¹ For summaries of these findings for many nations, see McClosky, Herbert, “Political Participation and Apathy,” International Encyclopedia of the Social Sciences (New York, 1968), XII 252–65Google Scholar; Milbrath, Lester, Political Participation (Chicago, 1965)Google Scholar; and Lane, Robert, Political Life (New York, 1959).Google Scholar

² Although less common, some research has focused on context variables. For example, Tingsten, Herbert, Political Behavior (Totowa, NJ, 1963, originally published in 1937)Google Scholar, examined the effect of homogeneity of neighbourhood on workers' voting turnout; Berelson, Bernardet al., Voting (Chicago, 1954)Google Scholar, explored the role of “community norms” in inhibiting party switching; Matthews, Donald and Prothro, James, Negroes and the New Southern Politics (New York, 1966)Google Scholar, take into account norms, electoral rules, and social structure in their analysis of Negro voter registration and participation in the American South; Siegfried, André, Géographie electorate de l'Ardèche sous la III^eRépublique, Cahiers de la Fondation nationale des sciences politiques, no 9 (Paris, 1949)Google Scholar, examined many animate and inanimate aspects of regional contexts; and Rae, Douglas, The Political Consequences of Electoral Laws (New Haven, Conn., 1967)Google Scholar, provides broad coverage of a number of electoral rules and conventions which influence turnout and voting direction. Despite these scattered efforts, however, the general thrust of research on political participation has been at the level of individual decisions and micro influences.

³ One might profitably make a distinction between regional contexts and psychological contexts, the latter consisting primarily of individual “definitions of the situation.” Lacking individual level data, I must restrict this discussion to regional contexts.

⁴ It is, of course, possible that context variables are important even though one can “explain away” regional differences in terms of the proportions of individuals possessing “appropriate” characteristics. This appears to be the case with literacy as an explanatory variable in Kerala, for example, though the data are not conclusive. At any rate, for purposes of argument, I concede that the relative proportions of such individuals may be important in accounting for regional differences in rates. This form of explanation has also been termed “demographic explanation” by Stinchcombe, Arthur, Constructing Social Theories (New York, 1968), 60–79.Google Scholar

⁵ Indeed it is higher at the state level than the equivalent contests in the United States, where 40 per cent is an average turnout figure (see Table I).

⁶ Within the United States the overt regional differences cannot be fully accounted for in terms of the proportions of individuals with certain characteristics either. See, for example, Campbell, Anguset al., The American Voter (New York, 1960), 227, 453, 477, 480, 490–1, and 495.Google Scholar

⁷ Similarly, changes over time in individual characteristics such as literacy, religion, caste groupings, and the like do not appear to be substantial enough to account for the large changes in participation discussed below. In particular, most such variables change monotonically whereas the patterns of changes over time in participation follow several different paths. See Tables I and II of this paper and chaps. 4 and 5 of my PHD dissertation, “Regional Contexts of Political Participation,” University of California, Berkeley, 1971.

⁸ A recent study which uses some of the same regional features as the present investigation is Zagoria, Donald S., “The Ecology of Peasant Communism in India,” American Political Science Review, LXV (1971), 144–60.CrossRefGoogle Scholar Although superficially similar, Zagoria's analysis does not actually interpret land tenure and hierarchy as regional contexts but as sources of individual hostility to the political system. Thus, the level of analysis is somewhat different. Furthermore, he frequently uses aggregate data on districts to make inferences about individual motives, thus committing the “ecological fallacy” (see n. 47 below).

⁹ In fact, the concept and the phenomenon are quite common throughout India. In the north, however, there are features which qualify the statements made here about the south. For example, the greater incidence of landholding by Brahmins in some parts of north India, the higher proportion of Muslims, the dislocation due to Partition, and the history of foreign incursions have resulted in somewhat less clear-cut lines of organization. On the other hand, the left-right division of society, common in Tamilnadu, Mysore, and Andhra in earlier periods, is not so prevalent in the north.

¹⁰ The literature directly on work teams is sparse. The most cogent is Neale, Walter C., “Land Is to Rule,” in Frykenberg, Robert, ed., Land Control and Social Structure in Indian History (Madison, Wis., 1969).Google Scholar For extensive citations of related literature, see chap. 2 of my thesis and Bailey, F.G., Stratagems and Spoils (Toronto, 1969).Google Scholar

¹¹ Beck, Brenda E.F., Peasant Society in Konku (Vancouver, 1972).Google Scholar

¹² Dugal, Brijinder Singh, “The Village Chief in the Indian Construction Industry,” Human Organization, XVIII (1959–60), 174–6CrossRefGoogle Scholar; and Ruth S. Simmons, personal communication to author.

¹³ Dugal, “The Village Chief.”

¹⁴ The word was first popularized by William, and Wiser, Charlotte in their book, Behind Mud Walls (Berkeley, Calif., 1964, originally published in 1930)Google Scholar, though similar words are used throughout India. For example, in Tamilnadu we find the term ejamanka. For more comprehensive views, see Kolenda, Pauline, “Toward a Model of the Hindu Jajmani System,” Human Organization, XXII (1963), 11–31CrossRefGoogle Scholar; Epstein, T. Scarlett, “Productive Efficiency and Customary Systems of Rewards in Rural South India,” in Firth, Raymond, ed., Themes in Economic Anthropology (London, 1967)Google Scholar; and Beck, Peasant Society in Konku.

¹⁵ Beck, Brenda E.F., “The Right/Left Division of South Indian Society,” Journal of Asian Studies, xxix (1970), 779–98.CrossRefGoogle Scholar

¹⁶ See Neale, “Land Is to Rule,” and Beck, “The Right/Left Division of South Indian Society,” and the references cited in both of these articles.

¹⁷ The work team arrangement is, of course, not limited to India. Under various names it may be found in many “traditional” politics. For example, clear parallels abound in Landé, Carl, Leaders, Factions, and Parties: The Structure of Philippine Politics (New Haven, 1965)Google Scholar; Tarrow, Sidney, Peasant Communism in Southern Italy (New Haven, 1967)Google Scholar; Barth, Frederick, Political Leadership among the Swat Pathans (London, 1959)Google Scholar; Powell, John Duncan, “Peasant Society and Clientelist Politics,” American Political Science Review, LXIV (1970), 411–25CrossRefGoogle Scholar; Zolberg, Aristide, “Mass Parties and National Integration: The Case of the Ivory Coast,” Journal of Politics, xxv (1963), 36–48CrossRefGoogle Scholar; Beardsley, Richard K.et al., eds., Village Japan (Chicago, 1959)Google Scholar, especially chap. 13; and urban political machines in the United States as analysed by Merton, Robert, Social Theory and Social Structure (New York, 1957).Google Scholar

¹⁸ The term was popularized in Srinivas, M.N., “The Dominant Caste in Rampura,” American Anthropologist, LXI (1959), 1–16CrossRefGoogle Scholar; but see also the important contribution by Mayer, Adrian in Caste and Kinship in Central India (Berkeley, 1960).Google Scholar

¹⁹ Srinivas seems to use the term only for large castes, and indeed he implies that dominance rests “ultimately” on force of numbers. His own data and every other village ethnography demonstrate, I believe, that dominance rests on control of the work teams, that is, control of the relevant economic resources of an area, even if the caste is a small one. Certainly wherever a Brahmin group is dominant it cannot be due to numbers since they are generally only a small percentage of the village population. For an extended discussion of more adequate ways of measuring dominance, see chap. 7 of my thesis.

²⁰ For detailed discussions of clan organization and clan dominance of particular areas of north and south India, see Beck, Peasant Society in Konku, and Singh, Kashi N., “The Territorial Basis of Medieval Town and Village Settlement in Eastern Uttar Pradesh, India,” Annals of the Association of American Geographers, LVIII (1968), 203–20.CrossRefGoogle Scholar

²¹ See Pocock, David, “The Bases of Faction in Gujerat,” British Journal of Sociology, VIII (1957), 295–306CrossRefGoogle Scholar; Nicholas, Ralph, “Factions: A Comparative Analysis,” in Bandon, Michael, ed., Political Systems and the Distribution of Power (New York, 1965)Google Scholar; F.G. Bailey, “Decisions by Consensus in Councils and Committees: With Special Reference to Village and Local Government” in ibid.; and Beck, Peasant Society in Konku.

²² As an aside, it should be noted that one argument sometimes adduced for the role of caste in politics concerns the fact that in many areas of India all the major parties choose their candidates from the same caste or subcaste. The present model of social organization places this “practice” in a wider framework. In particular, the local branches of political parties often reflect social divisions and factions. Thus, the parties' candidates are drawn from among bloc and faction leaders or their designated representatives. In areas where one caste is dominant, all of these will be from the same caste; in areas where no one caste is dominant, they will be from several different castes. This suggests that, rather than look for some overly simple relation of caste to politics, it may be fruitful to think of caste's role as almost that of an epiphenomenon. The crucial variables according to the present model are social diversity, the degree of inter- and intra-caste fragmentation and factionalism, and related degrees of party competition.

²³ Most of the sociological material is drawn from the 1961 Census. The caste data, however, were gleaned from the 1911 Census. Since no data on caste were collected by the census enumerators after 1931, one must rely on these early reports. It is my judgment that the caste data in the 1921 and 1931 reports were highly deficient; therefore I relied on the 1911 figures. This poses certain problems dealt with extensively in appendices A and C of my thesis.

²⁴ A full discussion of problems with the data, and suggestions for more refined measures when better data are available, may be found in chap. 7 and appendices A and C of my thesis.

²⁵ The identification of landholding castes was accomplished by several means. First, ethnographies indicate the principal landholding castes of the area where fieldwork was done. Second, studies of particular castes, such as Hardgrave's, RobertThe Nadars of Tamilnad (Berkeley, 1969)Google Scholar, yield some clues. Third, using the lists of castes for each district (taken from the census), I looked up those not known to be untouchables in the “castes and tribes” compendia such as Thurston and Ananthakrishna Iyer. Only castes that were clearly identified as major land-holders were then included. Obviously there is some danger of changes since the compendia were assembled, but it is unlikely that these are substantial since it has been difficult for large land transfers to occur in recent decades.

²⁶ There are, of course, problems of definition in deciding what constitutes a caste. I have relied on the 1911 Census for data, and therefore I have generally counted as castes those groups that are listed therein as separately named groups. In a few cases, where it is widely known that two or more names are simply equivalent alternatives for the same groups, I have combined them.

²⁷ That is, large proportions of labourers yield a low score on the Hardy Peasant Index, whereas high ownership or tenancy rates result in high scores on this index.

²⁸ This necessarily weights the degree of dominance higher in rural than in urban areas, but as we shall see this is reasonable given the south Indian case.

²⁹ In constructing indices of this sort, one can either take an average score on the components, essentially a “Likert” procedure, or one can take a cumulative score. In all of the indices described in this paper, I have used cumulative scores so as to maximize the range of scores. For example, in the Hardy Peasant Index each of the four components was trichotomized and scored as 0,1,2: two being the “high” score (after reversing the scoring of agricultural labourers). Thus, the range was from zero to eight for the composite index.

³⁰ The reader will note that there is no direct measure of the strength of work team organization, which is one of the most significant parts of the model of social organization. This is not a serious omission, since the work teams and dominance are closely related phenomena. After all, the way a caste becomes dominant is by controlling a large proportion of the work teams. Thus, to the extent that the Hardy Peasant Index, the Caste Fragmentation Index, and the Index of Caste Dominance measure the degree to which a single group controls the work teams, to that same extent the indices measure the strength of the work teams. The Hardy Peasant Index seems particularly suited to this interpretation.

³¹ The figures for the percentage of Muslims in a district were coded for both the 1911 and 1961 Censuses with virtually identical results. Therefore, I have used the 1961 figures since they are more up-to-date and correspond more exactly to the present district boundaries.

³² Rather than measure the competitiveness of any one seat, this index focuses on the control of a region by combining information on the division of the seats between the largest and second largest parties, whether there have been changes in which party has the majority and the number of uncontested seats. Considering the obstacles involved in devising adequate measures of party competition in multi-party systems as compared to two-party systems, the present index is quite satisfactory. For further discussion, see my paper, “Problems of Measurement of Party Competition with Special Reference to Multi-Party Systems,” forthcoming.

³³ Milbrath, Political Participation; McClosky, “Political Participation and Apathy,” and Norman, Nie G.Powell, Bingham, and Prewitt, Kenneth, “Social Structure and Political Participation: Developmental Relationships,” American Political Science Review, LXIII (1969), 361–78 and 808–32.Google Scholar

³⁴ For detailed evidence on these units, see chaps. 1 and 3 of my thesis, and the references cited there. Briefly, one can say that the district boundaries which the British used in the earliest censuses (roughly mid-nineteenth century) followed social and political boundaries extant at that time. For extensive description and documentation of such an area in Tamilnadu, see Beck, Peasant Society in Konku; for north India, see Kashi N. Singh, “The Territorial Basis,” and Day, Winifred, “Relative Permanence of Former Boundaries in India,” Scottish Geographical Magazine, LXV (1949), 113–22.CrossRefGoogle Scholar

³⁵ After analysis for the present study was virtually complete, I became aware of a recent book by Baxter, Craig, comp., District Voting Trends in India: A Research Tool (New York, 1969)Google Scholar, which presents (but does not analyse) electoral turnout data aggregated at the district level and generally following the 1967 district boundaries. Thus, it parallels my data closely. The figures, however, differ slightly from mine. This is partly due to different ways of averaging constituency results and partly due to different ways of resolving ambiguities in the delineation of constituency boundaries in earlier elections. For a discussion of discrepancies, see appendix A of my thesis.

³⁶ For details, see chap. 3 of my thesis.

³⁷ As delineated here, Telengana consists of the following districts: Adilabad, Nizamabad, Karimnagar, Medak, Warangal, Hyderabad, Nalgonda, and Mahbubnagar. Rayalseema consists of: East and West Godavari, Krishna, Guntur, Khammam, Kurnool, Nellore, Anantapur, Cuddapah, and Chittoor. The Northern Circars region, as defined here, consists of Srikakulam and Visakhapatnam. Historically these regional labels refer to slightly different groupings, but I felt it better to use familiar labels than to make up new ones.

³⁸ For details, see chap. 3 of my thesis.

³⁹ Thus the grand means do not always correspond exactly to the (equally weighted) average of the year-by-year means in the tables.

⁴⁰ The correlations of number of candidates with voting turnout are presented in Table III for the information of the reader but are not essential to evaluating the model, since the model does not specify what the correlation should be, except in the limiting case where only the social forces in the model are operative. In that case the correlation should be negative. Note that this is true for Tamilnadu.

⁴¹ Since there are only two districts in the region called Northern Circars, one cannot calculate a correlation coefficient. Hence I have omitted this region from some parts of the analysis.

⁴² Since all nine districts in Kerala have the same very high score on Dominance, a correlation is meaningless. Since these districts have the highest levels of turnout in the south, in an analysis of all sixty-one districts together, rather than region by region, they strongly support the hypothesis.

⁴³ For reasons which cannot be examined here, there is no linear relationship between the number of candidates (or the number of parties) and party competition. Thus, it is not feasible to make predictions from the model of social organization to the correlation of number of candidates and party competition. For a discussion of the curvilinearity of the relationship, see my paper on party competition in multi-party systems (n. 32 above).

⁴⁴ These results may be found in chap. 4 of my thesis.

⁴⁵ This relationship is not invariate. The Mahars, for example, are highly participant in the politics of Maharashtra, despite being a very low status group. See Zelliott, Eleanor, “Dr. Ambedkar and the Mahar Movement,” unpublished PHD thesis, University of Pennsylvania, 1969.Google Scholar

⁴⁶ A similar pattern obtains when one looks at the number of candidates in SC/ST areas, thus further strengthening this conclusion.

⁴⁷ When one infers from data about a group or an aggregate to the individuals (or sub-groups) composing the aggregate, one makes what is called an “ecological inference.” Such inferences are not logically warranted, and hence the term “ecological fallacy.” Under some circumstances the inference is more or less accurate, and a considerable body of literature exists which discusses this possibility. See Yule, G. Udny and Kendall, M.G., An Introduction to the Theory of Statistics (14th ed., London, 1950), 310–23Google Scholar, for an excellent discussion. See also Dogan, Mattei and Rokkan, Stein, eds., Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass., 1969)Google Scholar, for extended examples and discussion; and Shively, W. Phillips, “ ‘Ecological’ Inference: The Use of Aggregate Data to Study Individuals,” American Political Science Review, LXIII (1969), 1183–96.Google Scholar

⁴⁸ “Forcing” literacy and per cent urban into the equation raises the R ² to 0.939, a minuscule gain

⁴⁹ Tamilnadu's standard deviation on voting turnout is only 3.02 compared to 5.29 for the next lowest SD and 6.82 for the maximum SD.

⁵⁰ No such problem of attenuation exists for Tamilnadu on the number of candidates per constituency, where its standard deviation falls in the middle for the six regions. In that case 77 per cent of the variance is accounted for in Tamilnadu, indirectly supporting the argument that attenuation is a factor in the low R ² for voting turnout.

⁵¹ All of the transfers are detailed in chap. 3 of my thesis.

⁵² There is no adequate way to summarize these data in a single coefficient. See the full data in chap. 5 of my thesis.

⁵³ Note that a positive correlation indicates a negative relationship, since the larger the change from the previous election in the number of candidates the lower the present number if there was a large number of Independents in the previous election. For example, in 1952, there was a very large number of Independents in most areas, nearly all of whom lost their deposits; this was followed by very few Independents and thus few candidates in 1957, and the correlation is +0.36.

⁵⁴ The correlation of 0.13 is not quite significant at the .05 level, but the other two correlations are very significant.

⁵⁵ This is the approach taken by Matthews and Prothro in their excellent book, Negroes and the New Southern Politics, 146 ff. They “predict” the level of voter registration expected in a county on the basis of a regression equation containing twenty-one social and demographic factors which affect individual registration. Counties with residuals nearly equal to zero are considered to be explained by the proportions of individuals with the “appropriate” characteristics; while counties with large residuals are set aside for further examination in terms of structural, historical, or contextual factors not subsumed in the individual characteristics.

⁵⁶ Towns tend to resemble rural areas in voting turnout but to resemble cities in the number of candidates per constituency. See chap. 4 of my thesis.

⁵⁷ Deutsch, Karl, “Social Mobilization and Political Development,” American Political Science Review, LV (1961), 493–514.CrossRefGoogle Scholar See also my paper on “Social Mobilization, Social Structure, and Politics,” presented at the American Political Science Association annual meetings, Chicago, Sept. 1971.