay. Finally, research
studying public service motivation shows that a heightened desire to contribute to society
is an important reason public servants opt to work in the public sector (Steijn 2008). A
strong commitment to serve the public might make bureaucrats less willing to quit their
job.
On the other hand, however, psychological contract theory provides theoretical
reasons to expect that experiencing problems caused by Phoenix might lead some public
servants to quit their job in search of employment elsewhere. According to Sandra L.
Robinson (1996: 575) psychological contracts “entail beliefs about what employees
believe they are entitled to receive, or should receive, because they perceive that their
employer conveyed promises [either implicitly or explicitly] to provide those things.”
Scholars have identified two key categories of psychological contracts: relational
contracts and transactional contracts (Rousseau 1995). Relational contracts refer to
beliefs about the affective bonds between employees and employers. Important elements
of relational contracts include demonstrations of loyalty and support by the employer
through investments in training, career development and job security (Grimmer and Oddy
2007: 155). Transactional contracts refer to financial compensation and other material
benefits associated with employment.
Morrison and Robinson (1997) define an employee’s belief that their employer
has failed to deliver what was promised to them as a psychological contract breach.
Importantly, psychological contract breaches can induce a variety of emotions in
employees, including distrust, dissatisfaction, resentment and anger (Zhao et al. 2007)
and can affect an employee’s relationship with their employer. Psychological contract
breaches can reduce employees’ organizational commitment and their willingness to
partake in activities that benefit the organization but that are not part of their formal job
(Robinson 1996).
Research studying the private sector has shown psychological contract breaches to
lead employees to quit their jobs (Robinson and Rousseau 1994; Clinton and Guest
2014). Despite the unique motivations and greater stability of public servants, Lemire and
Rouillard (2005) also found a positive relationship between psychological contract
breaches and turnover intention within the public sector.
Public servants experiencing a psychological contract breach may consider
quitting the public service for several reasons. First, a common reaction to a negative
workplace event is for employees to express an intention to quit their job (Lum et al.
1998). By leaving their job, an employee no longer continues to participate in a
relationship with the employer who has brought about negative experiences stemming
from the contract breach. Second, employees may believe that quitting their job punishes
their employer who has broken their contractual obligations (Zhao et al. 2007).
Moreover, the impact of a psychological contract breach might be more
pronounced in a country like Canada than in countries from other administrative
traditions. A common distinction in comparative public administration is between closed
and open civil services systems (Suzuki and Hur 2020). Closed civil service systems—
found, for instance, in several Continental European countries such as France and
Germany—are characterized by a bureaucracy where entrance is tightly limited through
formal examination, employees have long careers with tenure protection, and labour
protection laws for public sector employees are different than those for private sector
employees. Open civil service systems—found for instance, in Westminster and Nordic
countries (particularly since New Public Management reforms)—are characterized by a
more flexible approach to entering the public service, greater mobility of personnel
between public and private sector jobs, and weaker distinctions in the labour protection
laws between the public and private sectors. Importantly, Suzuki and Hur (2020) find that
bureaucrats in closed systems are more committed to remain in their jobs than those in
open systems. Being closer to the open side of the continuum, Canadian public servants
experiencing Phoenix problems might be more open to the idea of looking for a new job
outside the federal public service.
In sum, drawing upon psychological contract theory and noting that pay
obligations are the bedrock of any employee-employer psychological contract, we put
forward the following hypothesis:
H1: The greater the Phoenix problems an employee has endured the more likely
they are to intend to quit the federal public service in search of employment
elsewhere.
It might also be the case that the relationship between experiencing problems with
Phoenix and turnover intention is particularly pronounced among younger and more
educated employees. Studies frequently show that because of fewer financial and family
obligations, young employees tend to have higher turnover intentions than older
employees (Lewis and Cho 2011; Pitts et al. 2011).
Some research finds that education is associated with higher levels of turnover
(Cho and Lewis 2012). Education increases an individual’s labour market value and
provides them with more employment opportunities. Because younger and more educated
employees may consider the prospects of beginning a new job to be less costly than older
and less educated public servants, the relationship between experiencing Phoenix
problems and turnover intention might be more pronounced among public servants
belonging to these groups. Accordingly, we put forward two additional hypotheses:
H2: The relationship between experiencing Phoenix problems and intention to
quit the federal public service is more pronounced among younger employees.
H3: The relationship between experiencing Phoenix problems and intention to
quit the federal public service is more pronounced among employees with a
higher level of education.
To be clear, our second and third hypotheses are not that turnover intention will
be higher among younger and more educated employees because of the Phoenix pay
scandal. It is a well-established fact, documented in a number of studies cited above, that
younger and more educated workers have higher turnover rates than other employees. As
such, it is perhaps not surprising that the challenge of retaining younger Canadians in the
public service has been the object of a number of reports over the last decade (Public
Policy Forum 2017). Rather, our second and third hypotheses emphasize that among
those who have experienced problems with the Phoenix Pay system, one would expect a
higher intention to quit the public service in search of employment elsewhere among
younger and more educated public servants, because their lack of family/financial
obligations and higher labour market demand, respectively, makes it less costly for them
to walk out the door.
Data and Methods
Data. We test our three hypotheses with data from the 2017 PSES, a survey conducted as
the pay backlog soared. We use the 2017 PSES because it was the first survey since
Phoenix was introduced and it is the most recent PSES whose microlevel data is
accessible to approved researchers through the Canadian Research Data Centre Network.
The 2017 PSES is a census of federal public servants (including indeterminate and term
employees) working in 86 departments and agencies. The 2017 PSES was administered
by the Treasury Board Secretariat and Statistics Canada between August and September
2017, 18 months after Phoenix was first introduced. Employees are informed that all
answers to the survey are anonymous. Overall 174,544 employees completed the survey
(a response rate of 61.3%).
Although it remains underused by public administration scholars (Charbonneau et
al. 2020), the PSES contains several questions about employees’ work and organizational
environment, including their turnover intention. As such, the 2017 PSES is an excellent
source of data to better understand the attitudes and behaviour of Canada’s federal public
servants, although it does have some limitations.
One limitation is that turnover intention does not capture actual levels of turnover.
Although turnover intention is a strong predictor of actual turnover (Dalton et al. 1999), it
is not a perfect one. Another limitation with the PSES is the possibility of common
source bias (Charbonneau et al. 2020), where observed covariation between two variables
may stem from having a shared measurement source, rather than being causally related.
The questionnaire design of the 2017 PSES, however, followed core procedural strategies
—anonymity of respondents, using different scale properties, providing proximal
separation between measures—to reduce the detrimental effects of this possible bias
between our hypothesized dependent and independent variables (George and Pandey
2017).
The microdata for the 2017 PSES is not publicly available, but is restricted to
approved researchers through application to the Canadian Research Data Centre Program
overseen by Statistics Canada. Two primary advantages of this is that our analysis
includes more fine-grained measurement of variables than in publicly available versions
of the PSES, and includes variables that are removed in publicly available data. For
instance, with specific reference to our research, we have several categories of age group,
and are able to measure certain types of turnover such as internal mobility and retirement.
One limitation of using this data, however, is that we are not able to include additional
data into the dataset nor are we able to report statistics that could jeopardize anonymity.1
With these limitations in mind, as a large survey of federal public servants
administered 18 months after Phoenix was first implemented that measures many
important variables shown to affect turnover intention, the 2017 PSES remains a good
source of data to test our hypotheses. Table 1 outlines the operationalization of the
dependent, independent and control variables. Table 2 shows the central tendencies of the
variables and the internal consistency of variables operationalized with more than one
question.
[Table 1 approximately here]
[Table 2 approximately here]
Dependent variable. Turnover intention was measured by asking employees whether they
intended to leave their current position in the next two years. This operationalization of
turnover intention focused on losing employees who could have continued to pursue a
career within the public service. Employees who were uncertain of their turnover
intention, and employees who intended to leave their position for another job within the
public service, or who expected to leave the public service because of retirement or the
end of a contract, were excluded from the analysis. Employees intending to quit the
federal public service within the next two years in search of employment outside the
public service were coded as leavers. Employees intending to remain in their public
service position were coded as stayers. This is consistent with our theoretical framework
claiming that a strong psychological contract breach will lead employees to consider
ending their relationship with their employer. Table 2 shows that in 2017 4 percent of
employees intend to quit the federal public service in search of employment elsewhere.
Independent variables. The PSES survey included a question on whether employees had
experienced problems with Phoenix. It was measured on a five-point scale from “not at
all” (1) to “a very large extent” (5). Respondents were also given an option of answering
“don’t know” or “not applicable” (these responses were excluded from the analysis). Age
was measured in nine categories according to five year increments, except for the
youngest (24 and under) and oldest (60 and above) groups. Education was measured in
four categories; the lowest being a high school diploma or less, and the highest being a
university diploma above the bachelor’s level. Age and education are both treated as
dummy variables in the analysis.
Control variables. The analysis also controls statistically for several additional factors
that previous research suggests to affect turnover. These variables are measured with a
mix of single-item questions and indices composed of two or more questions.2 An
employee’s satisfaction with their job (Caillier 2011), the fit between an employee’s
interests and the organization (person-organization fit) (Moynihan and Pandey 2008),
their decision-making autonomy (Kim and Fernandez 2017), their work-life balance
(Ertas 2015), and the recognition they receive for their work (Luthans 2000), have all
been shown to reduce an employees’ turnover intention. Employees in indeterminate
positions and full-time positions have also been shown to have lower levels of turnover
intention than employees in casual and part-time positions (Arcand et al. 2010). Our
analysis controls for these variables.
Our models also control for a number of socio-demographic variables. First, we
control for the respondent’s gender. Whereas older studies frequently found turnover
intention to be lower among men than women (Lewis and Park 1989), some more recent
research suggests that as the “male-breadwinner” family model has become less common
women no longer have a positive relationship with turnover (Moynihan and Landuyt
2008).
In addition to age, we also control for the number of years an employee has
worked within the public service. Time within an organization has been shown to have a
U-shaped relationship with turnover, with newer and more veteran employees displaying
the greatest propensity to leave (Pitts et al. 2011). There are three reasons for this. First, it
is quite common for some new employees to quit their job within an initial period
because their job does not meet their expectations. Employees remaining after this initial
period, however, are generally satisfied with their job and are less likely to leave until
they approach retirement. Second, over time employees tend to develop a great deal of
knowledge and experience specific to their present job that will not necessarily be
appreciated by another employer. Finally, more experienced employees tend to have
accrued a number of benefits with their employer, which a new employer might not be
able to match. To account for this U-shaped relationship we also include in the models a
squared value of years within the public service.
We also control for an employee’s first official language. A number of reports
have found that Francophones are more likely than Anglophones to state difficulties in
working in the language of their choice (Borbey and Mendelsohn 2019), and some
studies suggest that the language environment at work can be associated with higher
turnover intention (Froese et al. 2016).
Methods. The hypotheses are tested with multivariate binomial logistic regression in four
separate models. Survey weights adjusting for differences in the distribution of
occupational groups between respondents and the public service population are used in
all models.
The first hypothesis is tested in Models 1 and 2. Model 1 regresses turnover
intention against experiencing problems with Phoenix while controlling for an
employee’s age, gender, language, education, years working in the public service,
whether they have a full-time position, whether they have an indeterminate position and
whether they are supervisor. Model 2 regresses turnover intention against experiencing
problems with Phoenix, but in addition to the controls used in Model 1, it controls for job
satisfaction, work-life balance, decision-making autonomy, job-interest fit, job-skills fit,
workplace recognition, and the support employees receive from their supervisor.
The second hypothesis is tested in Model 3 that includes all of the control
variables in Model 2, but introduces an interaction term between experiencing problems
with Phoenix and an employee’s age group. The third hypothesis is tested in Model 4 that
uses the controls found in Model 2, but introduces an interaction term between
experiencing problems with Phoenix and an employee’s level of education. Models 3 and
4 thus allow us to examine whether the relationship between Phoenix problems and
turnover intention is particularly stronger or weaker across differing age groups and
education levels.
The results from the logistic regression models are reported with odds ratios. We
follow recommended practice (Mize 2019) and do not judge whether there is evidence of
a moderated relationship in our nonlinear model by interpreting the p-value of the
interaction term. Rather, we judge the presence of a moderated relationship by examining
the predicted probability of turnover intention across varying levels of experiencing
Phoenix problems at each age group and education level and consider overlap in
confidence intervals (Mize 2019). Statistically significant differences between age groups
and education levels are interpreted by non-overlapping confidence intervals (Cumming
2009).
Results
The results from Models 1 and 2 are reported in Table 3. They show that experiencing
problems with Phoenix has a positive and statistically significant relationship (p<.001)
with turnover intention. Specifically, with reference to Model 2, a one-unit increase in
experiencing problems with Phoenix is associated with a 16 percent increase in the odds
of intending to quit the public service to find employment elsewhere (95 percent
confidence interval ranging from 12 to 21 percent increase in odds).
The results from these models also show that many of the control variables have a
statistically significant relationship with turnover intention. With specific reference to
Model 2, women (p<.001) and Francophones (p<.001) have a lower intention to quit
one’s position in search of employment outside the public service than men and
Anglophones. Having a full-time position (p<.001) and having an indeterminate position
(p=.015) both have negative relationships with intending to quit one’s job in search of
employment elsewhere. Model 2 also shows that being a supervisor has a positive
relationship with turnover intention (p<.001), which is consistent with research
suggesting that managers tend to experience more pushing forces—such higher levels of
stress—and pulling forces—such as opportunities from other employers—leading them
to have higher quit rates than non-managerial staff (Kangas et al. 2018).
[Table 3 approximately here]
The results also show that many aspects of an employee’s job have a negative
statistically significant relationship with turnover intention. A higher level of general
workplace satisfaction (p<.001), a better work-life balance (p=.001), a job that reflects an
employee’s interests (p<.001), workplace recognition (p<.001) and supervisor support
(p<.001) are negatively related to an employee’s intention to quit the public service.
Not all of the controls, however, have a statistically significant relationship with
turnover intention. With specific reference to Model 2, the degree of fit between an
employee’s skills and their job (p=.294) as well as their degree of decision-making
autonomy (p=.236) do not appear to affect turnover intention.
The results from Models 1 and 2 show that generally speaking, an employee’s age
has a negative relationship with turnover intention and that education has a positive
relationship. But is the association between experiencing Phoenix problems and turnover
intention more pronounced among younger and more educated employees?
This is examined in Models 3 and 4. The regression results of these two models
are displayed in Tables 4 and 5. Figures 1 and 2 display the predictive probability that
experiencing Phoenix problems has with turnover intention across differing age groups
and education levels. To preserve legibility, the Figures only display 95 percent
confidence intervals for select values of age group and education level. The complete list
of predicted probabilities and confidence intervals for every age group and education
level are reported in Tables 6 and 7, respectively.
[Table 4 approximately here]
[Table 5 approximately here]
Figures 1 and 2 support the second and third hypotheses, showing that
experiencing Phoenix problems does lead to a larger increase in the probability of
turnover intention among younger and more educated public servants than it does among
older and less educated bureaucrats. Figure 1 shows that among the youngest age group,
experiencing Phoenix problems leads to a substantive increase in their intention to quit
the public service, meanwhile for those public servants belonging to older age groups,
experiencing Phoenix problems is not associated with as much of an increase in turnover
intention. For instance, the predictive probability of intending to quit the public service
for public servants under 25 with the highest level of Phoenix problems is 14 percent (95
percent confidence interval, .10; .18), which is double the 7 percent probability
(confidence interval, .05; .09) among employees belonging to the same age group but
who did not experience any Phoenix problems.
Looking at the second youngest age group of public servants—those between 25
and 29—Figure 1 shows that experiencing the highest level of Phoenix problems is
associated with a 8 percent predicted probability (confidence interval, .07; .10) of
intending to quit, whereas it is 6 percent (confidence interval, .05; .07) among those who
did not have any problems with Phoenix.
Such a substantive increase in turnover intention as Phoenix problems rise is not
found among older age groups. The probability of intending to quit the public service
among 30 and 34 year old public servants reporting the highest level of Phoenix problems
is 6 percent (confidence interval, .05; .07) and among those between 35 and 39 it is 5
percent (confidence interval, .04; .06). Meanwhile the probability of intending to quit
among employees in both of these age groups who have not experienced any problems
with Phoenix is 4 percent (confidence intervals, .03; .04). Figure 1 also shows that the
substantive increase in the probability of intending to quit the public service alongside
experiencing Phoenix problems is very small among public servants in their 40s, 50s and
60s. For example, employees between 55 and 59 as well as those 60 and older who
reported the highest level of Phoenix problems only had a probability of intending to quit
the public service that was approximately one percentage point greater than employees
belonging to the same age group who reported no problems at all (4 percent vs. 3
percent).
[Figure 1 approximately here]
[Figure 2 approximately here]
Figure 2 shows that experiencing Phoenix problems is associated with a greater
intention of quitting the public service among employees with higher levels of education,
although these differences are not as stark as that found between age groups. For
example, the probability of intending to quit among employees experiencing the most
severe level of Phoenix problems whose highest level of education is above a bachelor’s
degree is 8 percent (confidence interval, .07; .10), and 5 percent among public servants
whose highest level is a bachelor’s degree (confidence interval, .05; .06). These levels
are both approximately two percentage points higher than employees with the same level
of education but who did not experience any Phoenix problems (confidence intervals, .05;
.06 and .03;.04, respectively).
Experiencing phoenix problems does not lead to as much of an increase in
turnover intention among employees whose education is no higher than a high school
diploma. The probability of intending to quit among those experiencing the greatest level
of problems is approximately 3 percent (confidence interval, .02; .03), which is only one
percentage point greater than employees at this educational level who reported no
problems from Phoenix (confidence interval, .01; .02). Interestingly, the greatest
substantive increase in turnover attention is found among employees with a university
certificate. The probability of intending to quit among these public servants who have
experienced the greatest problems from Phoenix is 6 percent (confidence
interval, .04; .07), meanwhile those who have not experienced any problems have a quit
intention of approximately 3 percent (confidence interval, .02; .03).
[Table 6 approximately here]
[Table 7 approximately here]
Robustness Checks. We tested several alternative models to examine the robustness of the
results, including an alternative operationalization of the dependent variable considering
individuals who intended to leave their present job but remain in the federal public
service as “stayers” and models excluding different control variables. These alternative
models did not significantly alter the central findings pertaining to our three hypotheses.
Our results are robust, and our conclusions are not contingent upon a particular model
specification. We are confident in concluding that the Phoenix fiasco is pushing public
servants, particularly the young and more educated, to intend to quit their job within the
federal public service in search of employment elsewhere.
Conclusion
This article has shown that problems with the Phoenix pay system are leading a number
of public servants to consider leaving the public service. The more a public servant
experienced problems with Phoenix, the greater their probability of intending to quit their
job to find something new outside the federal public service. Our results are consistent
with psychological contract theory, which states that a breach in the contract bonding an
employee and an employer is likely to increase an employee’s intention to leave their job.
We have also shown that this relationship is particularly strong among younger,
and to a lesser extent, more educated public servants. This is worrying considering that
attracting and retaining talent in the public service, especially young and educated talent,
has long been prioritized by governments. The UK’s Nothcote-Trevelyan report ([1854]
1954), for example, stated that “[...] endeavours should be made to secure the services of
the most young men of the day [...]” (7) and also that they “[...] should be made
constantly to feel that [...] that with average abilities and reasonable application they may
look forward confidently to a certain provision for their lives [...]” (6). Similarly, in
Canada, the Courtney report of 1908 lamented that “[...] able and worthy young men
attracted by high emoluments have left the service. It is becoming more and more
difficult to fill their places” (Royal Commission on the Civil Service 1908: 14). Without
exaggerating its impact, our study clearly demonstrates that Phoenix has led a significant
number of young public servants to question their decision to pursue a career in the
federal public service.
By focusing on turnover, however, this article possibly underestimates the long-
term negative consequences Phoenix may be having on the public service. Phoenix might
also be reducing other aspects of a high-performing organization such as employee
motivation and the ability to recruit talented personnel. As the head of the Public Service
Alliance of Canada stated in early 2017, “I really think that they’re [the government]
going to have a problem attracting talented individuals to come and work for them when
they can’t pay them...” (Dawson 2017). As the problem persists to this day, albeit at a
lower intensity, it is likely that the impact of Phoenix will be felt for a number of years to
come in a myriad of ways.
Although our conclusion is rather grim, there is another, more optimistic,
possibility. In his damning report, Michael Ferguson hoped that Phoenix would lead to a
“wake-up call” within the public service; to shed its overly obedient and fearful
organizational culture, and to rediscover its identity as the conveyor of hard truths.
Although it remains an open question for future research, there are theoretical reasons to
justify Ferguson’s hope. After all, those bureaucrats having opted to remain in the public
service despite such a large transgression by their employer may be more loyal to the
public service and its mission to serve Canada’s elected government with “open, candid
and impartial” advice in a “spirit of openness, honesty and transparency” (Treasury
Board of Canada Secretariat 2011: 4-5). And having chosen to stay, these public servants
upset with the present state of affairs might just resort to voice, in an effort to bring about
such a change.
References
Abrams, Lisa C., Rob Cross, Eric Lesser, and Daniel Z. Levin. 2003. “Nurturing
interpersonal trust in knowledge-sharing networks.” The Academy of
Management Perspectives 17 (4): 64–77.
Arcand, Guy, Geneviève Tellier, and Lise Chrétien. 2010. “Le défi de la rétention des
jeunes travailleurs: Le cas de la fonction publique québécoise.” Canadian Public
Administration 53 (2): 201–20.
Allen, David. G., Phillip C. Bryant, and James M. Vardaman. 2010. “Retaining talent:
Replacing misconceptions with evidence-based strategies.” The Academy of
Management Perspectives 24 (2): 48–64.
Benay, Alex. 2018. Next Generation HR and Pay, and how we’re going to do things
differently. Retrieved February 5th 2020: https://tbs-blog.canada.ca/en/next-
generation-hr-and-pay-and-how-were-going-do-things-differently
Borbey, Patrick and Matthew Mendelsohn. 2019. The Next Level: Normalizing a Culture
of Inclusive Linguistic Duality in the Federal Public Service Workplace. Ottawa:
Privy Council Office.
Boyne, George A., Oliver James, Peter John, and Nicolai Petrovsky. 2011. “Top
management turnover and organizational performance: A test of a contingency
model.” Public Administration Review 71 (4): 572–81.
Burris, R. Etha, James. R. Detert, and Dan S. Chiaburu. (2008). “Quitting before leaving:
The mediating effects of psychological attachment and detachment on voice.”
Journal of Applied Psychology 93 (4): 912–22.
Buurman, Margaretha, Josse Delfgaauw, Robert Dur, and Seth Van den Bossche. 2012.
“Public sector employees: Risk averse and altruistic?” Journal of Economic
Behavior & Organization 83 (3): 279–91.
Boräng, Frida, Agnes Cornell , Marcia Grimes, and Christian Schuster. 2018. “Cooking
the books: Bureaucratic politicization and policy knowledge.” Governance 31 (1):
7–26.
Caillier, James Gerard. 2011. “I want to quit: A closer look at factors that contribute to
the turnover intentions of state government employees.” State and Local
Government Review 43 (2): 110–22.
Cho, Yoon Jik, and Gregory B. Lewis. 2012. “Turnover intention and turnover behavior:
Implications for retaining federal employees.” Review of Public Personnel
Administration 32 (1): 4–23.
Corbett, Jack, Denis C. Grube, Heather Lovell, and Rodney Scott. 2018. “Singular
memory or institutional memories? Toward a dynamic approach.” Governance 31
(3): 555–73.
Cohen, Aaron. 1993. “Organizational commitment and turnover: A met a-analysis.”
Academy of Management Journal 36 (5): 1140–157.
Charbonneau, Étienne, Geneviève Morin, Itizez Slama, and Fatou Bèye. 2020. “Sondage
auprès des fonctionnaires fédéraux: Synthèse des contributions canadiennes à la
recherche.” Canadian Public Administration 63 (3): 409–28.
Clinton, Michael E., and David E. Guest. 2014. “Psychological contract breach and
voluntary turnover: Testing a multiple mediation model.” Journal of
Occupational and Organizational Psychology 87 (1): 200–07.
Cooper, Christopher A. 2020. At the Pleasure of the Crown: The Politics of Bureaucratic
Appointments. UBC Press.
Cornell, Agnes. 2014. “Why bureaucratic stability matters for the implementation of
democratic governance programs.” Governance 27 (2): 191–214.
Cumming, Geoff. 2009. “Inference by eye: Reading the overlap of independent
confidence intervals.” Statistics in Medicine 28 (2): 205–20.
Dalton, Dan R., Jonathan L. Johnson, and Catherine M. Daily. 1999. “On the use of
‘intent to…’ variables in organizational research: An empirical and cautionary
assessment.” Human Relations 52 (10): 1337–50.
Dawson, Tyler. 2017. “Public service fails to attract the millennials; No creativity, no
flexibility; we have other options, thanks” The Ottawa Citizen, April 11, 2017:
A.7.
Ertas, Nevbahar. 2015. “Turnover intentions and work motivations of millennial
employees in federal service.” Public Personnel Management 44 (3): 401–23.
Froese, Fabian Jintae, Kwanghyun Kim, and Aileen Eng. 2016. “Language, cultural
intelligence, and inpatriate turnover intentions: Leveraging values in multinational
corporations through inpatriates.” Management International Review 56 (2):
2830–301.
Gilbert, Stephanie L., and E. Kevin Kelloway. 2018. “Leadership, recognition and well‐
being: A moderated mediational model.” Canadian Journal Administrative
Sciences 35 (4): 523–34.
Goss Gilroy. 2018. Lessons learned from the transformation of pay administration
initiative. Retrieved at:
https://www.canada.ca/en/treasury-board-secretariat/corporate/reports/lessons-
learned-transformation-pay-administration-initiative.html
Grimmer, Martin, and Matthew Oddy. 2007. “Violation of the psychological contract:
The mediating effect of relational versus transactional beliefs.” Australian
Journal of Management 32 (1): 153–74.
Hickey, Alexandra, and Scott Edward Bennett. 2012. “Canadian public service employee
satisfaction and its main drivers.” Canadian Public Administration 55 (1): 5–23.
Hood, Christopher. 2011. The Blame Game: Spin, Bureaucracy, and Self-preservation in
Government. Princeton, NJ: Princeton University Press.
House of Commons, Standing Committee on Justice and Human Rights, February 27,
2019.
Kangas, Maiju, Muel Kaptein, Mari Huhtala, Anna-Maija La¨msa¨, Pia Pihlajasaari, and
Taru Feldt. 2018. “Why do managers leave their organization? Investigating the
role of ethical organizational culture in managerial turnover.” Journal of Business
Ethics 153: 707–23.
Kim, Sun Young and Sergio Fernandez. 2017. “Employee empowerment and turnover
intention in the U.S. federal bureaucracy.” The American Review of Public
Administration 47 (1): 4–22.
Lee, Soo-Young, and Andrew B. Whitford. 2008. “Exit, voice, loyalty and pay: Evidence
from the public workforce.” Journal of Public Administration Research and
Theory 18 (2): 647–71.
Luthans, Kyle. 2000. “Recognition: A powerful, but often overlooked, leadership tool to
improve employee performance.” Journal of Leadership Studies 7 (1): 31–39.
Llorens, Jared J., and Edmund C. Stazyk. 2011. “How important are competitive wages?
Exploring the impact of relative wage rates on employee turnover in state
government.” Review of Public Personnel Administration 31 (2): 111–27.
May, Kathryn. 2016. “‘People in tears’ over pay delays, credit union says.” The Ottawa
Citizen. 10 Sep 2016: A.10.
McGrandle, Jocelyn. 2019. “Job satisfaction in the Canadian public service: Mitigating
toxicity with interests.” Public Personnel Management 48 (3): 369–91.
Miller, Gary. J., and Andrew. B. Whitford. 2016. Above Politics: Bureaucratic
Discretion and Credible Commitment. New York: Cambridge University Press.
Mize, Trenton D. 2019. “Best practices for estimating, interpreting, and presenting
nonlinear interaction effects.” Sociological Science 6: 81–117.
Nistotskaya, Marina, and Luciana Cingolani. 2016. “Bureaucratic structure, regulatory
quality, and entrepreneurship in a comparative perspective: Cross-sectional and
panel data evidence.” Journal of Public Administration Research and Theory 26
(3): 519–34.
Nova Scotia. 1934. Report of the Royal Commission, Provincial Economic Inquiry.
Halifax: King’s Printer.
Office of the Auditor General. 2017. 2017 Fall Reports of the Auditor General of
Canada. Report 1 - Phoenix Pay Problems. Retrieved from: https://www.oag-
bvg.gc.ca/internet/English/parl_oag_201711_01_e_42666.html
Office of the Auditor General. 2018. 2018 Spring Report of the Auditor General of
Canada. Message from the Auditor General of Canada. Retrieved from:
https://www.oag-bvg.gc.ca/internet/English/parl_oag_201805_00_e_43032.html
Office of the Parliamentary Budget Officer. 2019. Costs Associated with Replacing the
Federal Pay System. Retrieved from:
https://www.pbo-dpb.gc.ca/web/default/files/Documents/Reports/2019/Federal-
pay-system-phoenix/
Costs_Associated_with_Replacing_the_Federal_Pay_System_EN.pdf
Pitts, David, John Marvel, and Sergio Fernandez. 2011. “So hard to say goodbye?
Turnover intention among U.S. federal employees.” Public Administration
Review 71 (5): 751–60.
Public Policy Forum. 2017. Building a Dynamic Future: The Next Generation of Public
Service Talent. Ottawa: Public Policy Forum.
Robinson, Sandra L. 1996. “Trust and breach of the psychological contract.”
Administrative Science Quarterly 41 (4): 574–99.
Rousseau, Denise. 1995. Psychological Contracts in Organizations: Understanding
Written and Unwritten Agreements. Thousand Oaks: Sage Publications.
Royal Commission on the Civil Service. 1908. Report of the Commissioners. Ottawa:
S.E. Dawson, King’s Printer.
Savoie, Donald J. 2008. Court Government and the Collapse of Accountability in
Canada and the United Kingdom. Toronto; University of Toronto Press.
Savoie, Donald J. 2013. What Ever Happened to the Music Teacher? Montreal-Kingston:
McGill-Queen’s University Press.
Senate. 2018. The Phoenix Pay Problem: Working Toward a Solution. Ottawa: Standing
Senate Committee on National Finance.
Shaw, Jason. D. 2011. “Turnover rates and organizational performance: Review, critique,
and research agenda.” Organizational Psychology Review 1 (3): 187–213.
Steijn, Bram. 2008. “Person-environment fit and public service motivation.”
International Public Management Journal 11 (1): 13–27.
Suzuki, Kohei, and Hyunkang Hur. 2020. “Bureaucratic structures and organizational
commitment: Findings from a comparative study of 20 European countries.”
Public Management Review 22 (6): 877–907.
Treasury Board of Canada Secretariat. 2011. Values and Ethics Code for the Public
Sector. Ottawa: Government of Canada.
“The Northcote-Trevelyan Report.” 1954. Public Administration 32 (1): 1–16.
Wang, Yau-De, Chyan Yang, and Kuei-Ying Wang. 2012. “Comparing public and
private employees’ job satisfaction and turnover.” Public Personnel Management
41 (3): 557–73.
Whitford, Andrew B., and Soo-Young Lee. 2015. “Exit, voice, and loyalty with multiple
exit options: Evidence from the U.S. federal workforce.” Journal of Public
Administration Research and Theory 25 (2): 373–98.
Zhao, Hao, Sandy J. Wayne, Brian C.Glibkowski, and Jesus Bravo. 2007. “The impact
of psychological contract breach on work-related outcomes: A meta-analysis.”
Personnel Psychology 60 (3): 647–80.
Table 1. Variable operationalization
Variable Question(s) Response categories and coding
Turnover Do you intend to leave your current position in the next two years?
(Pitts et al. 2011)
Yes – 1
No – 0
Phoenix problems To what extent has your pay or other compensation been affected by
issues with the Phoenix pay system?
Not at all – 1
To a small extent – 2
To a moderate extent – 3
To a large extent – 4
To a very large extent – 5
Job-interests fit My job is a good fit with my interests.
(McGrandle 2019)
Likert-scale: strongly disagree (1) to strongly
agree (5)
Job-skills fit My job is a good fit with my skills.
(McGrandle 2019)
Likert-scale: strongly disagree (1) to strongly
agree (5)
Work-life balance I have support at work to balance my work and personal life.
(Hickey and Bennett 2015)
Likert-scale: strongly disagree (5) to strongly
agree (1)
Supervisor support My immediate supervisor creates an environment where I feel free to
discuss with him or her matters that affect my well-being at work.
(Ertas 2015)
Likert-scale: strongly disagree (1) to strongly
agree (5)
Job satisfaction I get a sense of satisfaction from my work;
Overall, I like my job.
(McGrandle 2019)
Likert-scale: strongly disagree (1) to strongly
agree (5)
(Cronbach’s Alpha .88)
Recognition Overall, I feel valued at work;
I receive meaningful recognition for work well done.
(Gilbert and Kelloway 2018)
Likert-scale: strongly disagree (1) to strongly
agree (5)
(Cronbach’s Alpha .88)
Decision-making autonomy I have opportunities to provide input into decisions that affect my work;
I am encouraged to be innovative or to take initiative in my work.
(Caillier 2011; Ertas 2015)
Likert-scale: strongly disagree (1) to strongly
agree (5)
(Cronbach’s Alpha .85)
Supervisor position Are you a supervisor? Yes – 1
No – 0
Full-time employee Do you work full-time or part-time? Full-time – 1
Part-time – 0
Indeterminate position What is your current employee status? Indeterminate – 1
All other responses – 0
Age To what age group do you belong? Categorical dummy variables (0;1):
Under 25 years
25 to 29 years
30 to 34 years
35 to 39 years
40 to 44 years
45 to 49 years
50 to 54 years
55 to 59 years
Over 59 years
Years in public service In total, how many years have you been working in the federal public
service?
Continuous
Female What is your gender? Female – 1
Male – 0
Education What is the highest level of education you have ever completed? Categorical dummy variables (0;1):
High school
Community college diploma
University certificate/diploma
University bachelor’s degree
Above bachelor’s degree
Francophone What is your first official language? French – 1
English – 0
Justification for variable operationalization in parenthesis.
Table 2. Summary Statistics
Mean Standard deviation 1st percentile 99th percentile
Turnover .04 .19 0 1
Phoenix problems 2.41 1.35 1 5
Job-interest fit 4.38 .83 1 5
Skills-interest fit 4.46 .80 1 5
Work-life balance 4.16 1.06 1 5
Supervisor support 4.39 .97 1 5
Job satisfaction 4.31 .82 1 5
Recognition 3.87 1.08 1 5
Decision-making autonomy 3.94 1.05 1 5
Supervisor position .22 .41 0 1
Full-time employee .95 .21 0 1
Indeterminate position .85 .35 0 1
Age (categorical dummy variables)
Under 25 .03 .17 0 1
25-29 .06 .24 0 1
30-34 .09 .29 0 1
35-39 .13 .34 0 1
40-44 .15 .36 0 1
45-49 .13 .37 0 1
50-54 .18 .39 0 1
55-59 .13 .33 0 1
Over 59 .06 .24 0 1
Years in public service 12.73 9.01 0 35
Education (categorical dummy variables)
High school or less .15 .36 0 1
Non-university diploma .29 .45 0 1
University certificate .05 .21 0 1
Bachelor’s degree .30 .46 0 1
Above bachelor’s .21 .40 0 1
Female .55 .50 0 1
Francophone .33 .47 0 1
Table 3. Regression Models 1 and 2
I II
Odd ratios SE Odd ratios SE
Phoenix problems 1.28*** (.021) 1.16*** (.023)
Job-interest fit .76*** (.031)
Skills-interest fit .95 (.036)
Work-life balance .91** (.025)
Supervisor support .88*** (.025)
Satisfaction .40*** (.017)
Recognition .76*** (.030)
Decision-making autonomy .96 (.034)
Supervisor position .89* (.051) 1.75*** (.127)
Full-time employee .66*** (.072) .63*** (.083)
Indeterminate position 1.02 (.077) .80* (.074)
Age
Under 25 (reference category) (reference category)
25-29 .68** (.082) .59*** (.088)
30-34 .39*** (.047) .30*** (.046)
35-39 .33*** (.040) .27*** (.041)
40-44 .25*** (.032) .22*** (.035)
45-49 .20*** (.026) .17*** (.027)
50-54 .19*** (.061) .15*** (.025)
55-59 .22*** (.030) .19*** (.032)
Over 59 .18*** (.029) .20*** (.039)
Years in public service 1.04*** (.010) .95*** (.011)
Years in public service (squared) 1.00 (.000) 1.00*** (.000)
Female .61*** (.027) .69*** (.037)
Francophone .52*** (.028) .59*** (.037)
Education
High school or less (reference category) (reference category)
Non-university diploma 1.47*** (.151) 1.44** (.174)
University certificate 2.63*** (.354) 2.47*** (.416)
Bachelor’s degree 2.49*** (.243) 2.83*** (.322)
Above bachelor’s 3.77*** (.368) 5.50*** (.636)
Constant .05*** (.001) 96.52*** (22.65)
N 66, 889 63, 910
Pseudo R-squared .07 .36
*=p<.05; **=p<.01; ***=p<.001
Reporting odds ratios. Robust standard errors in parentheses
Table 4. Regression Model 3
III
Odds ratios SE
Phoenix problems 1.30** (.122)
Job-interest fit .76*** (.032)
Skills-interest fit .96 (.036)
Work-life balance .91** (.025)
Supervisor support .88*** (.025)
Satisfaction .40*** (.017)
Recognition .76*** (.030)
Decision-making autonomy .96 (.034)
Supervisor position 1.76*** (.127)
Full-time employee .62*** (.083)
Indeterminate position .80* (.074)
Age
Under 24 (reference category)
25-29 .99 (.353)
30-34 .40* (.152)
35-39 .40* (.145)
40-44 .36** (.128)
45-49 .31** (.113)
50-54 .19*** (.071)
55-59 .27*** (.100)
Over 59 .31** (.130)
Years in public service .95*** (.011)
Years in public service (squared) 1.00*** (.000)
Female .69*** (.037)
Francophone .59*** (.037)
Education
High school or less (reference category)
Non-university diploma 1.43** (.173)
University certificate 2.46*** (.413)
Bachelor’s degree 2.81*** (.320)
Above bachelor’s 5.49*** (.634)
Phoenix problems*Age
24 and under (reference category)
25-29 .85 (.091)
30-34 .92 (.101)
35-39 .88 (.092)
40-44 .86 (.093)
45-49 .81 (.090)
50-54 .94 (.104)
55-59 .92 (.104)
60 and over .87 (.113)
Constant 66.60*** (25.22)
N 63, 910
Pseudo R-squared .36
*=p<.05; **=p<.01; ***=p<.001
Reporting odds ratios. Robust standard errors in parentheses
Table 5. Regression Model 4
IV
Odds ratios SE
Phoenix problems 1.22** (.086)
Job-interest fit .76*** (.032)
Skills-interest fit .96 (.036)
Work-life balance .91** (.025)
Supervisor support .88*** (.025)
Satisfaction .40*** (.017)
Recognition .76*** (.030)
Decision-making autonomy .96 (.034)
Supervisor position 1.75*** (.127)
Full-time employee .63*** (.083)
Indeterminate position .80* (.074)
Age
Under 24 (reference category)
25-29 .59*** (.088)
30-34 .30*** (.046)
35-39 .27*** (.041)
40-44 .22*** (.035)
45-49 .17*** (.027)
50-54 .15*** (.025)
55-59 .19*** (.033)
Over 59 .20*** (.039)
Years in public service .95*** (.011)
Years in public service (squared) 1.00*** (.000)
Female .69*** (.037)
Francophone .59*** (.037)
Education
High school or less (reference category)
Non-university diploma 1.67* (.432)
University certificate 1.91 (.721)
Bachelor’s degree 3.32*** (.806)
Above bachelor’s 6.55*** (1.588)
Phoenix problems*Education
High school or less (reference category)
Non-university diploma .95 (.078)
University certificate 1.08 (.126)
Bachelor’s degree .95 (.074)
Above bachelor’s .94 (.073)
Constant 84.21*** (25.52)
N 63, 910
Pseudo R-squared .36
*=p<.05; **=p<.01; ***=p<.001
Reporting odds ratios. Robust standard errors in parentheses
Table 6. Predicted probability with confidence intervals. Turnover intention and Phoenix
problems by age group
Age Phoenix problems Predicted probability 95 percent confidence interval
Low High
Under 25 Not at all (1) .071 .048 .094
To a small extent (2) .085 .066 .104
To a moderate extent (3) .101 .083 .118
To a large extent (4) .119 .096 .143
To a very large extent (5) .141 .103 .178
25-29 Not at all (1) .063 .053 .074
To a small extent (2) .068 .059 .077
To a moderate extent (3) .073 .064 .081
To a large extent (4) .079 .067 .090
To a very large extent (5) .084 .069 .100
30-34 Not at all (1) .035 .028 .043
To a small extent (2) .041 .035 .047
To a moderate extent (3) .046 .041 .052
To a large extent (4) .053 .046 .059
To a very large extent (5) .060 .049 .070
35-39 Not at all (1) .035 .030 .041
To a small extent (2) .039 .034 .043
To a moderate extent (3) .043 .039 .047
To a large extent (4) .047 .042 .053
To a very large extent (5) .052 .044 .061
40-44 Not at all (1) .031 .027 .036
To a small extent (2) .034 .031 .038
To a moderate extent (3) .037 .033 .041
To a large extent (4) .041 .035 .046
To a very large extent (5) .044 .035 .053
45-49 Not at all (1) .027 .023 .032
To a small extent (2) .028 .025 .032
To a moderate extent (3) .030 .026 .033
To a large extent (4) .031 .026 .036
To a very large extent (5) .032 .025 .040
50-54 Not at all (1) .021 .018 .025
To a small extent (2) .025 .022 .028
To a moderate extent (3) .029 .026 .033
To a large extent (4) .034 .028 .040
To a very large extent (5) .039 .031 .048
55-59 Not at all (1) .027 .022 .031
To a small extent (2) .030 .026 .032
To a moderate extent (3) .034 .030 .039
To a large extent (4) .039 .032 .046
To a very large extent (5) .044 .033 .056
Over 59 Not at all (1) .028 .021 .036
To a small extent (2) .031 .025 .038
To a moderate extent (3) .034 .028 .041
To a large extent (4) .038 .028 .048
To a very large extent (5) .042 .027 .057
Predictive probability and 95% confidence intervals for Model 3. Controls held constant at mean value.
Table 7. Predicted probability with confidence intervals. Turnover intention and Phoenix
problems by education
Education Phoenix problems Predicted probability 95 percent confidence interval
Low High
High school or less Not at all (1) .015 .011 .019
To a small extent (2) .018 .014 .021
To a moderate extent (3) .020 .017 .024
To a large extent (4) .024 .019 .028
To a very large extent (5) .027 .020 .035
Non-university diploma Not at all (1) .022 .018 .025
To a small extent (2) .024 .022 .026
To a moderate extent (3) .027 .024 .029
To a large extent (4) .030 .026 .033
To a very large extent (5) .033 .028 .038
University certificate Not at all (1) .026 .018 .035
To a small extent (2) .032 .025 .039
To a moderate extent (3) .039 .032 .046
To a large extent (4) .047 .037 .058
To a very large extent (5) .057 .040 .074
Bachelor’s degree Not at all (1) .035 .032 .039
To a small extent (2) .039 .036 .042
To a moderate extent (3) .043 .041 .046
To a large extent (4) .048 .044 .052
To a very large extent (5) .052 .046 059
Above bachelor’s Not at all (1) .057 .051 .062
To a small extent (2) .062 .058 .066
To a moderate extent (3) .068 .064 .072
To a large extent (4) .074 .068 .080
To a very large extent (5) .081 .071 .091
Predictive probability and 95% confidence intervals for Model 4. Controls held constant at mean value.
Figure 1. Predictive probability of turnover: Phoenix problems by age group
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
Under 25 25-29 30-34 35-39 40-44 45-49 50-54 55-59 Over 59
Turnover intenon
Note: Predicted probability of turnover intention by Phoenix problems moderated by age group. Displaying 95 percent confidence intervals for “under 25”, “25-29” and “over
59”. Based on Model 3 (Table 4) holding all other covariates constant.
Figure 2. Predicted probability of turnover: Phoenix problems by education
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
HS or less Non-university diploma University cer%cate Bachelor's degree Above bachelor's
Turnover intenon
Note: Predicted probability of turnover intention by Phoenix problems moderated by education. Displaying 95 percent confidence intervals for “Above bachelor’s degree”,
“Bachelor’s degree” and “High school or less”. Based on Model 4 (Table 5) holding all other covariates constant.
1Notes
For instance, we are not able to report the maximum and minimum value range of our variables. We thus report the
value of the 1st and 99th percentile among our observations.
2 The internal validity for some multi-question indices used in past studies was too low. For these variables we included
each question independently in the models. As a robustness test, we tested alternative models that used these indices.
The main results between our independent variables and turnover intention were unchanged.
