Human Resources predictive risk modeling uses various data points about an employee,
including an employee’s previous work experience, job performance, and personality traits to
identify an employee’s likelihood to persist in a particular job role (Feffer, 2014). Citing a case study
of Xerox deploying predictive risk modeling to assess high turnover in a customer service role,
Feffer claims that predictive risk modeling “offers a myriad of opportunities for cost savings and
efficiencies related to a company’s recruiting and talent management efforts” (Feffer, 2014).
METHODOLOGY
Analytical Approach
The aim of the present study was to describe the employment movement of teachers across
DCPS during the school years 2013-2014 through 2017-2018. In order to describe these
movements, historical snapshot data were obtained via public records requests to DCPS. Since this
data is historical, one limitation of the data presented here is that employee motivations are not
captured. This study also differs from studies which pair surveys with employment data (Goldring,
Taie, and Riddles, Boyd, et. al.) because it does not focus solely on novice teachers.
In addition to using public records requests, some of the data analyzed in the present study
were obtained through FLDOE publicly searchable databases. The data collected are described in
some detail in the following sections. The FLDOE standardizes the collection of many of these
variables for a variety of reasons. Student and Staff information systems must conform to the
standards included in an annual FLDOE publication entitled Information Database Requirements.
These data elements are then able to be compared across multiple school districts.
In order to eliminate variables far outside of the control of the researcher, employment and
school-level data were limited to schools that meet the following criteria. First, the school must be
a traditional public school operated solely by DCPS within the boundaries of Duval County. This
eliminates any school whose express purpose is serving a narrowly-defined population of
students, such as exceptional education schools and alternative learning centers. This also
eliminates all public charter schools and private or parochial, since the day-to-day operations of
these institutions is not controlled by DCPS. Each school studied must have been in continuous
operation during the entire five-year study period; this eliminates any schools that were closed or
reconstituted. Each school studied must also have had a school grade reported each year during
the regular school grade reporting period. Any school with a late or incomplete school grade
during even one year of the survey period was eliminated. Finally, schools that combined across
multiple grade level bands (such as a K-8 school or a combined middle and high school) were
eliminated. After eliminating these schools, 132 schools were within the study population. See
Appendix Item 1 for a full listing of these schools.
Employment data was also limited to control for some externalities. Employment data was
collected as a mid-year snapshot of employed individuals for each year during the five-year period.
This eliminates any vacant positions but may also obscure multiple turnovers within a single job
position. Employment data included demographic characteristics of both teachers and adminis-
trators. Employee data included only individuals listed as full-time teachers or full-time adminis-
trators with state certification. Teachers certified only to teach vocational or career and technical
education courses were excluded from analysis due to the differing credential requirements for
career and technical education teachers. For more context on the various pathways to education,
see Appendix Item 2.
11