The Journal of College and University Student Housing - Volume 45, No. 2 - 57

James C. Stoner

models but now included the 5-point increment
weights. Considering all models, the largest
variance explanations occurred between approximately 15-35 for application weights, 20-35
for reference weights, and 40-65 for interview
weights. In an effort to exercise caution when
generating weighting combinations, some
models outside the 5-point increment ranges
were also created. Every combination within
those ranges was established at 1-point increments (e.g., Model 320 [23A | 29R | 48I], Model
321 [23A | 30R | 47I], and Model 322 [23A | 31R |
46I]), creating a total of 270 additional models.
Each model was entered into a simple linear regression with the total weighted score serving as
the predictor variable and perfor-mance evaluation serving as the dependent variable.
Of all 394 models run on the collected Spring
2016 recruitment data, 349 (or 88.6%) revealed
significant prediction capabilities of Fall 2016
performance evaluations. The maximum explanation of variance in performance evaluations
was 13.1% (p < .01), an amount achieved by 21
of the models (see Table 2 for combinations and
statistics). This concluded the analysis of previous data which informed the hiring committee's decisions moving forward.

Year Two: SP17 Recruitment and FA17
Evaluation Processes
Since the 21 strongest models revealed equal
R2 values and little variation existed in Fvalues, the decision on which combination
of weights to apply to the SP17 recruitment
scoring seemed inconsequential. Within all
21 models, an application weight of 20% appeared in four models, a reference weight of
25% appeared in five models, and an interview
weight of 55% appeared in four models. Therefore, the studied department opted to employ
58

Model 116 [20A | 25R | 55I] to calculate the final
scores for candidates within the SP17 hiring
process. Put another way, applying Model 116's
weights to future candidate assessment would
create final candidate scores where application short-answer responses would contribute
the least (20%), references would contribute
slightly more (25%), and their individual interview would contribute the most to their final
score (55%).
After the Spring 2017 application and interview process concluded, the established
weights [20A | 25R | 55I] were applied to the
candidate scores to create the final weighted
scores for consideration in hiring decisions. At
the end of the Fall 2017 semester, those candidates who were offered and accepted positions
received annual performance evaluations with
the same evaluation rubric used in the previous year. The Model 116 weights significantly predicted performance evaluation scores, F(1, 45) =
8.337, p < .01, explaining 13.8% of the variance
in performance evaluation scores, an increase
from the previous year.

LIMITATIONS
This research study was ultimately limited by
the number of available vacancies for which
the studied department was hiring in each
year. Since candidates who were not selected
to serve in the RA role did not receive performance evaluations, the upper boundary of the
population was determined each year by the
number of available positions. Furthermore, as
this study measured the actual job performance
of selected candidates, the research design did
not consider the potential performance of candidates who were not selected to serve as an
RA; therefore, while the study yielded signifi-

THE JOURNAL OF COLLEGE AND UNIVERSITY STUDENT HOUSING

The Journal of College and University Student Housing - Volume 45, No. 2

Table of Contents for the Digital Edition of The Journal of College and University Student Housing - Volume 45, No. 2