¹ See Cutright, Phillips, “Activities of Precinct Committeemen in Partisan and Non-partisan Communities,” Western Political Quarterly, 17 (March, 1964), 93–108CrossRefGoogle Scholar; Cutright, Phillips, “Measuring the Impact of Local Party Activity on the General Election Vote,” Public Opinion Quarterly, 27 (Fall, 1963), 372–386CrossRefGoogle Scholar; Cutright, Phillips and Rossi, Peter H., “Grass Roots Politicians and the Vote,” American Sociological Review, 23 (April, 1958), 171–179CrossRefGoogle Scholar; Katz, Daniel and Eldersveld, Samuel J., “The Impact of Local Party Activities upon the Electorate,” Public Opinion Quarterly, 25 (Spring, 1961), 1–24CrossRefGoogle Scholar; Kramer, Gerald H., “The Effects of Precinct-Level Canvassing on Voter Behavior,” Public Opinion Quarterly, 34 (Winter, 1970–1971), 560–572CrossRefGoogle Scholar; Putnam, Robert, “Political Attitudes and the Local Community,” American Political Science Review, 60 (September, 1966), 640–654CrossRefGoogle Scholar; and Rossi, Peter H. and Cutright, Phillips, “The Impact of Party Organization in an Industrial Setting,” in Community Political Systems, ed. Janowitz, Morris (Glencoe, Illinois: Free Press, 1961), pp. 81–116Google Scholar.

² This view of the party as a dependent variable underlies Robert K. Merton's classic discussion of the functions of the party machine. See Social Theory and Social Structure (New York: Free Press, 1968), pp. 126–127Google Scholar.

³ Crotty, William J., “The Party Organization and Its Activities,” in Approaches to the Study of Party Organization, ed. Crotty, William J. (Boston: Allyn and Bacon, 1968), pp. 247–306Google Scholar; and Key, V. O., Southern Politics in State and Nation (New York: Alfred A. Knopf, 1949), pp. 386–405Google Scholar.

⁴ Eldersveld, Samuel J., Political Parties: A Behavioral Analysis (Chicago: Rand McNally, 1964), p. 424Google Scholar.

⁵ Cutright, Phillips, “Measuring the Impact of Local Party Activity …,” pp. 372–386Google Scholar; and Putnam, Robert, “Political Attitudes and the Local Community,” p. 643Google Scholar.

⁶ There are several methodological reasons why these research findings might be so contradictory: (1) both the independent and dependent variables were measured differently in each; (2) different types of party behavior were examined; (3) different levels of analysis were employed (some studies worked at the precinct level while others worked with counties); and (4) variations in the range of the political environment captured by the independent variable, party competition, were quite small in most cases. Whatever the reason, the resultant problem in cumulating these diverse findings into a theory of environment and party behavior underscores the need for truly comparative studies of party performance across a wide range of political environments.

⁷ See Parsons, Talcott, The Social System (New York: Free Press, 1951), pp. 61–63Google Scholar, for the meaning of particularistic. My use of the term in the context of political machines is derived from James Scott, C., “Corruption, Machine Politics, and Political Change,” American Political Science Review, 63 (December, 1969), 1145–1149CrossRefGoogle Scholar.

⁸ Scott, , “Corruption, Machine Politics, and Political Change,” p. 1150Google Scholar.

⁹ Ibid., pp. 1148–1149.

¹⁰ A clear distinction should be made between the “theory” concerning the relationship of the demographic environment to party behavior and the “findings” concerning the relationship of the political environment to party behavior. The expectations embodied in the theory were derived from studies of party organizations, particularly the political machine, which, although highly insightful, were not systematically empirical. With the exception of the findings reported in the Key study, on the other hand, the discussion of the political environment's impact relies on systematically empirical studies.

¹¹ See Przeworski, Adam and Teune, Henry, The Logic of Comparative Social Inquiry (New York: Wiley, 1970), pp. 106–110Google Scholar, for an excellent discussion of the problems of establishing equivalency in comparative research.

¹² Crotty, , “The Party Organization and Its Activities,” pp. 254–257 and 269–273Google Scholar.

¹³ Eldersveld, , Political Parties, p. 347Google Scholar.

¹⁴ For a good example of the operationalization of this dimension, see Eldersveld, pp. 410–433.

¹⁵ Nonetheless, this assumption underlies some research on party behavior, particularly that which implicitly views the political machine as the “ideal” party organization. For example, see Cutright, “Activities of Precinct Committeemen.” In this comparative study of two cities, Cutright uses indicators which tap only the service dimension of party activity—(1) contacts with voters, (2) requests for aid (3) responsiveness to requests, and (4) whether the precinct leader held a patronage job. While a measure of party activity based exclusively on such indicators may reflect party behavior in a “machine” city such as Gary (Cutright's partisan city), it is surely not generalizable to a non-machine and non-partisan city. In other words, the political machine is only one form of party organization and can not serve as the model against which the activities of all party organizations should be measured.

¹⁶ Wolfinger, Raymond, “The Influence of Precinct Work on Voting Behavior,” Public Opinion Quarterly, 27 (Fall, 1963), 387–398CrossRefGoogle Scholar; Froman, Lewis, “A Realistic Approach to Campaign Strategies and Tactics,” The Electoral Process, ed. Jennings, M. Kent and Zeigler, Harmon (Englewood Cliffs, New Jersey: Prentice-Hall, 1966), pp. 1·19Google Scholar; and Kramer, “The Effects of Precinct-Level Canvassing on Voter Behavior.”

¹⁷ See Froman, “A Realistic Approach to Campaign Strategies and Tactics”; Berelson, Bernard, Lazarsfeld, Paul, and McPhee, William, Voting (Chicago: University of Chicago Press, 1966), p. 171Google Scholar; and Kramer, , “The Effects of Precinct-Level Canvassing on Voter Behavior,” p. 572Google Scholar.

¹⁸ Richard T. Frost lists a number of different campaign activities which were performed by party organizations in eight New Jersey counties. See his “Stability and Change in Local Politics,” Public Opinion Quarterly, 25 (Summer, 1961), 221–235CrossRefGoogle Scholar.

¹⁹ The county chairmen data were collected as a supplement to the Survey Research Center's 1964 election study and were drawn from both parties in all counties where voters were interviewed. Questionnaires were mailed to the chairman of each party in 129 counties and the District of Columbia. The return rate from these mailings was unusually high (more than 90 per cent) and was not noticeably biased in favor of any particular type of county environment. For most of the analyses which follow, the data were weighted by county population to ensure representativeness of the findings. The District of Columbia was deleted from all analysis.

The use of the county as the level of analysis and county party chairmen as the source for perceptions of party activity seems optimal for my present purposes. Because of the decentralization of American party organizations, several levels of analysis are appropriate to the study of environment and party. Most previous studies have utilized either the precinct or the county. Of the two, it seems to me that the county chairmen may be freer to respond to environmental conditions and more knowledgeable of party performance at and below their level than any other party leaders. The use of judges' perceptions to measure performance of party activities can not be justified as easily. The inherent difficulties in measuring party performance directly have forced all previous studies to rely upon indirect measurements. Two types of these indirect measures have been utilized. Most studies have depended upon estimates by party leaders, usually precinct captains or county chairmen depending upon the level of analysis. An alternative method has been to compile estimates of party performance by aggregating voter perceptions of party contact from mass samples. (See Putnam, , “Political Attitudes and the Local Community,” p. 643Google Scholar; and see Kramer, , “The Effects of Precinct-Level Canvassing on Voter Behavior,” pp. 564–565.)Google Scholar The estimates of party leaders, while still indirect, are clearly preferable to those aggregated from a sample of voters. Not only do they capture a broader range of party behavior, but they are less contaminated by errors caused by selective perception and weak memories or by large sampling variances. Nonetheless, the use of an indirect measure is not wholly satisfactory and more work needs to focus on the development of measures of party performance.

²⁰ The Varimax solution is presented in Table 1. An eigenvalue of 1.00 was used as the cutoff for the admission of additional factors. This criterion is suggested by Rummel, R. J., Applied Factor Analysis (Evanston, Illinois: Northwestern University Press, 1970), pp. 362–364Google Scholar. The matrix which was factored to produce this solution contained coefficients based on equally weighted counties in both the South and the North. This weighting procedure was followed to produce a solution which reflected the behavior of county chairmen as nearly as possible. The results remain substantially the same when the counties are weighted by population size.

Precinct Organization was not included in the factor analysis presented in Table 1 because I expected it to form a separate, yet correlated, dimension of activity. The positive correlations between it and all campaign activities but literature distribution support this expectation. Furthermore, a two-factor Varimax solution for all seven measures of party activity virtually replicates Table 1 and adds Precinct Organization as an important item to each factor. While a three-factor oblique solution seems warranted given my expectations, the results of this operation are meaningless because too little variance is left to be explained by the additional factors. More items, and particularly more items which reflect organizational activity, are required before a meaningful three-factor oblique solution can be extracted.

²¹ This procedure for calculating factor scores, called the composite method, uses only those items with the highest loadings on the factor in scoring that factor; the factor loadings used here have been underlined. See Rummel, , Applied Factor Analysis, pp. 441–442Google Scholar, for a discussion of the composite method.

²² This time base is long enough to smooth out the irregularities which may result from the operation of short-term forces, yet not so long that it overlaps two electoral eras. (It would have had to extend back to 1930 to cross into another electoral era.) The offices chosen represent three of the four offices for which county-level data are readily available and are diverse enough so that local idiosyncrasies in voting for a particular office will not distort the measure of the general pattern of partisan division.

²³ Scott, , “Corruption, Machine Politics, and Political Change,” pp. 1145–1149Google Scholar

²⁴ This factor analysis was performed in exactly the same manner as that presented in Table 1, except that it was done for the two regions separately. The population compositions of the southern and nonsouthern counties were too different to allow for strictly equivalent measures and a meaningful general factor solution. The counties were weighted according to size in calculating correlation coefficients so that the results would be representative of the universe of county voter environments.

²⁵ The selection of .70 as a cutoff point for this table in contrast to the .50 used in Table 1 is some-what arbitrary, but it was felt that this level captured a more meaningful factor structure than the lower level would have.

²⁶ These measures of the characteristics of county constituencies are undeniably crude. In an attempt to estimate the concentration of voters whose orientations to politics are what has been described as particularistic, I have been forced to use proxy measures. Instead of information on the distribution of voter orientations, I have used information on the distribution of voters who are likely, according to “machine theory,” to have particularistic orientations —the concentration of minority group members and the median levels of “poverty.” A good deal of measurement error must be anticipated in such measures.

²⁷ Crotty, , “The Party Organization and Its Activities,” and Key, , Southern Politics, pp. 386–405Google Scholar.

²⁸ Eldersveld, , Political Parties, p. 424Google Scholar.

²⁹ Cutright, “Measuring the Impact of Local Party Activity,” and Putnam, “Political Attitudes and the Local Community.”

³⁰ Parallel analyses of the two regions are required for two methodological reasons as well as the substantive one. First, important regional differences in the clustering of demographic features have precluded a factorial structure which is common to both regions. Second, the nationwide distribution of the political environment variable is bimodal and, as a result, seriously violates the normality assumption underlying correlation analysis. The regionally-specific distributions of this variable, however, are close enough to normality to satisfy the assumption. The South includes all the states of the Old Confederacy: Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia.

³¹ Three additional points require elaboration prior to this analysis. First, party performance was measured for 1964 and the environment was measured prior to 1964 so that party performance could be used as a dependent variable. Second, the hypothesized relationships between the environment and the “dependent” activity variables are linear and have been tested with a measure of linear association. Examination of the bivariate scatterplots uncovers no theoretically meaningful departures from linearity. Third, since this is an analysis which is based on data from the 1964 election, generalizations to other years should be made with caution.

³² Some criterion is needed to differentiate between those correlations which are meaningful and those which are not. Since a high probability that the relationship has not occurred by chance seems a prerequisite for meaningfulness, significance at or beyond the .05 level will be the criterion used in the analysis.

³³ This test for spuriousness assumes the following causal ordering among the variables:

Since competitiveness and Democratic concentration are highly intercorrelated, as are many other pairs of independent variables in this study, it is clearly possible that the relationships of either of them to the dependent variable may really reflect the influence of the other on both. Thus, the partial correlation coefficients must be used to sort out the influences of the independent variables. When the original zero-order correlation is substantially reduced by partialing, one may conclude that it is the control variable which is responsible for the bulk of the relationship.

³⁴ Competitiveness and Democratic concentration are almost perfectly correlated at − .94, and this confounds substantive interpretations as will be seen later. The relationship between nonwhite concentration and competitiveness is −.64, while that between nonwhite concentration and Democratic concentration is .63.

³⁵ The scarcity of counties with pro-Republican political environments is clearly reflected in the near-perfect correlation between Democratic concentration and competitiveness. Because of this, scores on either variable reflect the other as well. Any inferences made regarding the relative importance of the two political variables rest on those few counties which are pro-Republican. These inferences must be regarded as highly tentative.

³⁶ The hypotheses involving the service dimension went untested as is indicated by the letter “U.” Some hypotheses were supported in the analysis using the party-specific performance measures as is indicated by the letter “S.” The few hypotheses which did not receive empirical support in the party-specific analysis but did receive it where the relative performance measures were utilized are indicated by “S*.” Rejection of the hypothesis is symbolized by the letter “R,” and each R is followed by a subscript to indicate the reason for rejection. Where rejection was warranted because no significant relationship was found, the subscript is “n.” Where a significant relationship was found which vanished when partials were employed, the hypothesis was rejected due to a spurious relationship as is indicated by the letter “s.” Where the hypothesis was rejected because the relationship was significant but not in the direction predicted by the hypothesis, the subscript “o” was used.

³⁷ For example, the finding that both parties in the North do not perform persuasion activities very well in those counties which contain large concentrations of nonwhites and foreign-borns suggests that persuasion was a rarely practiced activity in major American cities during the late 1800s and early 1900s. To the extent to which these county characteristics reflect parochial orientations to political life, one may argue also that as group identifications become less salient, persuasion will be practiced more by both parties. Similar inferences could be drawn from the impact of concentrations of these types of citizens on Democratic mobilization activities. There has probably been a decline in the emphasis on mobilization as immigrant groups have been assimilated into the mainstream of American society, although this decline has been slowed partially by the emergence of “new” immigrant groups—blacks from the rural South and Puerto Ricans. As these new immigrants are integrated into American society (if that ever happens) and their places are not taken by new waves of immigrants, performance of mobilization activities will decline and political machines will finally vanish.