¹ Morris, Cynthia Taft and Adelman, Irma, Where Angels Fear to Tread: Quantitative Studies in History and Development (Stanford, California, forthcoming).Google Scholar

² Our selection and definition of variables and our typology are nevertheless heavily dependent on this literature–a literature rich in “partial” hypotheses about institutional influences on poverty with a few “grand” theories. Because of space constraints, we omit our discussion of this literature.Google Scholar

³ John Coatsworth's comments remind us to explain that the set of variables in this paper differs in two respects from our earlier published sets: we have added foreign economic dependence, export expansion, shift in export structure, changes in real agricultural wages, and changes in real industrial wages. Since excessive proliferation of variables appears to reduce the stability of components results, we have offset these additions by forming three composite indicators of market institutional development out of the previous 12 market variables. Turkey was omitted when we could not classify it on the new variables.Google Scholar

⁴ Brief descriptions of most of the classification systems may be found in Adelman, Irma and Morris, Cynthia Taft, “Patterns of Industrialization in the Nineteenth and Early Twentieth Centuries: A Cross-Sectional Quantitative Study,” Research in Economic History, 5 (1980), 1–83.Google Scholar

⁵ John Coatsworth's comments remind us of a more important reason than those mentioned in the text for unclassified countries. Our experience with the disjoint technique has indicated that a minimum of three countries (yielding nine observations) is required in a class in order to obtain reliable results of general interest. Among the unclassified countries, the courses of poverty vere too diverse either to form a group of three having reasonably similar courses of poverty or to add any of these countries to Classes One through Four.Google Scholar

⁶ Wold, Svante, “Pattern Recognition by Means of Disjoint Principal Components Models,” Pattern Recognition, 8 (1976), 127–137.CrossRefGoogle Scholar

⁷ Because of space constraints we discuss only (I) in this paper. (See footnote 12.)Google Scholar

⁸ Because the samples are not random, conventional tests of differences among samples are not applicable.Google Scholar

⁹ The inclusion of the land variables shows higher levels and rates of economic change associated with the predominance of cultivator-owned holdings rather than tenant holdings and with the predominance of neither very large nor very small holdings. The inclusion of the classification by socioeconomic character of political leadership shows higher levels and rates of economic change associated with greater political influence in national leaderships of rising industrial, commercial, and working classes. We limit our discussion to variables with weights or “loadings” rounding to at least. 17. This ad hoc cut-off was selected on the basis of sensible interpretations across all six chapters of final results for our book; it is less restrictive than our earlier cut-off based on only a few sets of results. We simplify the discussion in the text by not distinguishing between primary and secondary high loadings. Comparison of the whole set of our earlier published results with the current final set indicates substantial stability in the composition of components, allowing for differences due to our new variables; however, the division of variables between primary and secondary high loadings is less stable. John Coatsworth's comments remind us to mention that the choice of number of components is based on the sensibleness and theoretical interest of associations in additional components. We think the presence of three rather than two interesting components in the present results has occurred because of the new variables we added.Google Scholar

¹⁰ In the interpretation of a component as a process of change, the signs may be reversed for convenience since it is only the relationship among signs within a component that matters. The continuum for “net immigration” extends from high net immigration at the top of the scale to high net emigration at the bottom of the scale. Since all Scandinavian countries were at the lower end of the scale, then a negative association between agricultural wage movements and net immigration means faster wage increases with more net emigration.Google Scholar

¹¹ The indicator of foreign economic dependence is a composite of seven dimensions of foreign dependence with heavier weights for those listed first: foreign control of trade and distribution, foreign ownership and control of modern industry, dependence on foreign skills, locus of entrepreneurial initiative, share of foreign capital in domestic investment, reliance of governments on foreign loans, and dependence of production structure on trade. The freeing of land from communal restrictions is a major element of our composite measure of the development of market systems at the lower end of the spectrum.Google Scholar

¹² Because of space constraints, we have omitted the section in our paper on measures of statistical fit (including Table 6). In pair-wise comparisons among Classes One, Three, and Four, the own-class variance of each class is substantially lower than the variance obtained by fitting each class's observations to the components models of the other two classes (a comparison indicative of the distance between pairs of classes). Similar calculations for Class Two, however, show it to be fairly close to both Classes One and Four. This reflects the similarity of the Scandinavian primary pattern to that of Class One and the similarity of its secondary patterns to those of Class Four.Google Scholar

¹³ Because of space constraints we have omitted Table 7 giving statistical measures of the importance of individual variables in “explaining” within-class variations and in discriminating between classes.Google Scholar