# What Is the Hawthorne Effect? Definition and Examples

The Hawthorne effect is the tendency of people to change their behavior when they know they are being observed or singled out for attention. It is named after a series of studies at the Western Electric Hawthorne plant in the late 1920s and early 1930s, where productivity rose even when working conditions got worse. The effect matters because it can inflate results in experiments, workplace studies, and quality-improvement programs, making a treatment look more effective than it really is.

## Quick Answer

- The Hawthorne effect is a change in behavior caused by the awareness of being observed, not by the treatment itself.
- It was named after studies at the Hawthorne Western Electric plant, where output rose when lighting was made brighter or dimmer [1].
- Elton Mayo and colleagues ran experiments from 1927 to 1932 on breaks, work hours, and incentives, and found that special attention motivated workers [1].
- The effect biases research by adding an observation component to the measured outcome, so the treatment effect is overestimated [2].
- You can spot it in data when performance jumps during observation and returns to baseline afterward, as in the worked example below.

## What the Hawthorne Effect Means

In plain terms, the Hawthorne effect is the change in people's behavior that happens because they know someone is watching. Workers who feel singled out for special attention tend to perform better, regardless of whether the change they are experiencing is actually an improvement [1].

The precise statistical definition is narrower. The Hawthorne effect is the tendency of study participation itself to affect the outcome being measured [2]. In a formal experiment, you compare a treatment group against a control group to isolate the effect of the treatment. The Hawthorne effect adds a second influence: being in the study at all. If both groups are observed, both may improve, which shrinks the measured difference between them. If only the treatment group is observed, the improvement from attention gets mixed into the treatment effect and inflates it.

This is why the effect is often described as a threat to internal validity. The outcome you measure is not purely the response to your intervention. It also contains a response to measurement.

## How It Works

There is no single formula for the Hawthorne effect, because it is a behavioral bias, not a fixed quantity. But you can model it as an extra term in the outcome equation.

$$Y_i = \beta_0 + \beta_1 T_i + \beta_2 O_i + \varepsilon_i$$

- $Y_i$ is the measured outcome for person $i$, such as units produced per week.
- $T_i$ is the treatment indicator, equal to 1 if person $i$ received the intervention and 0 otherwise.
- $\beta_1$ is the true treatment effect you want to estimate.
- $O_i$ is the observation indicator, equal to 1 when person $i$ knows they are being observed and 0 otherwise.
- $\beta_2$ is the Hawthorne effect, the shift in outcome caused by awareness of observation.
- $\varepsilon_i$ is random error.

If you omit $O_i$ from the model, its influence gets absorbed into $\beta_1$ whenever observation and treatment overlap. Your estimate of the treatment effect is then biased upward by roughly $\beta_2$. The practical fix is to observe the control group in the same way you observe the treatment group, so $O_i$ is the same for everyone and cancels out of the comparison.

## Worked Example

This example uses a small dataset of weekly productivity in units per worker across 12 weeks: 4 weeks before observation, 4 weeks during observation, and 4 weeks after.

| Week | Phase | Output |
|---|---|---|
| W1 | Before | 100 |
| W2 | Before | 102 |
| W3 | Before | 98 |
| W4 | Before | 100 |
| W5 | During | 115 |
| W6 | During | 118 |
| W7 | During | 116 |
| W8 | During | 117 |
| W9 | After | 101 |
| W10 | After | 99 |
| W11 | After | 103 |
| W12 | After | 100 |

Step 1. Compute the mean for each phase.

- Before-observation mean: $(100+102+98+100) / 4 = 100.0000$
- During-observation mean: $(115+118+116+117) / 4 = 116.5000$
- After-observation mean: $(101+99+103+100) / 4 = 100.7500$

Step 2. Compute the percent change against the before phase.

- During vs before: $(116.5000 - 100.0000) / 100.0000 \times 100 = 16.5000\%$
- After vs before: $(100.7500 - 100.0000) / 100.0000 \times 100 = 0.7500\%$

Step 3. Test whether the during-phase jump is larger than noise. A Welch t-test on before versus during gives a t-statistic of 15.8527 with 5.6966 degrees of freedom.

```python
import statistics
before = [100, 102, 98, 100]
during = [115, 118, 116, 117]
mb = statistics.mean(before)
md = statistics.mean(during)
pct = (md - mb) / mb * 100
print(round(pct, 4))  # 16.5
```

Output:

```
16.5
```

The pattern is the signature of a Hawthorne effect. Output rises 16.5% while workers are being observed, then falls back to 100.75, only 0.75% above the original baseline, once observation stops. The treatment did not produce a lasting change. The attention did.

## How to Interpret It

Read the during-phase jump as a combination of two things: the real effect of whatever changed, and the effect of being watched. The after-phase return to baseline is the strongest clue. If performance holds at the higher level after observation ends, you likely measured a real improvement. If it drops back, the gain was tied to the observation itself.

The t-statistic of 15.8527 tells you the during-phase difference is large relative to the week-to-week variation in this dataset. It does not tell you the cause. Statistical significance confirms that a difference exists, not why it exists. You still need the study design to separate attention from treatment.

Use the after-phase mean as your reality check. A 0.75% residual is small enough to treat as normal variation around the original baseline of 100.

## When to Use It (and when not to)

Apply the concept when you design or evaluate any study where people know they are participants. This includes workplace productivity trials, clinical quality-improvement programs, educational interventions, and usability tests. In these settings, plan for the effect before you collect data.

The standard defenses are straightforward. Observe the control group in the same way as the treatment group. Add a baseline period before the intervention and a follow-up period after it. Keep observation as unobtrusive as your ethics rules allow. Where possible, blind participants to which condition they are in.

Do not use the Hawthorne effect as a catch-all explanation for any result you cannot explain. It is a specific mechanism, and claiming it without evidence is just a guess. It also does not apply to studies where participants cannot know they are being observed, such as analysis of existing records with no contact with subjects. In those cases, look for other sources of bias.

## Hawthorne Effect vs Demand Characteristics

These two ideas are related but distinct. The Hawthorne effect is about behavior changing because of awareness of observation. Demand characteristics are about participants guessing what the researcher wants and acting to confirm or deny that guess. Both can distort results, and both come from the participant's interpretation of the study situation.

| Feature | Hawthorne Effect | Demand Characteristics |
|---|---|---|
| Trigger | Knowing you are observed | Guessing the study hypothesis |
| Direction of change | Usually improved performance | Toward or against the expected result |
| Typical setting | Workplace, field studies | Lab experiments, surveys |
| Main defense | Observe control group equally | Blind participants, hide the hypothesis |
| Key reference | Study participation affects outcome [2] | Participant beliefs shape responses |

## Common Mistakes

- Treating any improvement during a study as a Hawthorne effect. Fix: check whether performance returns to baseline after observation ends before naming the cause.
- Observing only the treatment group. Fix: apply identical observation to the control group so the attention component cancels out.
- Skipping the baseline period. Fix: collect pre-intervention data so you have a reference point to compare against.
- Confusing the Hawthorne effect with the placebo effect. Fix: remember the placebo effect comes from expecting a treatment to work, while the Hawthorne effect comes from being watched.
- Assuming the effect is always positive. Fix: allow for the possibility that observation can also make people more anxious or more careful, which can lower output.
- Reporting a single before-and-after comparison as proof. Fix: use a control group and a follow-up period, and report the uncertainty around your estimate.

## Limitations

The Hawthorne effect is hard to measure directly because you cannot easily separate attention from treatment in a real setting. The original studies have also been reexamined over the years, and some researchers argue that the productivity gains had other explanations, such as feedback, learning, or the removal of workers who performed poorly. The label is widely used, but the underlying evidence is more contested than the popular story suggests.

The concept also cannot tell you the size of the bias in advance. It gives you a direction and a warning, not a correction factor. You cannot subtract a fixed percentage from your results and call it adjusted. The only reliable approach is design: control groups, baseline periods, and follow-up measurement. As one clinical review puts it, when a proof is required, you should design a randomized study instead of relying on internal quality assurance alone [2].

## Frequently Asked Questions

### What is the Hawthorne effect in simple terms?

It is the tendency of people to behave differently when they know they are being watched. In the original studies, workers increased their output when researchers paid attention to them, even when the changes to their working conditions were not improvements [1]. The attention itself, not the change, drove the result.

### Where did the Hawthorne effect come from?

It is named after the Hawthorne Western Electric plant, where engineers first tested whether brighter lighting would raise productivity. Output rose when the light was made brighter and also when it was made dimmer [1]. Elton Mayo and a team of Harvard researchers joined the investigation in 1927 and ran further experiments through 1932 [1].

### Does the Hawthorne effect always increase performance?

No. The classic studies showed increases, but the effect can push behavior in either direction. People who feel watched may work faster, or they may slow down and become more cautious. The consistent part is that behavior changes, not the direction of the change.

### How do you control for the Hawthorne effect?

Observe the control group in the same way as the treatment group, add a baseline period before the intervention, and measure again after observation ends. If both groups are watched equally, the attention component affects both and cancels out in the comparison. Blinding participants to the study hypothesis also helps.

### Is the Hawthorne effect the same as the placebo effect?

No. The placebo effect comes from a person's expectation that a treatment will help them. The Hawthorne effect comes from the awareness of being observed or singled out for attention [1]. A participant can show a Hawthorne effect with no treatment at all, simply because someone is measuring them.

## References

1. [9.3: The Hawthorne Studies - Business LibreTexts](https://biz.libretexts.org/Courses/Benedictine_University/Introduction_to_Business_and_Professional_Responsibility/09%3A_Motivating_Employees/9.03%3A_The_Hawthorne_Studies)
2. [Vahl CF, Osswald BR, Meinzer P, de Simone R, Thomas G, Hagl S. (1997). (Internal quality assurance or Hawthorne effect?). Langenbecks Archiv fur Chirurgie. Supplement. Kongressband. Deutsche Gesellschaft fur Chirurgie. Kongress](https://pubmed.ncbi.nlm.nih.gov/9574140/)

## Further Reading

- [10.3: The Hawthorne Studies - Business LibreTexts](https://biz.libretexts.org/Courses/HACC_Central_Pennsylvania's_Community_College/Introduction_to_Business_(Gifford)/10%3A_Motivating_Employees/10.03%3A_The_Hawthorne_Studies)
- [Krzywinski M, Altman N (2013). Importance of being uncertain. Nature Methods](https://doi.org/10.1038/nmeth.2613)
- [NIST/SEMATECH e-Handbook of Statistical Methods](https://www.itl.nist.gov/div898/handbook/index.htm)
- [Ioannidis JPA (2005). Why Most Published Research Findings Are False. PLoS Medicine](https://doi.org/10.1371/journal.pmed.0020124)

## Related Articles

- [Confirmatory Factor Analysis: Definition and Example](/blog/data-analysis/confirmatory-factor-analysis)
- [Interval Scale Questions: Examples and How to Use Them](/blog/data-analysis/interval-scale-questions-examples)
- [Generalizability in Research: Definition and Examples](/blog/data-analysis/generalizability-in-research)
- [Ordinal Survey Questions: Examples and How to Write Them](/blog/data-analysis/ordinal-survey-question-examples)
- [Dependent Variable Examples: Definition and Study Design](/blog/data-analysis/dependent-variable-examples)