One of the most critical issues in Argentina’s recent political history concerns the source of popular support for Juan Perón. His followers ardently maintain that El Líder won devotion from “the common people;” detractors insist that he hoodwinked and exploited gullible illiterates. Historical and sociological interpretations have emphasized Perón’s connection with “the urban working class” and “displaced migrants,” but these observations raise as many questions as they answer.1 What sort of urban community, working class, or migration? How about other social sectors? What happened in the rural areas?
Voting returns from the 1946 presidential election offer one promising means for attacking such problems. It was this election, of course, which marked the beginning of Perón’s decade-long dictatorship. Starting his climb to national prominence in 1943, Perón launched his “impossible candidacy” in late 1945 and overcame the combined opposition of all former national political parties—plus the indiscreet resistance of the U.S. State Department—to win a solid 54 percent majority. Most commentaries upon this extraordinary achievement have derived from eyewitness accounts, partisan statements, and studies of fragmentary data.2 The situation calls for systematic and dispassionate analysis.
In response to this need I shall examine statistical relationships between various socioeconomic factors and the vote for Juan Perón in 365 partidos, or counties, during the 1946 election.3 This approach of course has drawbacks. Socioeconomic matters do not always determine political behavior, although the absence of a hyper-efficient Peronist party “machine” suggests that socioeconomic considerations played a crucial role in this particular instance. Some causal connections do not lend themselves to quantification. Election results can be ambiguous. Not everyone who voted for Perón could be called a “Peronist,”4 and political attitudes among the Argentine electorate undoubtedly changed over time. Nevertheless the 1946 election, one of the most free and open in the country’s history, provides an excellent way of exploring the social base of popular support for Juan Perón at the start of his regime.
Operationally this analysis will seek out mathematical associations between the percentage of votes cast for Perón5 and socioeconomic characteristics of the counties as reported in the national census of 1947.6 I must emphasize that these relationships are ecological: they refer to the proportional Peronist vote in collective communities, not among a sample or population of human individuals. A 55 percent plurality in a county with 55 percent urbanization does not mean that all the city-dwellers came out for Perón. Strictly speaking, this study deals with the kinds of places, not the kinds of people, that furnished support for Perón. Without any usable survey data historians have little choice but to acknowledge and accept this limitation. But as I hope to show, ecological statistics have considerable intrinsic importance; if urbanization and the Peronist vote show the same fluctuations from one county to another there must be at least some connection between the two variables, and it is likely, if not certain, that the pattern reflects voting tendencies among the urbanites. Thus ecological data help establish the context and probable limits of individual behavior.7
The methodological characteristics of this inquiry should not encourage false illusions about the imminence of a radically “new” interpretation of electoral trends in Argentina. Sometimes this sort of investigation can produce totally unexpected results. Other times it confirms prevailing views. But even in the latter case statistical analysis can make useful contributions, for it provides a systematic approach to the problem, and can indicate not only whether specific variables (e.g., urbanization) help explain the Perón vote, but also by how much. We are bound to refine the conventional wisdom and explore its factual foundations, not to overthrow it.
Data for this analysis include 40 variables for the 365 counties, with some data missing for some counties (see Appendix 1 for names, definitions, and statistical characteristics of the variables. “Political” information refers to the Peronist vote and electoral participation. “Social” variables focus on the size, urbanization, literacy, ethnic composition, and age structure of the population. “Economic” data concern land usage and land tenure, employment, unemployment, wages, and composition of the work force in commerce and industry; figures on electric horsepower usage also give a loose indication of industrial “modernization.” Countless hypotheses about possible relationships between these variables and the Peronist vote come quickly to mind.8 Instead of making a long and formal fist of such propositions, however, I shall simply ask: Which variables best explain the Peronist vote, to what degree, and in what combination?
Initial results indicate that statistical correlations between the Peronist vote and the socioeconomic variables are uniformly low. On a scale ranging from +1 (strong positive association) to −1 (strong negative association), the highest correlation is with the percentage of internal migrants; only +.34.9 Horsepower per industry is next with +.31. Several of the others show no significant deviation from zero.10 (See Appendix 2 for a complete list of all these correlations). Despite possible error in data compilation, the meaning is obvious: no single ecological factor accounts for Perón’s electoral popularity to any large degree.11
This finding thoroughly obliterates all fanciful hopes for a simple one-factor socioeconomic explanation of the Perón vote. El Líder did not draw support merely from urban areas, working-class districts, counties with lots of young voters, internal migrants or unemployed men. Whatever else the Peronist movement might have been, it was socially complex.
In order to explore this problem, it seems reasonable to divide the counties into groups and then study correlations within each separate group. One such classification might be regional, since historical literature on Argentina abounds with reference to conflict between the City of Buenos Aires or the Coast and the Interior. Nevertheless I have elected to separate counties according to the size of largest community (see the CITY SIZE variable in Appendix 1). I shall employ three basic categories: counties with communities of 50,000 or more inhabitants, which I shall call Big Cities; those with communities of 2,000-49,999 inhabitants, to be known as Townships; and the Countryside, without any communities of 2,000 or more people.
This urban-rural approach is promising partly because of its precision. CITY SIZE says something about the county itself, while regional location often refers to little more than geographical accident. The strategy also makes intuitive sense since many analysts have stressed the role of “the city” in politics and in Argentine history.12 According to Table 1 regional and community-size distinctions tend to overlap anyway. Big City counties clustered in Greater Buenos Aires (69.4 percent), Township districts fell predominantly on the Coast (63.9 percent), and Countryside areas spread over the Interior (89.1 percent). To a considerable extent the exploration of urban-rural differences implies the simultaneous exploration of regional differences.
The demographic data in Table 1 convey some clear impressions about the relative political importance of the various categories. The Big Cities contained more than 46 percent of the nation’s population,13 the Townships had almost as much, and the Countryside had less than 12 percent. But if these figures show the electoral predominance of urban areas, they also obscure the fact that roughly 38 percent of the national populace lived in rural communities with less than 2,000 inhabitants; and nearly 60 percent of the people in “Townships” actually lived outside of urban districts. Since there is no way of distinguishing between the “urban” and “rural” vote in mixed counties, we might wish to assume that voting patterns for the Countryside apply to other rural communities too.14 Despite its small population, therefore, the Countryside acquires substantive significance in two different ways: on an empirical level, it helps illuminate the political behavior of Argentina’s sizable rural population; on a theoretical level, it provides important contrast with the urban communities and stands as a check or “control” upon observations and inferences concerning the political characteristics of life in the cities.
Preliminary electoral analysis promptly demonstrates political aspects of the urban-rural dimension. As shown in Table 2, Perón averaged 55.2 percent in the Big Cities, where he won three-quarters of the counties; he did less well in the Townships; and he averaged just 46.0 percent in the Countryside, where he carried only 40 percent of the counties and lost many others by decisive margins.15 Even then these results contradict the view that Perón drew all his “important” support from the cities, since he won quite a few votes in the Countryside too.
Level of urbanization not only reveals a direct relationship with the Peronist vote, which I shall symbolize with the notation PERON, but also exerts strong influence upon associations between Perón’s vote and the other variables. Some correlations change sign between CITY SIZE categories. In the Big Cities voter turnout, population size, population growth rate, and percent industrial workers acquire notable positive correlations; the literacy rate, commercial employees per establishment, and wage levels become significantly negative. In the Townships such variables as percent internal migrants, employees per industry, and horsepower per industry gain considerable explanatory power. Few patterns emerge in the Countryside, where many correlations concentrate about zero and tend to lose statistical significance. (See Appendix 2 for all correlations.)
The procedure known as “multiple regression” makes it possible to carry on the analysis by expressing the relationship between the Perón vote and any number of socioeconomic variables by the equation
where X1 through Xk are the “independent” variables (e.g., turnout, internal migrants) and a and b1 through bk are statistical constants. To get the “expected” or “predicted” value of the Perón vote in an individual county, we simply add a plus b1 times the value for the independent variable X1 for that county plus b2 times X2 up through bk and Xk. If the “expected” values for PERON are close to the actual or “observed” values for all the counties involved, then the equation expresses a strong relationship; if not, the relationship is weak.16
Table 3 displays multivariate equations for each of the three urban-rural categories. Since a computer chose independent variables according to standard mathematical criteria,17 these equations represent statistically optimal combinations of socioeconomic factors.18 As presented here, the variables appear in order, from left to right, as selected by the computer. The figures in parentheses give the number of percentage points that would be added to or subtracted from the Peronist total by the product of the constant (b) and the average value for each variable (X), thus providing some notion about the relative impact or political “weight” of the different variables. (For a more detailed version of these equations see Appendix 3.)
Results for the Big City counties are highly suggestive. The consecutive appearance of percent industrial workers, internal migrants, literacy, and population size clearly reveal that Perón obtained strongest support in heavily populated industrial areas with large numbers of illiterates and migrants. These findings strongly confirm the general idea that Perón cultivated his urban following among the native industrial proletariat and displaced migrant population.19
The Big City equation can also provide some fresh insights. The largest change in the electoral result would come from LITERACY, whose average value would depress the Peronist total by —94.94. Next in importance is percent industrial workers, whose mean would increase the Perón figure by +48.17. Percent internal migrants and population size would exert relatively little impact on the final outcome (+13.23 and +3.23).
The unquestionable implication in these figures is that the presence of an industrial work force, rather than internal migrants, furnished the essential precondition for Peronist victory in major urban areas. Percent industrial workers exerts more statistical influence than percent internal migrants. The position of LITERACY strengthens this view because, somewhat surprisingly, it correlates positively with MIGRANTS and negatively with IND WORKERS. In short, Perón seems to have obtained his most crucial electoral support from the “old” working class, not from recent countryside-to-city migrants.20
The very absence of some variables from the Rig City equation bears additional meaning. Voter turnout does not appear to have affected Perón’s vote in any significant way, for instance, suggesting that he had no inordinately effective means for mobilizing mass participation; both sides felt strongly about this election and turnout was generally high in all areas. Horsepower-per-industry’s low explanatory value indicates that the technological level of industrial development made little political difference. The exclusion of percent working class and any data on commercial activity implies that Perón might not have drawn uniformly ardent support from every sector of the proletariat. The apparent irrelevance of unemployment and wage disparities raises some doubts about the functional importance of economic antagonism.21
The equation for Townships presents a complicated picture.22 The first four variables to enter—industrial wage rate, internal migration, industrial horsepower, unemployment—show the continuing relevance of industrialization and the laboring class. If the last two variables reflect individual behavior, they moreover suggest that Perón did well among two specific kinds of workers: the unemployed and the relatively well-paid, both liable to be keenly aware of economic issues (the unemployed for obvious reasons, the well-paid because they probably had to fight for wage increases). The subsequent appearance of INEQUALITY, which measures unequal land distribution, reveals a rural aspect of the situation. With an average contribution of +11.12, INEQUALITY would have considerable positive influence upon the statistical likelihood of Peronist victory.
To me these findings imply that Perón attracted support from a broader “class-wide” coalition in the Townships than in the Big Cities. Concern with the rural economy and land distribution seems to have nearly as much independent impact on probable voting as technological development and other essentially urban matters. The contrast between the positive constant for percent working class and the negative constant for percent industrial workers further strengthens this idea. It appears that Perón’s social base in the Townships did not consist of merely the industrial work force; it depended upon a broad and loosely defined “lower class” which perhaps found unity in economic consciousness.
The Countryside shows yet another pattern. MIGRANTS comes in quite early with a positive constant, much as in the other equations, thus confirming standard impressions about the political significance of internal migration. Yet the strong negative impact of percent Europeans among immigrants (—10.71 on the average) combines with this finding to suggest that Perón did well in rural areas undergoing minimal social change—change that in no way threatened traditional social structures. Internal migration might normally alter community relationships but on this small scale (here MIGRANTS averaged 9.1 percent, compared to a national mean for all counties of 14.2) could not exert much pressure in the Countryside. On the other hand immigration, particularly the European immigration which is negative in this equation, reflects the kind of demographic influx which modernized and revamped society along the Coast and in Greater Buenos Aires.23 Also percent older men (age 50 or more) probably shows the result of out-migration by young males and further testifies to economic underdevelopment. Negative partial correlations for population size and commercial activity variables, controlling for the variables in the equation, further indicate that Perón did well in relatively small, economically backward rural areas. All in all these findings suggest that “social change,” at least according to these variables, would reduce Perón’s electoral chances in the Countryside.
The positive impact of the SMALL HOLDINGS variable—percent landholdings under 100 hectares—demonstrates that Perón obtained support in small farm and ranch districts, but there is not much sign that “class antagonism” would substantially improve his prospects. In fact land INEQUALITY actually has a negative partial correlation with Perón, revealing that he tended to gain votes in areas with relatively equal land distribution. One might guess that Perón attracted support from small-scale agricultural entrepreneurs rather than from a rural peasantry or proletariat.
These results for the Countryside provide a striking contrast with the urban areas, where Perón displayed electoral strength among industrial districts. Within the rural constituency, it appears, he drew votes in the least developed counties; within the Big Cities and particularly within the Townships, with horsepower per industry entering the latter equation, he drew votes in the most developed counties. This paradoxical result implies that, in Argentina, socioeconomic modernization24 exerted a differential political impact. In rural areas it may have engendered frustration among those who were feeling left out or behind; in urban areas it angered those who were taking part in the process but felt exploited by it. In this way Peronism represented a protest against the Argentine pattern of modernization. The movement grew out of the process of social change by providing a means of expression for both its orphans and its victims.
But these inferences for the Countryside have a tentative quality, and this brings up a crucial point. The socioeconomic variables used in this study have much stronger predictive power in the urban areas than in rural ones. The Big City equation explains 53 percent of the “variance” or fluctuation in the Peronist vote, which is to say that “expected” values for PERON are pretty close to the “observed” values. The Township regression, albeit with eight independent variables, explains a respectable 47 percent of the Peronist variance. Yet the Countryside equation accounts for only 19 percent of the fluctuation in PERON, which is to say that expected values are not particularly close to the observed ones.25
This remarkable difference gives rise to another hypothesis: Argentines usually voted on the basis of personal or collective socioeconomic concerns in urban areas, especially in big urban areas, while they often voted according to traditional patterns of loyalty in rural districts. Without proving the point, Table 4 offers some suggestive evidence. The combined average margin of victory for the two sides was 18 percent in the Big Cities, 21 percent in the Townships, almost 27 percent in the Countryside. Thus elections in urban districts were much more competitive than in rural ones. People seem to have been fighting for their interests in the Cities and, to a lesser degree, in the Townships. In the Countryside they might frequently have been following the dictates of local caudillos who could impose an almost county-wide consensus.26
The average pluralities for Perón and his opponent, José P. Tamborini, show further distinctions between the urban-rural categories. The Tamborini figures reveal a gradual increase from roughly 54 percent in the Big Cities to 61 percent in the Countryside—again the apparent transition from urban competition to rural domination. On the other hand Perón won about as handily in the Cities as in the Countryside, a fact which might reflect the social composition of his political constituency. A class-based movement would be likely to win large local victories in cities, just as traditional affiliation would yield lopsided results in rural areas.
In summary, statistical analysis of the 1946 election returns leads to several conclusions about the social base of the early Peronist movement. Since my findings are “ecological” they cannot be conclusive, but they have a demonstrable factual basis and, at the very least, they offer some promising hypotheses for future research. Other investigators might well test these propositions by non-statistical methods; ultimately, of course, a comprehensive understanding of Peronism will have to depend upon qualitative criteria as well as purely quantitative ones.
No single socioeconomic variable explains Perón’s electoral popularity to any large degree. Facile generalizations about urbanization, internal migration, young voters and other such factors do not stand up to empirical analysis.
In the Big Cities, Perón drew greatest support in large, economically developed areas containing both industrial workers and internal migrants. This much generally agrees with standard literature on the subject. But electoral statistics yield two additional refinements: Perón did well among specifically industrial districts, not in all working-class neighborhoods; and the “old” laboring groups played a more crucial political role than did internal migrants.
In the Townships, Perón attracted votes from both urban and rural lower-class groups.
Socioeconomic issues played relatively little political role in the Countryside, though Perón won numerous votes in underdeveloped areas whose constituents might have been eager for modernization. In rural counties traditional allegiance or local strongmen seem to have determined electoral outcomes; incidentally, one possible reason that internal migrants did not exert more positive impact on the Peronist vote in urban areas might be that they retained their customary loyalties for some time after moving to the city.
To the extent that socioeconomic factors can provide an explanation, Peronism began as a protest movement against Argentina’s pattern of modernization, a process which made many people feel exploited or abandoned and thus exerted a differential political impact.
Partly for this reason, Peronism began as a loose and potentially unstable coalition of differing social groups. What was good for Big City workers would not always be good for commercial employees in Townships or agricultural entrepreneurs in the Countryside. The unexplained portions of variance in all three urban-rural categories also imply that many people followed Perón for non-economic reasons. Perón was leading a fragile alliance, not a monolithic class or mass movement.27
In pursuit of the previous point, these interpretations combine to produce some provocative speculation about the subsequent development of the Peronist movement. El Líder’s rural following would probably remain secure as long as he could keep up friendly relationships with local leaders, though he could withstand some loss in districts containing fairly small proportions of the population. For Perón, as for most other politicians, the key to power lay in urban areas. It seems likely, and history appears to show, that the self- identity and sectoral solidarity of the industrial work force could survive any crisis and maintain Perón’s base in Big Cities.
One cannot be so sure about the Townships. If a loose class-wide coalition of Peronist supporters found unity in socioeconomic awareness, as implied by variables about unemployment, wage scales, and land distribution, political allegiance in those areas would probably depend upon tangible gratification. Job-seeking, wage-conscious laborers and land-hungry farmers would be likely to react to the performance of the Perón regime, rather than to political bargains or ideological affinity. This reasoning leads to the hypothesis that, in the Townships, (a) inflation and other economic difficulties after 1950 alienated urban workers, (b) governmental deemphasis of agrarian reform disappointed small farmers and rural laborers, and (c) consequent weakening of the lower-class alliance undermined an essential portion of Perón’s political base and helped make possible his downfall in 1955.
Names, Definitions, and Statistical Characteristics of Variables
Correlations with Peronist Vote by Urban-Rural Categories
Statistical Properties of Regression Equations
Recent analyses include Gino Germani, Política y sociedad en una época de transición: de la sociedad tradicional a la sociedad de masas (Buenos Aires, 1963), Ch. 9; Samuel Baily, Labor, Nationalism, and Politics in Argentina (New Brunswick, N. J., 1967); Peter G. Snow, “The Class Basis of Argentine Political Parties,” American Political Science Review, 63:1 (March 1969), 163-67; Peter H. Smith, “Social Mobilization, Political Participation, and the Rise of Juan Perón,” Political Science Quarterly, 84:1 (March 1969), 30-49. For some conflicting views on the general question of Perón’s popularity see Joseph R. Barager (ed.), Why Perón Came to Power: The Background to Peronism in Argentina (New York, 1968). An important new study, based on a survey taken in 1965, is Jeane Kirkpatrick, Leader and Vanguard in Mass Society: A Study of Peronist Argentina (Cambridge and London, 1971), esp. Chs. 4 and 5.
Biased but informed descriptions of the campaign and election can be found in Robert J. Alexander, The Perón Era (New York, 1951) and George I. Blanksten, Perón’s Argentina (Chicago, 1953). Among the votes cast for the two major presidential candidates, Perón won a resounding 56 percent.
The twenty electoral and census units or circunscripciones of the City of Buenos Aires are included among the 365 “counties.”
For the sake of simplicity, however, I shall make interchangeable reference to “Peronism” and “electoral support for Perón” throughout this essay.
Among Argentine male citizens age 18 or over, women excluded by law. Election data have been taken from Ministerio del Interior, Las fuerzas armadas restituyen el imperio de la soberanía popular (Buenos Aires, 1946), II, and subsequently checked against Darío Cantón, Materiales para el estudio de la sociología política en la Argentina (Buenos Aires, 1968), II, 157-174. The two sources are in complete agreement on most presidential results, with discrepancies falling into three categories: (1) instances where Cantón combined data for two or more partidos, as explained in Vol. I, Apéndice C, pp. xlvii-xlix; (2) non-reportage of marginal third- or fourth-party votes (in Las fuerzas) or blank ballots (in Cantón), which would usually alter the Peronist percentage by one or two points; (3) clerical errors of infinitesimal importance, except two which I changed in accordance with Cantón. Thus the sources are virtually identical with reference to the presidential vote. There are numerous minor discrepancies regarding voter turnout, however; I have used Canton’s figures for the provinces of La Rioja and Tucumán and in several other isolated counties.
Socioeconomic data are from Dirección Nacional del Servicio Estadístico, IV° censo general de la Nación  (Buenos Aires, 1948-52), 3 vols.
See Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge and London, 1969), esp. Part One on “The Logic of Ecological Inference”; W. Phillips Shively, “‘Ecological’ Inference: The Use of Aggregate Data to Study Individuals,” American Political Science Review, 63:4 (December 1969), 1183-1196; and Charles M. Dollar and Richard J. Jensen, Historian’s Guide to Statistics: Quantitative Analysis and Historical Research (New York, 1971), pp. 97-104.
To say nothing of relationships among the socioeconomic variables, which could throw a good deal of light upon Argentina’s social structure in the mid- 1940s. Other kinds of data (for instance on income distribution) are desirable but unavailable. The codebooks and datasets compiled for this study are on file in the Data and Program Library Service at the University of Wisconsin and can be obtained upon request.
On the Pearson product-moment correlation coefficient, which I have used here, see Hubert M. Blalock, Jr., Social Statistics (New York, 1960), Ch. 17, esp. pp. 285-99; or Dollar and Jensen, Historian’s Guide, pp. 56-65.
Relationships are statistically significant to the extent that they would probably not result from a random selection of numerical values for the particular variables. Significance at the .05 level means that random data would be likely to produce the association in question no more than 5 times out of 100. In a way significance tests are irrelevant for this study because I am not trying to generalize beyond the limits of the data, but they still provide a useful check on spurious relationships and data reliability.
These correlation coefficients refer only to linear relationships. In search of possible curvilinear associations I have drawn numerous scattergrams and found nothing.
For example see James R. Scobie, Argentina: A City and a Nation (New York, 1964; second ed., 1971).
Though the fairly small N for Big Cities (49) produces some associations which are not statistically significant at the .05 level, their political importance gives them great historical significance.
This assumption involves some risk because rural residents in Townships lived more or less in the immediate vicinity of urban settlements, so their societal environment might have been quite different from the purely rural Countryside.
The urban-rural classification actually derives from political behavior, since I examined voting results in eight different CITY SIZE categories (described in the notes to Appendix 1) and then condensed the categories on the basis of electoral performance.
On multiple regression see Blalock, Social Statistics, Ch. 19; or Dollar and Jensen, Historian’s Guide, pp. 87-90. For a good application of this technique see Edward J. Mitchell, “Some Econometrics of the Huk Rebellion,” American Political Science Review, 63:4 (December 1969), 1159-71. For more detailed explanations consult N. R. Draper and H. Smith, Applied Regression Analysis (New York, London, and Sydney, 1966); and Blalock, Causal Inferences in Nonexperimental Research (Chapel Hill, N.C., 1964).
With the aid of STEPREG1, a packaged program from the STATJOB series at the University of Wisconsin Computing Center, I employed a stepwise regression procedure and set the F criterion at .10. Most major computer centers have similar programs available. See Draper and Smith, Applied Regression, pp. 171-72.
Since STEPREG1 cannot handle missing data I deleted the land tenure variables from the input for Big Cities and the industrial data from the input for the Countryside; otherwise it would have been necessary to reduce the number of observations drastically for each category. Loss from this procedure is almost certainly negligible, since land tenure bears little relationship to urban areas and industrialization has little impact in rural districts. Several additional variables were omitted from all program runs because of high correlations with other socioeconomic variables (which presented the danger of multicollinearity) or because of interpretive redundance: VOTES, MALES, SA/POPULATION, EU/POPULATION, HECTARES, HP/CAPITA.
Such concepts as “proletariat” and “working class” naturally demand strict definition. Rather than make involved theoretical statements on each of these matters, however, I would refer readers to the operational definitions of the variables in Appendix 1.
To state the argument another way, in an equation involving only IND WORKERS and MIGRANTS as independent variables, IND WORKERS has a beta “weight” of .62 and MIGRANTS has one of only .27. It should be noted that the MIGRANTS variable refers only to migration between provinces and not within provinces, as from rural areas in the Province of Buenos Aires to communities outside the Federal Capital. This limitation undoubtedly distorts interpretation—how much, I do not know.
These variables also have low partial correlations with PERON, controlling for the independent variables already in the equation.
The computer went on to create a more extensive equation for the Township counties, but for purposes of presentation I have selected the eighth step—after which the incremental change in the multiple correlation coefficient was always less than .01.
South American immigration, on the other hand, often reflects the arrival of Bolivian Indians who formed a sort of “underclass” in Salta and Jujuy.
For one definition and application of the quickly changing concept of “modernization” see my Politics and Beef (fully cited in the notes to Appendix 1), Ch. I, “Argentina: The Process of Modernization.”
The difference in coefficients of determination is partly due to the difference in variance for PERON (70.7 in the Big Cities, 188.8 in the Townships, 319.8 in the Countryside for data in the equations). Thus socioeconomic variables have much more to explain in rural areas than in urban ones. It should be possible to equalize the variance through standard statistical transformation procedures and make the equations strictly “comparable” in this sense, but I believe that the result would be misleading: the amount of variance strikes me as an important fact in itself, not something to be overlooked or transformed out of existence.
Of the 49 Countryside districts with margins exceeding 24 percent, Perón won 25 and Tamborini won 24. Note how this pattern, which I would attribute to the traditional political structure of these communities, helps explain the unequal variances in PERON (see footnote 25).
See how these points relate to Markos J. Mamalakis’ theory of “sectoral clashes,” amply expounded and discussed in the Latin American Research Review, 4:3 (Fall 1969).
The author is Associate Professor of History at the University of Wisconsin, and would like to thank the Graduate School at Wisconsin for helping to fund this research. David R. Olson provided expert assistance in the computer programming; Michael Leavitt and James R. Taylor gave valuable advice on the presentation of the data.