The time series regression assumptions (TS.1 to TS.6)
The six time-series OLS assumptions, why TS.1 to TS.5 make OLS BLUE, and how stationarity and weak dependence keep it consistent.
A 5 minute 8 second animated lesson on the six classical time-series OLS assumptions. Built for ECON3006 Economic & Financial Modelling at Western Sydney University.
The video takes each assumption in turn: TS.1 linearity with finite distributed lags, TS.2 no perfect collinearity, TS.3 strict exogeneity (the most demanding, since it rules out feedback from past errors to future regressors), TS.4 homoskedasticity over time, TS.5 no serial correlation, and TS.6 normality. TS.1 through TS.5 are the Gauss-Markov conditions that make OLS the best linear unbiased estimator; TS.6 is a separate assumption that buys exact t and F inference in finite samples.
It then explains why OLS still works on realistic time-series data through stationarity and weak dependence, which let the law of large numbers and the central limit theorem apply, and contrasts that with a unit root whose shocks never die out. Pair it with the Atlas concept page for the formulas, a worked Stata example, and a quick quiz.