Choose your reading route
eduKateSG · Why Mathematics? · Omitted Variable Bias: Confounding Direction and Magnitude
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 5–9
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 10–15
Check assumptions
See where design and calibration matter.Route 4 · Chapters 16–20
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 21–35
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important when leaving out a relevant variable can reverse or magnify a regression coefficient? The omitted-variable-bias formula separates the target relationship from the omitted variable’s effect and its association with the included predictor.
Did You Know? Bias direction depends on two signs: how the omitted variable affects the outcome and how it relates to the included predictor.
Full chapter index
Chapters 1–4 · Understand the question
Chapters 5–9 · Reproduce the example
Chapters 10–15 · Check assumptions
Chapters 16–20 · Explore applications
Chapters 21–35 · Study and verify
- A spreadsheet workflow
- A coding workflow
- Applications and careers
- A student investigation
- Common misconceptions
- Connection to recent mathematics articles
- Guidance for students and parents
- Practice questions
- Frequently asked questions
- A reporting checklist
- A sign table before calculation
- Sensitivity rather than false certainty
- Why adding every variable can backfire
- Final bias audit
- Useful next reading
A short regression can mix the target coefficient with pathways carried by an omitted variable.
MIT summarises the omitted-variable formula as short equals long plus the omitted effect times the regression of omitted on included. Read MIT OCW.
If y=alpha+beta x+gamma z+e and z=pi x+v, omitting z gives short slope beta_tilde=beta+gamma pi under the linear projection setup.
| beta | gamma | pi |
|---|---|---|
| 2 | 3 | 0.5 |
Let beta=2, gamma=3 and pi=0.5. The short-regression slope is 2+3×0.5=3.5, an upward bias of 1.5.
| Bias | Short slope |
|---|---|
| 1.5 | 3.5 |
Positive gamma and pi produce upward bias; opposite signs produce downward bias. Either can move an estimate toward, away from or across zero.
Omitting z biases the x slope only when z affects y conditional on x and z is associated with x in the relevant projection.
If pi=0, the scalar bias term is zero, though precision or other coefficients may still change.
A noisy proxy may reduce some confounding but leave residual bias. Measurement error requires its own model.
An omitted nonlinear term or interaction can create bias even when a simple pairwise correlation appears small.
With several omissions, the bias is a sum or vector product of omitted effects and auxiliary-regression coefficients.
A causal graph helps distinguish confounders from mediators and colliders. Including every available variable is not automatically safe.
Conditioning on a variable caused by the treatment or on a collider can create bias rather than remove it.
Because omitted quantities are unknown, vary plausible gamma and pi values and show how the target conclusion changes.
Random assignment can break association between treatment and pre-treatment confounders in expectation, reducing the classic bias pathway.
A variable may improve prediction without identifying a causal effect. Conversely, a causal design may focus on an estimand rather than maximum predictive fit.
List plausible omitted causes, timing, measurement quality, expected signs and whether controls belong in the estimand.
A study-time coefficient may absorb prior attainment if prior attainment affects marks and is related to study time.
State the short-model slope, the plausible bias direction and the assumptions needed to interpret the adjusted coefficient.
The formula does not reveal unknown effects from data alone, guarantee a chosen control set or replace study design.
Students can practise sign reasoning before numbers, then use the formula to test whether a story is algebraically coherent.
Write the long and auxiliary regressions, calculate gamma pi and add it to beta; then run sign and sensitivity checks. Keep every matrix dimension, unit and intermediate quantity visible, then verify the defining identity.
Use tested linear-algebra routines, save the design and model formula, check rank and conditioning, and test against a hand-worked example.
Omitted-variable reasoning matters in economics, epidemiology, education research, policy, business analytics and AI fairness audits.
Simulate y from x and z, compare short and long slopes, then vary gamma and pi.
A high R-squared, many controls or a significant coefficient does not prove omitted-variable bias is absent.
The FWL guide explains adjusted slopes. The R-squared guide explains why fit is not causality.
Students should draw the sign pathway and identify timing. Parents can ask what plausible causes were not measured.
- Reverse one sign.
- Set pi to zero.
- Explain why a collider is not a harmless control.
### Can the bias reverse sign? Yes.
### Do many controls solve it? Not automatically.
### Is the formula causal by itself? No.
Report the target, model, data unit, assumptions, calculation, diagnostics, validation and practical threshold. Separate descriptive evidence from causal interpretation.
Write the sign of gamma, the omitted variable’s outcome effect, and pi, its association with x. Same signs imply upward bias; opposite signs imply downward bias. Then ask whether “upward” means farther from or closer to the scientifically relevant value. If beta is negative, an upward bias can make the estimate less negative and may even reverse its sign. A sign table prevents loose verbal stories.
The omitted effect and auxiliary slope are often unknown. Choose plausible ranges from subject knowledge or external evidence and calculate a grid of beta_tilde−gamma pi. Mark where the substantive conclusion changes. This does not identify the true bias, but it reveals how strong the hidden relationship must be to overturn the result. Preserve assumptions beside every scenario.
A pre-treatment common cause can reduce confounding, but a mediator can remove part of the effect being studied and a collider can create a spurious association. Measurement timing and causal structure matter more than the number of controls. Use a causal diagram to define the estimand, then apply regression and the bias formula consistently.
List each plausible omitted variable, its expected outcome effect, its association with the included predictor and the evidence behind those signs. Calculate sensitivity ranges instead of one speculative correction. Explain which variables are confounders, mediators or colliders for the chosen estimand. This record shows where mathematics ends and subject-matter judgement begins. This matters.
Continue through the Mathematics Learning Hub and reproduce the worked example before using software.
Extended practice laboratory
Practice case 1: Write long model
Setup: Start from the article’s omitted-variable-bias worked example for case 1. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 2: Write auxiliary model
Setup: Start from the article’s omitted-variable-bias worked example for case 2. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 3: Record beta
Setup: Start from the article’s omitted-variable-bias worked example for case 3. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 4: Record gamma
Setup: Start from the article’s omitted-variable-bias worked example for case 4. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 5: Record pi
Setup: Start from the article’s omitted-variable-bias worked example for case 5. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 6: Multiply bias
Setup: Start from the article’s omitted-variable-bias worked example for case 6. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 7: Find short slope
Setup: Start from the article’s omitted-variable-bias worked example for case 7. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 8: Reverse gamma
Setup: Start from the article’s omitted-variable-bias worked example for case 8. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 9: Reverse pi
Setup: Start from the article’s omitted-variable-bias worked example for case 9. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 10: Set pi zero
Setup: Start from the article’s omitted-variable-bias worked example for case 10. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 11: Add second omitted
Setup: Start from the article’s omitted-variable-bias worked example for case 11. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 12: Add proxy
Setup: Start from the article’s omitted-variable-bias worked example for case 12. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 13: Add nonlinearity
Setup: Start from the article’s omitted-variable-bias worked example for case 13. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 14: Draw DAG
Setup: Start from the article’s omitted-variable-bias worked example for case 14. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 15: Check mediator
Setup: Start from the article’s omitted-variable-bias worked example for case 15. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 16: Check collider
Setup: Start from the article’s omitted-variable-bias worked example for case 16. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 17: Audit timing
Setup: Start from the article’s omitted-variable-bias worked example for case 17. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 18: Run sensitivity
Setup: Start from the article’s omitted-variable-bias worked example for case 18. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 19: Compare short long
Setup: Start from the article’s omitted-variable-bias worked example for case 19. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 20: Inspect design
Setup: Start from the article’s omitted-variable-bias worked example for case 20. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 21: Check randomisation
Setup: Start from the article’s omitted-variable-bias worked example for case 21. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 22: Reject fit claim
Setup: Start from the article’s omitted-variable-bias worked example for case 22. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 23: Reject significance claim
Setup: Start from the article’s omitted-variable-bias worked example for case 23. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 24: Report uncertainty
Setup: Start from the article’s omitted-variable-bias worked example for case 24. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 25: Write conclusion
Setup: Start from the article’s omitted-variable-bias worked example for case 25. Task: Change the named quantity, recompute dependent values and compare with the original.
Check: Check the defining identity, preserve units and state one diagnostic or validation step before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.