There are a variety of
procedures used for statistical modeling. While we will not dwell on their
range and specifications here, this section will take you through how we use one
such approach, referred to as Path Analysis, to help tell a more complete story
of a company's business and the important relationships that drive it.
Path analysis allows us to
look at the relationships among a number of interrelated areas on a survey and
see how they work individually and together in helping to produce desired
outcomes. The diagram below is a sample path diagram similar to the ones
that we often obtain when we use modeling procedures to understand the findings we obtain on employee surveys. This approach is also very
useful in customer surveys. Sound statistical modeling, however, requires that you
have a fairly large sample size, generally of several hundred or more survey
respondents.
A sample path
diagram is presented below. The boxes represent the factors in the model that are influencing
employee loyalty and commitment. Within each box is a statistical value
labeled R2. This value is an expression of the amount of
variability in that factor that is accounted for in the data. It can vary
from 0 (none at all) to 1.0 (completely accounted for). When these values
are low (well below 0.5), then we are not accounting for that factor or
dimension well and we may wish to add or refine the questions in that area on
future surveys.
Each box or each line of
text has arrows associated with it. With each arrow (or with each line of
text on the right hand side) there is a value labeled "r". This is the
correlation that the factor or survey item has with the other factors or outcome
variables that it is associated with. These values also vary between 0 and
1.0. But, they will tend to be a bit lower than the R2 values
because they are linking only a single item or a single factor to another.
Generally, correlations of 0.2 are considered low (modest), those around
0.4-0.6 are moderate, and those approaching or exceeding 0.8 are considered high
(strong).
Path modeling requires
that we define a desired outcome. In this case, we selected employee
commitment and loyalty, which was assessed using a series of questions that
asked employees about how committed they are to a career with the company and
how likely they would be to stay with the company even though they were offered
a comparable job elsewhere.
What does the path diagram
tell us? First, we are accounting for a respectable 67% of the variation
in our outcome measure - labeled Employee Commitment and Loyalty in the
yellow box on the left hand side of the diagram above. This, combined with
some statistical tests that we can run, helps us to have confidence that we have
a good model.
The path diagram, or
statistical model, will not include all items included on a survey.
Rather, it will help us to identify the most important dimensions related to our
desired outcomes and to pinpoint which items are contributing the most to those
dimensions. This helps us to identify action areas and to make strategic
recommendations to companies on how to move their key relationships forward.
The items in text only (not enclosed in boxes) represent those that best
captured each dimension uncovered in the current survey. The correlations
in the parentheses that follow give an indication of which items have the most
impact.
Second, we can see that
job satisfaction (in this case assessed by items on how much employees say they
like their jobs and on how likely they would be to recommend the company to
others as a good place to work) is an important predictor of loyalty and
commitment. Obviously, it is important that employees like their jobs.
But, simply liking one's job is not the only the only dimension that accounts
for loyalty and commitment. Other important factors are views of
leadership, perceptions of their managers, having a sense of empowerment,
satisfaction with compensation and benefits, opportunities for advancement, and
a feeling of teamwork. Day-to-day job satisfaction , however, is a
necessary precursor to employee loyalty and commitment so leadership will need
to address areas in which weaknesses are contributing to suboptimal levels of
employees' job satisfaction.
Views of leadership have
relatively little impact on job satisfaction per se, but they have a moderate
impact on employee loyalty and commitment. In order for employees to be
committed to long term careers with the company, they need to have confidence in
their leadership. This is especially true for how well leadership
addresses and manages issues internal to the company, but it is also influenced
by how well they address external issues and the degree of trust that employees
have in them. By looking at how the company fares on these individual
items and examining open-ended comments, we can see where the company's
strengths and weaknesses are in the employees' eyes and we can get an idea of
some of the issues that employees consider most pressing and how they feel about
how leadership is or is not addressing these issues.
Perceptions of one's
manager, in contrast, have a relatively strong impact on job satisfaction.
It is important to the satisfaction of employees that their managers are
effective team leaders, that they mentor and guide them in the work and careers,
and that they are both accessible to them and are open to the employees' input.
Conversely, satisfaction with managers, on its own, has a relatively
smaller impact on loyalty and satisfaction. This may be because managers
are mobile in this company and are not considered by the employees to be a
relatively fixed parts of their long term work environment. Again, by
looking at the strengths and weaknesses among the ratings of managers on
individual items (and open-ended comments if they are available) we can tell a
fuller story. Perhaps, for example, managers are not fairing well overall
on items reflecting team leadership or in mentoring their employees (suggesting
the need for more in-service training and management review in these areas).
It is also interesting to
note that perceptions of leadership and of managers are not correlated very
strongly with each other. Having a good manager, or good leadership, is
viewed as relatively independent by the employees. A good manager is not
going to have a very strong influence on views of corporate leadership nor is
strong leadership likely to ensure that employees are happy with their managers.
This may be the result of having collected data from employees with a large,
multinational firm where there are quite a few levels separating most managers
from upper level corporate leadership.
A sense of empowerment
among employees is also important, both for their job satisfaction and for their
loyalty and commitment to the company. Employees in this survey are saying
that they value having the authority to do their jobs and having input into
decision making. Interestingly, the path suggests (see the correlation
between ratings of managers and of empowerment) that managers play a key role in
enabling a sense of empowerment among employees. This then becomes an area
worth focusing on with managers in helping them to improve how they work with
their direct reports.
As we would expect,
compensation and benefits are important as well. Of note, however, is the
fact that opportunities for promotion are even more important for job
satisfaction and for long term commitment. By looking at how the company
is faring on individual ratings on the items that feed into these dimensions,
and by reviewing the comments received for any open-ended questions in these
areas, leadership may find ways to increase satisfaction and loyalty without
actually having to re-adjust compensation or benefits. Another interesting
point to note is that items relating to perceptions of equal opportunity and
opportunities for advancement are correlated with each other, though not highly.
This suggests that further analyses are in order to explore differences in
employees' perceptions of opportunities for advancement among relevant employee
subgroups (e.g., to look for gender or ethnic differences, to explore potential
differences among job types or levels, etc.) and if there are differences, to
pay particular attention to the comments made by respondents in groups that view
themselves as having less opportunity for advancement in the company.
Other conjoint analyses (looking at the relationship among two or more items)
might be run to evaluate the extent to which this appears to be a problem among
high performers in particular. If so, this would be further impetus for
addressing these issues within the company.
The last dimension
identified is the sense of teamwork among employees. This is especially
important for job satisfaction and it also impacts on employee loyalty and
commitment. This suggests that it is important for managers to foster
collaboration within and across workgroups and that internal communications that
bolster company spirit can serve to boost satisfaction and retention among
employees. These, in turn, represent important areas to investigate in
other company-wide research, including communications audits and management
reviews (e.g., 360o feedback studies).
This rather long
discussion actually represents only the tip-of-the-iceberg in what we can learn
from statistical modeling of the data obtained with well-designed surveys.
Indeed, in addition to surfacing important relationships and suggesting areas
for tactical and strategic initiatives, a good statistical model can raise
important questions for future investigations. These questions might
suggest the need for further interviews or focus group work and redesign of
survey content in future rounds. The accompanying analyses of strengths
and weaknesses and the identification of differences among subgroups also help
to identify areas to track in future surveys and to guide leadership towards
addressing the most critical issues and those that are most likely to have the
greatest impact on improving business relationships.
