Percentage agreement and phi: A conversion table.

Goldman et al. (1979) built a look-up table. You enter any percentage-agreement score. The table gives you the matching phi coefficient.

Phi is a chance-corrected number, like kappa. It lets readers compare reliability across studies even when base rates differ.

The table works for any agreement value from 0 to 100%. No math needed. One glance turns percent agreement into a statistic that accounts for chance.

Jones et al. (1977) came first. They gave a formula that also removes chance agreements. Goldman et al. (1979) make the idea usable by hiding the algebra in a table.

Friedling et al. (1979) disagree. They say the table is overkill. Their 10% disagreement rule is faster: if disagreements are 10% or less, chance correction is unnecessary.

The clash is only skin-deep. C et al. tested common, balanced data sets. When base rates get extreme, the phi table still saves you from inflated agreement.

Next time you write 95% agreement, add the phi from the table. Reviewers see you controlled for chance, and meta-analysts can fairly compare your numbers with other labs. It takes ten seconds and ends arguments about inflated reliability.