In the United States, past income helps predict air travel at some lags and past air travel never helps predict income; a stricter version of the test agrees. Both are tests of prediction only.
What this shows: The chart plots Aviation → GDP and GDP → Aviation on a vertical axis of P-value (below 0.05 = significant).
Conditions and method
Status: confirmed (estimated in a small bivariate system; omitted macro drivers could alter the result).
Each bar tests whether the past of one series helps predict the other at that lag, on the US quarters before the pandemic: below 0.05, it does. A stricter version of the test, valid on trending series, has GDP significant at all 5 lag settings tried and air travel at none. These tests show prediction, not cause.
The cross-market panel chart tests the same US quarters with a different test at a single lag, so its reading of the United States can differ from this chart’s. That is a difference of test and lag choice, not a contradictory finding.