08/06/2026
U is for “Unit Roots”
Some concepts in economics are less “flashy insight” and more “silent gatekeepers of credibility.”
Unit roots belong firmly to that second category.
At first glance, a time series with a unit root can look completely harmless—smooth trends, plausible movements, visually intuitive patterns. But statistically, it is doing something far more problematic: it refuses to “forget” the past.
In more technical terms, a unit root process is non-stationary. Shocks don’t fade away—they accumulate. A policy change in 2010 can still be influencing the series in 2020, not because of structural persistence, but because the data-generating process never properly stabilises.
Why does this matter?
Because most standard econometric tools assume stationarity. Regressing one non-stationary series on another can produce spurious relationships—beautiful t-stats, convincing R², and entirely misleading conclusions.
This is why testing (ADF, Phillips–Perron) and transformation (differencing, cointegration frameworks) are not just technical rituals—they are safeguards against being fooled by trends that only look meaningful.
So “U” is a quiet reminder: before interpreting relationships, ask whether the data is even speaking in a stable language.
Because sometimes the most convincing results are the ones that shouldn’t exist at all.