Here's a question that matters for every SaaS founder building onboarding sequences: does signing up for your product give users permission to disengage?
The concept is called moral licensing — the idea that doing something "good" (like signing up for a productive tool) creates psychological permission to slack off afterward. If true, the very act of signing up for your free trial might be the reason users never complete onboarding.
But the latest meta-analytic evidence tells a more nuanced — and ultimately more useful — story.
The Original Finding: d = 0.31
Blanken, van de Ven, and Zeelenberg's 2015 meta-analysis in Personality and Social Psychology Bulletin [1] analyzed 91 studies with 7,397 participants. Their estimate: after behaving morally, people are more likely to behave immorally, with an effect size of d = 0.31 (small-to-medium).
Two mechanisms were proposed:
- Moral credentials: "I proved I'm a good person, so this next thing won't define me"
- Moral credits: "I've earned goodness points, now I can spend them"
Applied to SaaS: signing up establishes credentials ("I'm the kind of person who invests in systems"), which theoretically licenses disengagement ("I've done the hard part — onboarding can wait").
The Bias Correction: d = -0.05 to 0.18
Then Kuper and Bott (2019) [2] applied publication bias correction methods to the same dataset. The results were devastating:
| Method | Corrected Effect Size | Significant? | |--------|----------------------|--------------| | PET-PEESE | d = -0.05 | No (p = .64) | | Three-parameter selection model | d = 0.18 | Yes (p = .002) |
After bias correction, the effect either disappears entirely or shrinks to d = 0.18 — barely above the threshold for "small." The authors concluded that "both the evidence for and the size of moral licensing effects has likely been inflated by publication bias."
The Replication Failure: d = 0.02 to 0.08
The foundational study by Monin and Miller (2001) — which launched the entire moral credentials literature — was subjected to a registered report replication with 932 US participants [3].
Results:
- Race scenario: d = 0.02 to 0.08 (essentially zero)
- Gender scenario: d = -0.50 to -0.38 (opposite direction!)
No consistent moral credential effect was found. The most-cited study in the field didn't replicate.
The 2025 Resolution: It's About Being Watched
This is where it gets genuinely interesting.
Rotella, Jung, Chinn, and Barclay (2025) [4] conducted a multi-level meta-analysis of 115 experiments with 21,770 participants and used Robust Bayesian Meta-Analysis (RoBMA) to account for publication bias.
Their pre-registered prediction: moral licensing should be stronger when the initial moral behavior is observed by others.
The results:
| Condition | Hedges' g (uncorrected) | g (bias-corrected) | Bayes Factor | |-----------|------------------------|---------------------|--------------| | Observed (others saw moral act) | 0.65 | 0.51 | BF = 9.14 (moderate evidence FOR licensing) | | Unobserved (private moral act) | 0.13 | -0.01 | BF = 0.11 (moderate evidence AGAINST licensing) |
When nobody is watching, moral licensing essentially does not exist (g = -0.01, with the Bayesian analysis providing moderate evidence against the effect).
When someone IS watching, the effect is substantial (g = 0.51) — because moral licensing is driven by reputation management, not by some internal moral thermostat.
What This Means for SaaS Onboarding
Online SaaS behavior is largely unobserved.
Nobody sees you sign up for a trial. Nobody sees you skip onboarding. Nobody knows you downloaded a guide and never opened it. And per Rotella et al.'s 115-experiment meta-analysis, moral licensing is a reputation phenomenon — it needs an audience.
This means:
1. Moral licensing is probably NOT causing your onboarding drop-off.
The effect is g = -0.01 in unobserved conditions. Your users aren't licensing themselves to disengage — they're facing cognitive load (Post #148), present bias (Post #157), and planning fallacy (Post #149). These are much larger effects.
2. Don't treat signup as an achievement anyway.
Even though licensing isn't the main driver, there's wisdom in not making signup feel like completion. Don't send "Congratulations on taking the first step!" emails. Send "Here's your first action" emails. The difference matters for framing, even if it's not about licensing specifically.
3. Be careful with public progress sharing.
If you add social features where users can show their community they signed up or completed Step 1, you may actually CREATE the licensing conditions that don't naturally exist online. Making moral behavior visible to others increases licensing (g = 0.51). Public accountability has benefits (commitment-consistency, social proof), but it also has this licensing cost.
4. Credentials vs. credits — design for credits.
Don't build identity too early ("You're a systems builder!"). This creates moral credentials that license coasting on the identity. Instead, build moral credits through actions — and immediately create opportunities to "spend" those credits on the next step. Credits deplete; credentials persist.
The Honest Assessment
After five meta-analyses and multiple replication failures:
| Analysis | Moral Licensing Effect | |----------|----------------------| | Blanken et al. 2015 (91 studies) | d = 0.31 | | Simbrunner & Schlegelmilch 2017 (106 studies) | d = 0.32 | | Kuper & Bott 2019 (bias-corrected) | d = -0.05 to 0.18 | | Ferguson et al. 2024 | g = 0.11 | | Rotella et al. 2025 (unobserved, bias-corrected) | g = -0.01 |
The trajectory is clear. As sample sizes increase and bias corrections improve, the effect shrinks. In the conditions most relevant to online SaaS (unobserved behavior), there's moderate Bayesian evidence that the effect doesn't exist at all.
Moral licensing is a reputation phenomenon wearing a self-concept costume. It requires an audience. Your SaaS onboarding doesn't have one.
The Nervous System Angle
For Fleshtimer readers, the observation finding connects to HRV research:
Social evaluation activates the hypothalamic-pituitary-adrenal axis — being observed triggers a cortisol response. The Trier Social Stress Test (the most reliable lab stress inducer) works precisely because of social evaluation threat.
When observed moral behavior triggers licensing, it may be because the social evaluation stress is "resolved" by the good deed — creating a physiological relief that permits relaxation. In unobserved conditions, no social evaluation stress was ever activated, so there's nothing to resolve.
This maps onto the HRV-cognition chain: observed moral behavior → social stress → cortisol → HRV dip → relief after good deed → HRV recovery → licensing as the subjective experience of that relief. No observation → no social stress → no relief → no licensing.
Sources
- [Blanken, I., van de Ven, N., & Zeelenberg, M. (2015). A Meta-Analytic Review of Moral Licensing. Personality and Social Psychology Bulletin, 41(4), 540-558.](https://journals.sagepub.com/doi/10.1177/0146167215572134)
- [Kuper, N. & Bott, A. (2019). Has the evidence for moral licensing been inflated by publication bias? Meta-Psychology, 3.](https://open.lnu.se/index.php/metapsychology/article/view/878)
- [Xiao, Q. et al. Licensing via Credentials: Replication Registered Report of Monin and Miller (2001). International Review of Social Psychology.](https://rips-irsp.com/articles/10.5334/irsp.945)
- [Rotella, A., Jung, J., Chinn, C., & Barclay, P. (2025). Observation Moderates the Moral Licensing Effect. Personality and Social Psychology Bulletin.](https://journals.sagepub.com/doi/10.1177/01461672251345512)
- [Ferguson et al. (2024). Moral Licensing and Foot-in-the-Door Meta-Analysis.]
- [Simbrunner, P. & Schlegelmilch, B. (2017). Moral licensing: a culture-moderated meta-analysis. Management Review Quarterly, 67(4).](https://link.springer.com/article/10.1007/s11301-017-0128-0)