Skip to content

FRM Exam Part I · Regression Diagnostics

Multicollinearity in Regression: Detection and Remedies

Updated 11 October 2026 · Fact-checked

Multicollinearity means two or more regressors in a multiple regression are highly correlated. OLS stays unbiased, but standard errors rise, t-statistics fall and coefficients become unstable. Typical sign: high R-squared with insignificant t-stats. Detect it with correlations or VIF; fix it by dropping or combining variables, or collecting more data.

Understand Multicollinearity

In a multiple regression, each slope coefficient measures the effect of one regressor holding the others constant. That only works if each regressor carries some information the others do not.

Multicollinearity arises when two or more regressors are highly correlated, or when one is nearly a linear combination of the others. Then the model cannot tell which variable is responsible for the movement in Y. Think of two analysts who always give the same forecast: you cannot say whose view drove the result.

Perfect multicollinearity means an exact linear relationship, for example including a dummy for every category plus the intercept. OLS cannot be computed at all. Imperfect multicollinearity is the usual case: OLS still runs, and the estimators remain unbiased and consistent, but they become imprecise.

The main effects are large standard errors for the affected coefficients, small t-statistics, wide confidence intervals, and coefficients that change a lot when you add or drop a variable or a few observations. Signs can even look wrong. The classic symptom is a high R-squared and a significant F-test, yet individual t-tests are insignificant.

The model as a whole can still forecast well, because the combined effect of the correlated variables is well estimated. The trouble is interpreting individual coefficients.

Key formulas to remember

Variance Inflation Factor
VIF_j = 1 ÷ (1 − R_j²)
R_j² comes from regressing regressor j on all the other regressors (not on Y). VIF = 1 means no collinearity with the others.
Effect on slope variance
Var(β̂_j) = σ² ÷ [(1 − R_j²) × Σ(x_ij − x̄_j)²]
The standard error is inflated by √VIF compared with uncorrelated regressors. Standard form for multiple regression.
Standard error inflation
SE inflation factor = √VIF
A VIF of 9 means the standard error is 3 times as large as it would be with no collinearity.
Rule of thumb
VIF > 10 suggests serious multicollinearity
A common guideline, not a strict test. Some texts use lower cut-offs such as 5.
Two-regressor case
R_j² = r² (squared correlation between the two regressors)
With only two regressors, VIF = 1 ÷ (1 − r²) for both.
Classic symptom
High R² and significant F, but insignificant individual t-stats
Pairwise correlations can miss collinearity involving three or more variables, so VIF is more reliable.

How to solve Multicollinearity questions

Use this routine for any multicollinearity question, whether it asks you to identify, compute or fix.

  1. 1Read the clue: highly correlated regressors, high R² with insignificant t-stats, unstable coefficients, or a large VIF all point to multicollinearity.
  2. 2Decide whether it is perfect (OLS cannot be estimated) or imperfect (OLS works but is imprecise).
  3. 3If asked to compute VIF, find R_j² from the auxiliary regression of regressor j on the other regressors, then use VIF = 1 ÷ (1 − R_j²).
  4. 4If asked about the effect on standard errors, take √VIF as the inflation multiple.
  5. 5State what is unaffected: unbiasedness and consistency of the OLS estimators, and the overall fit and F-test.
  6. 6State what is affected: standard errors, t-statistics, confidence widths and coefficient stability.
  7. 7If asked for a remedy, pick from dropping a redundant variable, combining variables (for example an index or principal components), getting more data, or ridge-type regularization.
  8. 8Check the answer against the options for traps such as 'coefficients become biased'.

Quickest way: Fast VIF and symptom check

When to use it: Use for MCQs that give an R² or correlation and ask for VIF, or ask which statement about multicollinearity is true.

  1. For VIF: compute 1 − R_j², then take the reciprocal. Use the calculator 1/x key.
  2. For two regressors with correlation r: square r first, then VIF = 1 ÷ (1 − r²).
  3. For standard error change: take √VIF.
  4. For true/false: bias is NOT a consequence; high standard errors and unstable estimates ARE.
  5. If the stem says high R² and insignificant t-stats, answer multicollinearity before reading the options.

Common mistakes in Multicollinearity

  • Saying multicollinearity biases the OLS coefficients.

    Students link any regression problem with bad estimates.

    Fix: Remember imperfect multicollinearity leaves OLS unbiased and consistent. It only inflates variances.

  • Computing VIF using R² from the regression of Y on X.

    Confusing the model R² with the auxiliary R².

    Fix: R_j² comes from regressing X_j on the other regressors. Y is not involved.

  • Treating √VIF and VIF as the same thing.

    Both are called inflation factors.

    Fix: VIF inflates the variance. The standard error is inflated by √VIF.

  • Concluding the model is useless for forecasting.

    Insignificant t-stats look like a failed model.

    Fix: The overall fit and F-test remain fine, and forecasts can be good if collinearity persists in the future data. Only individual coefficient interpretation suffers.

  • Relying only on pairwise correlations to rule it out.

    Low pairwise correlations seem reassuring.

    Fix: One regressor can be a combination of several others. VIF catches this; pairwise correlations may not.

  • Treating VIF above 10 as a formal test with a critical value.

    Rules of thumb are memorised as laws.

    Fix: Call it a guideline. Also weigh the symptoms and the purpose of the model.

Worked examples

Example 1

In a multiple regression, X2 is regressed on the other regressors and the auxiliary R² is 0.90. Calculate the VIF for X2 and the factor by which its standard error is inflated.

Show the solution
  1. VIF = 1 ÷ (1 − R_j²) = 1 ÷ (1 − 0.90).
  2. 1 − 0.90 = 0.10, so VIF = 1 ÷ 0.10 = 10.
  3. Standard error inflation = √10 = 3.162.

Answer: VIF = 10; the standard error is about 3.16 times what it would be with no collinearity. This sits at the common rule-of-thumb threshold for serious multicollinearity.

Example 2

A regression of fund returns on three market factors gives R² = 0.93 and a significant F-statistic, yet none of the three slope t-statistics is significant. Dropping one factor changes the other coefficients sharply. Which is the most likely cause, and which statement about the OLS estimators is correct? A) Heteroskedasticity; estimators are biased. B) Multicollinearity; estimators remain unbiased but have large standard errors. C) Serial correlation; estimators are inconsistent. D) Omitted variable bias; standard errors are too small.

Show the solution
  1. High R² and significant F with insignificant individual t-stats is the classic symptom of multicollinearity.
  2. Unstable coefficients when a variable is dropped confirm it.
  3. Imperfect multicollinearity does not cause bias or inconsistency. It inflates standard errors.
  4. Only option B matches both points.

Answer: B. The factors are likely highly correlated; OLS stays unbiased but standard errors are inflated.

Exam tips

  • Memorise the symptom pair: high R² with insignificant t-stats. It is the most common clue in the stem.
  • Always check whether an option claims bias or inconsistency. For imperfect multicollinearity it is wrong.
  • Do VIF arithmetic in two steps: 1 − R², then reciprocal. Write the intermediate value to avoid slips.
  • Know the remedies and their cost: dropping a variable can cause omitted variable bias if it truly belongs in the model.
  • Distinguish clearly from heteroskedasticity and serial correlation, which affect standard errors for different reasons.

Practice questions from Regression Diagnostics

Multicollinearity: frequently asked questions

What is multicollinearity in simple terms?

It is when two or more regressors move together so closely that the regression cannot separate their individual effects. The model still fits, but the individual coefficients become imprecise and unstable.

How do I detect multicollinearity using VIF?

Regress each explanatory variable on the others and take its R². Then VIF = 1 ÷ (1 − R²). A VIF near 1 is fine, and values above about 10 are commonly treated as a warning sign.

Does multicollinearity bias regression coefficients?

No, not when it is imperfect. OLS estimators remain unbiased and consistent. What rises is their variance, so t-statistics fall and confidence intervals widen.

How do you fix multicollinearity?

Drop or combine redundant regressors, use an index or principal components, or gather more data. Regularization such as ridge regression can also stabilise estimates. If the model is only for prediction, you may choose to leave it alone.