Universities Departments Methods
Fourteen international ranking systems, pooled with a dynamic Bayesian latent-trait model. Each ranking is treated as one noisy, censored instrument reading a single underlying quantity, rather than as an answer in itself.
Monash is pinned to every comparison. Add others to compare against it. The shaded band is the 90% credible interval; it widens in years when fewer rankings listed the institution, which is the honest way to show that we know less about those years. Dots mark years with at least one listing.
Position on the pooled scale each year, and the published ranks the estimate was built from.
How many of each country's institutions sit in the top 200 of the pooled scale — not of any one published table.
Reliability is α²/(α²+σ²): the share of a ranking's variation that the shared factor explains. A low value means that ranking is either noisy or is measuring something the others are not — which makes it the more informative one to have.
List length by ranking and reference year. Blank means no edition could be retrieved — the model treats those as missing, not as zero. THE, U.S. News and post-2013 QS editions are shifted back one year so a column refers to one real-world moment.