Do Retention Bonuses Work? What the Evidence Says

Evidence grade: Mixed. Credible causal studies exist and disagree, or the effect holds only under conditions the studies can name. This grade describes the published evidence. It is not a prediction about your workforce and not a result RetainScore has produced.

This is the rare retention question with real causal evidence

Most retention interventions have never been tested against a comparison group. Retention bonuses have, repeatedly, because public school systems ran them at scale with eligibility rules sharp enough to support a credible design. The results are genuinely mixed, and the pattern in the disagreement is more useful than any single number.

The same bonus, two different answers

Springer, Swain and Rodriguez evaluated Tennessee's US$5,000 retention bonus for highly effective teachers in the state's lowest performing schools, using a fuzzy regression discontinuity design built on the effectiveness rating that determined eligibility. For many recipients the bonus was about a 10 percent salary increase, so this was not a token payment.

Across all eligible teachers, it did nothing measurable. "Point estimates for the main effect of the bonus are not different from zero."

For one subgroup it worked clearly. "For teachers of tested subjects and grades, the program has a consistently positive effect that is both statistically and substantively significant." The authors' explanation is that for teachers of untested subjects, the evaluation system leaned on school-level results, so a teacher in a low performing school carried a rating they could not personally move. A bonus conditioned on a rating you cannot influence is not an incentive.

What the wider record shows

The same paper reviews the other rigorous evaluations, and they do not line up neatly:

• North Carolina, US$1,800 a year. Far smaller, and it worked: the bonuses "reduced turnover rates of eligible teachers by 17%, or 5 percentage points" among eligible math, science and special education teachers in hard to staff schools.
• Washington DC, up to US$25,000. Far larger, and impacts on retention of effective teachers were not statistically significant.
• Talent Transfer Initiative, US$20,000 over two years. It moved retention while it was being paid, but "the difference was no longer statistically significant after payments stopped."

Note what that ordering does to the intuitive theory. The smallest bonus produced the clearest effect and the largest produced none. Size is not the variable that decides this.

What actually seems to decide it

• Whether the recipient can influence eligibility. The Tennessee split is the cleanest evidence on this page. Same money, same schools, different answer, and the difference tracks whether the rating was under the teacher's control.
• Whether people understand the offer. The North Carolina evaluation found widespread misunderstanding of the incentive and skepticism that the amount would be enough. An incentive nobody has correctly understood is not being tested.
• Whether you can afford it permanently. The transfer-incentive result is the one most likely to catch an employer out. Paying for retention buys retention during the payment window and hands you the same problem afterwards, having reset expectations about what staying is worth.
• Whether it is administered competently. The Tennessee authors flag that "implementation concerns, including the timing of application process and observed noncompliance in bonus distribution, present obstacles" both to the program working and to measuring it. Some of what looks like a failed incentive is a failed rollout.

The boundary that matters most

Nearly all of this evidence comes from K-12 public education. That is not a small caveat. Teacher labor markets have salary schedules, licensure, a compressed hiring calendar, and mission-driven motivation, and none of those transfer cleanly to a hospital system, a distribution center, or a software company. What generalizes is the shape of the finding, which is that conditional design beats size, and that effects tied to payments end with the payments. The specific magnitudes do not generalize, and you should distrust anyone who quotes them at you as though they do.

Read the grade correctly. Mixed does not mean it depends as a way of avoiding the question. It means the studies disagree in a patterned way, and the pattern tells you which conditions to check before spending the money.

Size what a departure actually costs you before deciding what a retention payment is worth, or work through the retention diagnostic to see whether pay is the binding constraint in your situation.