This article shows how the difference between the observed frequencies of accounting policy choice and the outcome of a random policy choice, where each available method has an equal chance of being selected, may be fully explained with a statistical model. The process of harmonization is described in a way that identifies departures from equiprobable accounting policy choice as either: (a) the systematic effects of harmonization, or (b) the effects of systematic divergence from international harmony where the frequency of adoption of differing accounting methods varies across countries, or (c) the effects of company‐specific accounting policy choices.The understanding ofharmonythat underlies previous attempts to measure harmonization is such that, with respect to a particular financial statement item, a situation of maximum harmony is reached when all companies in all countries use the same accounting method. From the standpoint of modelling the harmonization process. however, a different concept of harmony may be more useful. In this article, therefore, we posit a state ofdistributional harnzonyin which, other things being equal, the expected distribution of accounting policy choices is the same in each country. In this theoretical state. the odds of selecting a given accounting method from those available for a particular financial statement item are identical for each country. A major advantage of this benchmark is that it provides a basis for distinguishing between two possibly conflicting components of the international harmonization process: between‐country harmonization and within‐country standardization.A hierarchy of nested statistical models is then used to describe accounting policy choices made by companies with an international shareholding and registered in Europe, where the European Union has been involved in a program of accounting harmonization. The accounting policies analysed in depth in this article comprise the treatment of goodwill and accounting for deferred taxation. The results are compared with the comparability index method used previously in harmonization research st