Bivariate and multivariate scaled association models. An application to homogamy of social origin and education in Hungary between 1930 and 1979