Are golden ages predictable?
Here, a golden age means an institutional improvement that continues over time. We tested whether the index could forecast one. The proposed signal failed its test. A weaker pattern based on a country's starting level appeared in one period, but it did not continue after 2012.
The registered prediction failed
Before examining the second dataset, we recorded a specific claim in Git: a sudden improvement in control of corruption would predict a sustained institutional climb. The pattern appeared among 89 countries used to develop the idea, but not among the 112 countries reserved for testing it.
| Set | Sustained climb | Overall rate | z score | Events |
|---|---|---|---|---|
| Discovery (89) | 31% | 18% | +3.0 | 87 |
| Reserved test set (112) | 17% | 17% | −0.0 | 137 |
Across 137 events in the reserved test data, countries with the proposed signal improved at the same rate as the full group. The original pattern was concentrated around post-communist transitions and EU accession, so it did not generalize to other settings. No other institutional component produced the proposed effect either.
The z score shows how far each result is from the overall rate after accounting for sample size. A value near zero means no detectable difference.
What appeared in the reserved data
Two other relationships appeared in the reserved data. Neither was based on the country's recent institutional movement.
Sustained climbs also clustered in time, especially from 2002 to 2005. The measured clustering was about twice the level expected from a random timing model. This suggests that wider historical conditions mattered, although this study did not isolate a single cause.
The low-starting-level pattern appeared when testing a different set of countries, but it did not appear when testing a later period. It was concentrated in the 2002 to 2011 window:
| Window | Low-start climb | High-start climb | z score |
|---|---|---|---|
| Discovery 2002 to 2011 | 15% | 6% | +6.2 |
| Reserved test 2012 to 2019 | 5% | 6% | −0.6 |
Since 2012, countries starting at a low level have improved no more often than countries starting at a high level. Resource-rich countries also showed no special boost, so a simple commodity-cycle explanation was rejected. EU enlargement, debt relief, and post-Cold-War reforms are possible explanations for the earlier period, but none has been isolated.
What the study currently supports
The study does not identify a reliable current predictor of a sustained institutional climb. Recent movement failed the registered test, and the low-starting-level pattern did not persist after 2012. The most defensible result is that the 2002 to 2011 period was different, even though the cause of that difference remains uncertain.
A low starting value also has more room to rise, so part of the earlier pattern may be simple movement back toward the average. The possible historical explanations remain candidates, not confirmed causes.
The failed forecast remains published because it sets a clear limit on what the project can claim.