Employee Retention and Turnover at CSI

Giovanni Circo, Ph.D.

Why Employees Stay

  • In short, inertia
    • Employees generally stay with a company until something causes them to leave
  • What factors cause them to leave?
    • Job satisfaction
    • Perceived opportunities (advancement)
    • Financial pressures
    • Work environment

The Question(s)

  1. What are factors that lead to longer or shorter employee retention?


  1. Can we predict who will be more likely to leave?


  1. What are the characteristics or patterns among “bad hires” ?

Some Terminology

  • CSI positions were broadly combined into 4 categories:
    1. AIC: Alternatives in the Community and Day Programs
    2. Therapists: FFT,MST1, and other therapists
    3. Residential: Community residential programs
    4. Other: Admin, support, and managerial staff


  • Separation: Voluntary or involuntary leaving of CSI
  • Bad Hire: Individual employed by CSI for 6 months or less

The Data

  • Employee HR data from 1,462 individuals hired by CSI from 2010 through 2021

Cumulative Employees Hired, by Position (2010-2021)

The Model

  • A statistical model was used to determine which variables are predictive of employee retention1.


  • Estimates from this model can be used to isolate factors relating to turnover, net of other factors.
# Final Model
fit1 <-
  gam(
    total_time_hired ~ sex + race + status + position +
      admin_category + exempt + city +
      s(salary, by = status) + s(age_at_hire, by = sex),
    weights = terminated,
    family = cox.ph(),
    data = gam_data
  )

Median Time Employed

CSI Position Types

The median time employed in CSI was just under 2 years for those in the AICs, compared to about 1 year for therapists, and less than 1 year for residential employees.

Time Employed
Other 1.91
Aic 1.77
Therapist 1.15
Residential 0.85

What Factors are Most Important?

  • The top two most important factors were salary and site location.
ChiSq Importance
Salary, Full-time 161.94 0.43
Site Location 102.61 0.27
Salary, Part-time 35.36 0.09
Admin Status 19.01 0.05
Age at Hire, female 17.28 0.05
Race 12.70 0.03
Position Type 11.39 0.03
Exception Status 8.65 0.02
Age at Hire, male 6.93 0.02
Sex 0.35 0.00
Full-Time, Part-Time Status 0.02 0.00

Salary and Age at Hire

Probability of Separation, Model Estimates

Site Location

Probability of Separation, by Site Location (2010-2021)

What are “bad hires” ?

  • A “bad hire” doesn’t imply a “bad” employee!
  • There can be a number of factors that lead to quick turnover
    • Not a good match for the position
    • Not ready for the specific demands
    • Not fully aware of the needs of the position

“Bad Hires” by Position

Position Types

  • Just over 50% of all “bad hires” were individuals in residential positions.
  • About 20% were in therapist positions.
  • Less than 1% were in the AICs.
Position Type Position N Prop.
Service Residential 143 0.43
Professionals Therapist 65 0.19
Professionals Residential 27 0.08
Professionals Other 23 0.07
Service Other 23 0.07
Managerial Other 13 0.04
Other Residential 13 0.04
Managerial Residential 12 0.04
Other Other 8 0.02
Service Aic 4 0.01

“Bad Hires” by Position and Year

Proportion of “Bad Hires” (2010 - 2021)

Employee Retention at CSI

What we know

  • Salary is important, especially for those at the lower end of the scale.


  • There are substantial differences in turnover based on the site location


  • Certain positions (residential) are more prone to short-term “bad” hires

What we need to know

  • Determine what programs with low turnover are doing right.


  • Find incentives and improve onboarding for high-turnover positions


  • Use data and organizational knowledge to select “good fits”

Questions?