How can some countries register large declines in their rankings, but these declines are not statistically significant?
When comparing country scores over time, or with other countries, due regard must be paid to margins of error. A simple rule of thumb is that when the indicated confidence intervals do not overlap for the two comparators, the change is not statistically significant. For some countries, particularly where data is scarce and our estimates are based on very few data sources, these confidence intervals can be quite large, properly indicating the uncertainty about governance in these countries. For such countries, even apparently large changes may not be statistically significant, simply because there is a great deal of transparently-noted imprecision in the data. A more detailed discussion of confidence intervals, standard errors and changes over time can be found in the Governance Matters III paper (pages 14-19, see also figure 3 on pages 49-51) and the Governance Matters IV paper (pages 15-26).
Related Questions
- How can some countries register large declines in their rankings, but these declines are not statistically significant?
- Is the relationship between the Great Lakes ice cover and climate patterns statistically significant?
- What happens if I refer someone who purchases an ad but declines to register/paricipate in the RRP?
