Heteroskedasticity vs Homoskedasticity: What You Need to Know

🔒 More advanced videos with Datasets (Excel) + Stata Do-files available as Members-Only - Join P.E.R.S here:    / @cruncheconometrix   P.E.R.S Full List (92 videos, 15 Courses):    • Members-only videos   📊 WELCOME TO CRUNCHECONOMETRIX! 📊 If you are writing your MSc or PhD dissertation and your supervisor flagged issues with error variances, this session is for you. We explore why heteroskedasticity happens, why standard OLS regression assumptions fail when variance is not constant, and how these errors compromise the validity of your p-values and t-statistics. Understanding Heteroskedasticity Error Variances GLS WLS OLS Homoscedasticity Tutorial for MSc Dissertation | UK, US, Canada Are you writing your MSc Economics, Finance, Banking or PhD dissertation and your supervisor said you need to understand heteroskedasticity - error variances - GLS, WLS, OLS, homoscedasticity - what is heteroskedasticity and how to understand? You are in the right place. WHAT YOU WILL LEARN: ✓ What is heteroskedasticity - heteroskedasticity means different or unequal spread or scatter - error variances not constant - e.g., variance of error term varies across observations - e.g., variance of earnings error larger for high education than low education - or variance of firm growth error larger for large firms than small firms - heteroskedasticity - vs homoskedasticity means equal spread - error variances constant - e.g., variance of error same across all observations - homoskedasticity - classical assumption of OLS - OLS assumes homoskedasticity - if heteroskedasticity exists, OLS still unbiased but inefficient and standard errors biased - t-stats invalid ✓ Why heteroskedasticity matters - if heteroskedasticity exists and you use OLS without correcting, OLS standard errors biased - t-stats and p-values invalid - you may conclude variable significant when not or not significant when significant - Type I and Type II errors - need to test for heteroskedasticity and correct if exists - using robust SE or GLS/WLS - UK/US examiners check heteroskedasticity test - need to report ✓ Common mistakes MSc students make: Not testing heteroskedasticity at all - examiners check heteroskedasticity test - need to report Breusch-Pagan or White test, Not correcting heteroskedasticity when exists - if heteroskedasticity exists and you do not correct, SE biased - t-stats invalid - need robust SE or GLS/WLS, Interpreting GLS/WLS coefficients as different from OLS - GLS/WLS coefficients should be similar to OLS if heteroskedasticity not severe - but SE smaller - more efficient - if coefficients very different, check - maybe model misspecified, Confusing heteroskedasticity with multicollinearity or autocorrelation - heteroskedasticity is non-constant error variance - variance of error varies - multicollinearity is correlation among X's, autocorrelation is correlation among errors over time WHO THIS IS FOR: MSc, MBA, PhD Economics, Finance, Banking, Management, Business Analytics students in UK (Warwick, Manchester, Leeds, Birmingham, Glasgow, Edinburgh, LSE), US, Canada, Australia, EU with cross-sectional or panel data - heteroskedasticity common in cross-sectional - e.g., earnings, firm size. OFFICIAL LINKS & COMMUNITY: 🌐 Website: https://cruncheconometrix.com 🛒 Data Shop: https://cruncheconometrix.com/view/da... 🎓 Members-Only:    • Members-only videos   ⭐ Join Membership:    / @cruncheconometrix   👥 Facebook:   / cruncheconometrix   💼 LinkedIn:   / cruncheconometrix   🔔 Subscribe: https://www.youtube.com/c/CrunchEcono... If this helped, LIKE, COMMENT your heteroskedasticity test p-value and country, SHARE. FAQ: Q: How to test heteroskedasticity in Stata? A: After reg Y X, use hettest - Breusch-Pagan test - null=homoskedasticity - if p below 0.05 heteroskedasticity exists - or imtest, white - White test - more general - if p below 0.05 heteroskedasticity - plot residuals vs fitted - if spread increases with fitted heteroskedasticity. Q: How to correct heteroskedasticity? A: Use robust SE - reg Y X, robust - White robust SE corrects SE for heteroskedasticity - simplest - recommended for MSc - or WLS/GLS if form of heteroskedasticity known - e.g., variance proportional to X - weight by 1/X - more efficient - but robust SE often sufficient. 0:00 Understanding Heteroscedasticity: Nature and Causes 1:27 Homoscedasticity vs Heteroscedasticity 4:41 Root Causes of Heteroscedasticity 6:36 Consequences of Heteroscedasticity on OLS #Heteroskedasticity #ErrorVariances #GLS #WLS #OLS #Homoscedasticity #Heteroscedasticity #BreuschPagan #WhiteTest #RobustStandardErrors #Econometrics #Stata #EViews #EconometricsTutorial #MScDissertation #UKUniversities #HeteroskedasticityTest #PhDResearch #DissertationHelp #PERBA #CrunchEconometrix