politics and geography || electoral geography: retrospect and prospect

22
Electoral Geography: Retrospect and Prospect Author(s): R. P. Woolstencroft Source: International Political Science Review / Revue internationale de science politique, Vol. 1, No. 4, Politics and Geography (1980), pp. 540-560 Published by: Sage Publications, Ltd. Stable URL: http://www.jstor.org/stable/1600753 . Accessed: 23/09/2013 06:05 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to International Political Science Review / Revue internationale de science politique. http://www.jstor.org This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AM All use subject to JSTOR Terms and Conditions

Upload: r-p-woolstencroft

Post on 17-Dec-2016

231 views

Category:

Documents


15 download

TRANSCRIPT

Electoral Geography: Retrospect and ProspectAuthor(s): R. P. WoolstencroftSource: International Political Science Review / Revue internationale de science politique, Vol.1, No. 4, Politics and Geography (1980), pp. 540-560Published by: Sage Publications, Ltd.Stable URL: http://www.jstor.org/stable/1600753 .

Accessed: 23/09/2013 06:05

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to InternationalPolitical Science Review / Revue internationale de science politique.

http://www.jstor.org

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

ELECTORAL GEOGRAPHY Retrospect and Prospect

R. P. WOOLSTENCROFT

In recent years there has been increased interest, especially from geographers, in the field of electoral geography. In this article three major approaches to the study of voting-the cartographic method, the "friends and neighbors" models, and the "nationalization" models-are discussed and evaluated. It is argued that the cartographic approach is very limited in its utility because of the static and imprecise character of the work as well as its low explanatory potential. The "friends and neighbors" models are most applicable in very restricted circumstances. Under other conditions they are likely to be of marginal value. The "nationalization" approach is of considerable importance, although some interpretative problems still exist. Despite some analytical and conceptual differences, these approaches to electoral geography are marked by an underlying similarity: a concern for the character and delineation of subnational political effects which are spatially demarcated. The "nationalizational" approach, it is argued, is likely to be the most useful for the description and explanation of such phenomena.

If it is the case that political systems are embedded in space, as Aristotle suggested, then it would seem incumbent on political scientists-as well as other social scientists-to develop ana- lytical models and techniques which explicitly incorporate spatial variables, processes, and relationships, to unlock the impact of spatial factors. However, the textbooks of political science, which, if Kuhn (1970) is right, tell us how practitioners perceive their discipline, are generally innocent of analyses of the geo- graphical underpinnings of politics. In a work intended to bring together geographers and political scientists, the majority of con- tributors had geography affiliations (Cox et al., 1974). It seems fair to suggest that political scientists, at least until recently, generally have not accorded space a high ranking on the agenda of research priorities.

Geographers, in comparison with political scientists, have devoted more of their intellectual resources to understanding

International Political Science Review, Vol. I No. 4, 1980 540-560 ? 1980 International Political Science Association

540

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 541

spatial dimensions of political systems as well as to staking out the frontiers and boundaries of the field of political geography. This commitment, readily tainted by easy confusion with the amorphous but politically disreputable concepts of geopolitics, has also met with more than passing skepticism, as evidenced by Hartshorne's (1954: 170) scathing comment that in "perhaps no other branch of geography has the attempt to teach others gone so far ahead of the pursuit of learning by the teachers." Berry's (1969: 450) derisive dismissal of political geography as a "mori- bund backwater" reflected the belief that much of political geography lacked systematic order and theoretical structure.

Whatever the accuracy of Muir's (1975: 203) observation that electoral geography, as a subfield of political geography, has enjoyed increased research activity since World War II, it has attracted neither much interdisciplinary work between political scientists and geographers, nor, indeed, much priority within either discipline. Nonetheless, some common themes have devel- oped, with those working in geography apparently more sensitive to questions of proper disciplinary specialization.

Muir for instance, following the lead of Prescott (1972), wonders if much that is included in electoral geography is not more the province of sociology and political science. Specifically, he argues (1972: 210) that until "voting decisions can be clearly related to spatial contexts, their geographical relevance is un- proved." On the other hand, Busteed (1975: 3) argues that geo- graphy is capable of providing "an entirely new dimension to the study of elections." R. J. Taylor and R. J. Johnston (1979)-al- though somewhat less sanguine-still find geographical perspec- tives important for the understanding of elections.

Political scientists, on the other hand, generally have studied elections without much reference to geographical factors or the works of electoral geographers. This disciplinary isolation, at least from the perspective of political scientists, stemmed partly from the often narrow definition of electoral geography put forth by those working in geography and the reflexive aversion to connotations of geopolitics and geographical determinism asso-

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

542 POLITICS AND GEOGRAPHY

ciated with Ratzel and Mackinder, or the tenuous distinctions drawn by such electoral geographers as Krehbiel (1916: 419) be- tween "natural conditions" and "artificial influences."

This inattention to locational considerations can be partially explained as well, no doubt, by the rapid diffusion since the 1930s of survey research techniques througout the various academic communities and the demonstration by Robinson (1950) that there was no necessary relationship between individual- and aggregate-level correlations. The impact of these developments on political science, among other things, was to create an overarching emphasis on the production and analysis of indi- vidual-level data, which, given the relative ease of cross-sectional studies, meant a concomitant contraction of interest, at least until the 1960s, in questions relating to the ecology of systems and variations in behavior between spatially demarcated units of analysis (Dogan and Rokkan, 1969: 1-15).

More fundamentally, however, there was a set of methodo- logical presuppositions that manifested itself in more than the primacy attributed to individual-level socioeconomic and psy- chological variables. They suggested that only at the far end of the explanatory chain-in some pool of unspecified variables- should matters of space be found, as represented by labels for systems and subsystems. Indeed, as Przeworski and Teune (1970: 74-76) argue, this part of the explanation is not of further theoretical interest unless it is possible to move beyond titles of places to further specification of variables and relationships peculiar to particular places. Geographical variables, then, are only residual containers tacked on at the end of descriptive and explanatory accounts.

Electoral geography has incorporated a number of different themes, ranging from the study of voting in legislatures and international assemblies, through the relationship between elec- toral and spatial systems, to geographical influences upon voting behavior. This essay will focus on the last theme, with specific reference to three different approaches-the cartographic, the "friends and neighbors," and the "nationalization" approaches-

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 543

in an attempt to evaluate their relative contributions to under- standing the electoral process, and to make an argument about the possibilities of further research.

THE CARTOGRAPHIC APPROACH

Although early examples of the cartographic approach, or its two main divisions-the areal-structural and the areal-ecological (Busteed, 1975: 25-39)-can be found in the work of Krehbiel (1916) and Wright (1932), it has undoubtedly been the work of the French political geographers, notably Andre Siegfried (1947) and Francois Goguel (1951), which has been identified closely with this approach to electoral geography. Although the research is extraordinarily time-consuming, a statement of the method is relatively straightforward: Maps of the distribution of election results and various socioeconomic characteristics are produced and then compared to ascertain relationships.

Prescott, a leading electoral geographer, in the first of three interpretations of the field, following the influence of Siegfried and Goguel, understood electoral geography to be the presenta- tion of election issues through the use of maps, provided that the issues are related to the raison d'etre of the state (1959: 304). Prescott argued that cartographic analysis would be most profitable in either recently formed systems or systems with electorates marked by intense cleavages along ethnic, linguistic, or religious lines. The object of such research was to delineate political regions and to ascertain degrees of cohesiveness (1959: 300-304).

Subsequently Prescott (1969: 379) deemphasized the necessity of restricting cartographic analyses to issues of state existence and expanded the definition of electoral geography to include politi- cal issues. But even here definite results would be limited to those occasions when electorates were given clear policy alternatives or the election involved one sharply defined issue. Failing these conditions, the analytical task was one of understanding "the continued emergence of the same areas of party preponderance, and the explanation of short- and long-term changes in party strengths and weaknesses" (Prescott, 1972: 87).

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

544 POLITICS AND GEOGRAPHY

Explanations for instances of long-term domination would include "population characteristics which appear to distinguish groups of common interest," such as "race, perceived social and economic status, nationality and age." Contrary patterns will probably be explained in terms of short-term issues, which will move even homogeneous electorates in unexpected directions, such as occurred in the 1964 American presidential election when the South abandoned the Democratic Party for the Republicans under Barry Goldwater; or, in those regions characterized by heterogeneous electorates but strong long-term support for one party, the explanation is likely to be found in the personal appeal of incumbents (Prescott, 1972: 88).

It is clear that Prescott's understanding of electoral geography is in terms of regional delineation, with a primary focus on socioeconomic variables. Regionalism is a pattern of subnational variations as reflected by the proportion of winners and losers. The main technique is the employment of maps, although the best results occur when they are used in conjunction with statistical methods. Prescott offers the cautionary injunction that statistical methods either will not reveal significant relationships or will offer only confirmation of what already is known. Indeed, it will be the results of cartographic studies that will provide common- sensical descriptive interpretations (Prescott, 1969: 381).

Although we do not have an abundance of cartographic analyses, we do have sufficient examples that we can speak with some confidence about their utility for the study of elections.

One example is found in Kinnear's The British Voter (1968). Maps were used for the display of the pattern of winning parties in each constituency for each general election form 1885 to 1965. While some overall spatial patterns are discernible, the maps, as noted by Prescott (1972: 81), lack linear scale; so, despite Kinnear's admonition (1968: 12) that we must bear in mind differences of size, the inevitable impression is one of land rather than people. Moreover, while some overall descriptive impres- sions of the relationships between "area," as a surrogate for socioeconomic variables, and party vote are generated, there is no reliable way to measure the strength of the relationships, which results in the fundamental limitation: The analyst cannot speak in

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 545

other than nominal terms about changes over time within a system, or in very broad terms across systems. We have in Kin- near's work, then, a clear illustration of the analytical weaknesses of descriptive electoral geography.

Another example of such research is found in Busteed's Geography and Voting Behaviour (1975), where the spatial distribution of seats and votes won by the Scottish Nationalist Party (S.N.P.) in the February 1974 British general election are shown through various cartographic techniques. However, Bus- teed's statement of the factors underlying the spatial dispersion of S.N.P. support depends on information and variables not pro- vided by the cartographic technique. Other than for the visual impact of election results, it is not clear what analytical advantage is gained from reliance on maps for the depiction of electoral processes.

Lewis's (1965) cartographic examination of Black migration and voting patterns in Flint, Michigan, represents a sophisticated example of the method. Through a series of maps and overlays, Lewis traced the movement of Blacks away from the Republican Party to the Democratic Party. Lewis concluded that the cartographic method of electoral analysis is useful, provided that the socioeconomic groups under analysis are located within well- defined geographical areas so that clear associations can be drawn, and the various groups exhibit strong support for the parties, so that clear and distinct regions can be displayed. Lewis argued as well that electoral maps would be useful as a means of identifying areas of aberrant political behavior (Lewis, 1965: 23-25).

Even with these limiting conditions-which in themselves seriously constrain electoral geography-some other important qualifications must be recognized. First, maps are static in character, even in those instances when time series are examined. Second, though Lewis went beyond winners and losers to include proportions of the total vote by various groups for parties-and thus utilization of isopleths to mark off regions of higher and lower support for the parties-the strength of relationships cannot be ascertained.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

546 POLITICS AND GEOGRAPHY

The question that arises is whether or not the cartographic approach allows for the generation of information or insight not otherwise available from other approaches, particularly the correlation of various aggregate-level data, such as was done by Roberts and Rumage (1965). In the absence of such data, of course, the cartographic approach will have some utility, and as Lewis suggests (1965: 24), it can perform the heuristic function of suggesting avenues for further research, but it is to be doubted that much will be gained from the utilization of both as opposed to reliance on statistical methods alone. It is noteworthy in this regard that the part of Lewis's analysis which is dynamic in character relies upon statistical evidence that came from the very sources used for the construction of the maps and led to a conclusion at odds with the cartographic evidence (namely that the evolutionary pattern of Black voting had not been "unstable and heterogeneous" but rather "sure and unsystematic" (Lewis, 1965: 17-18). The fact that Lewis accepted the authority of the statistical over cartographic evidence suggests that greater re- liability and certainty are to be found with the former. And as the analysis moves from description to explanation, the carto- graphic evidence becomes less and less useful.

Prescott (1969: 381) argued that the results of the study by Roberts and Rumage should be of no surprise for those who had read Krehbiel's study of eight British general elections from 1885 to 1910. Surely this is an overstatement. While Krehbiel, for example, on the basis of his cartographic analysis, can point to relationships between socioeconomic characteristics and party vote, their formulation lacks the specificity provided by Roberts and Rumage. And while Roberts and Rumage do not speak directly to the question of political regions, their study seems to produce results similar to the kind desired by Prescott, namely the spatial distribution of election results and the underlying socio- economic determinants. As Dogan and Rokkan (1969: 10) put it, "there is no obvious cartographic equivalent of the partial correlation," so that electoral geography is severely limited in terms of the kinds and numbers of variables which can be incorporated into any analysis.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 547

THE FRIENDS AND NEIGHBORS APPROACH

The cartographic approach, whether undertaken primarily by maps or statistical analyses, raised questions about the geograph- ical import of the research. Geographers such as K. R. Cox (1969) and D. R. Reynolds and J. C. Archer (1969) argued that the traditional emphasis on area and social structure did not incorporate spatial relationships or materials that were pertinent to geographical theory.

In an attempt to move away from the aspatial study of electoral behavior, Cox and Reynolds and Archer, argued for the value of spatial diffusion models of political behavior that would enable geographers to address the phenomena of neighborhood effects identified by Tingste-n (1963: 171-180), "friends and neighbors" voting identified by Key (1949), and the "breakage" effect identified by Berelson and his associates (1954: 98-101). These phenomena, while somewhat different in conceptual terms, are joined by the postulate that local factors and processes structure the vote in such a way as to produce either deviations from higher- order patterns or deflections, upwards or downwards, from community trends. Tingsten's work suggested that "electoral participation within a social group rises with the strength of the group in the area in question"(1963: 171). Key used the term "friends and neighbors" to describe voting in a number of southern American states which was marked by the strong proclivity of electors to support candidates on the basis of personal factors, particularly their residence. Key suggested that this kind of voting was most evident in primary elections, in small towns and rural areas, and where parties were not well-organized (Key, 1949: 37-41). Berelson's concept of a breakage effect referred to the tendency of voters with inclinations or loyalties to minority views as well as majority views to fall in line with the dominant opinion. Behavior, in part, reflected the impact of the local political environment.

Since Cox and Reynolds and Archer began work on spatial diffusion models or contagion models, a considerable amount of research has been published, with a majority dealing with local

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

548 POLITICS AND GEOGRAPHY

elections or national elections in local communities. A useful discussion appears in Taylor and Johnston (1979) and an interesting set of essays is found in Johnston (1977). Analytical themes have ranged from the specification of constituencies which deviate from predicted levels of party support (Crewe and Payne, 1976; Miller, 1977; Blake, 1978) to estimations of the impact of residential location upon behavior (Cox, 1969).

Cox (1972) identified three different strategies that had been adopted for the problem of ascertaining neighborhood effects: contextual studies, where aggregate variables and individual- level variables are cross-tabulated to create frequency counts, with differential frequencies measuring local effects; continguity analyses, which compare residuals from regression analysis in an attempt to discover structuring of electoral phenomena along spatially demarcated lines; and curve fitting, where theoretical curves are matched with actual curves, with deviations measuring neighborhood effects. Cox argued (1972: 166) that these stra- tegies, which "attempt to partial out the effects of individual resistance to information flow (individual effects) and the spatially structured process of information flow itself (aggregate effects)," suffer from reliance on large-scale data and "their over- whelming bias to the statics of political behavior [rather] than [to] dynamics." Cox employed an alternative technique: transition probability matrices, which, once computed, are regressed against "selected probabilities on territorial unit aggregate voting proportions [to] obtain a crude measure of neighborhood effect." With some concessions to his data base Cox was able to make some interesting conclusions-that voting behavior in Colum- bus, Ohio, in the 1964 presidential election was structured to a considerable extent by neighborhood effects, to the point that the estimation of transitional probabilities by ecological regression would be seriously led astray (Cox, 1972: 167-172).

Reynolds adopted a number of stochastic models in an attempt to ascertain their capacity to "mirror aspects of the spatial distribution of political behavior" rather than being concerned with showing that "regularities in voting behavior are consistent with notions of spatial contagion" (Reynolds, 1974: 258). The

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 549

three models-the contagious Poisson, the negative binomial, and Neyman's Type A distribution-are predicated on assump- tions relating to interpersonal communications between individ- uals, particularly as embodied in the two-step model of communi- cation (Reynolds, 1974: 261-262). Evidence supporting contagion hypotheses was found for the case of the 1967 mayoral election in Indianapolis, Indiana.

Fitton, who employed a radically different research strategy from those of the above-mentioned studies, found evidence of a neighborhood effect in three streets in Manchester, England. Sociometric analysis suggested that changes in party support were in the direction of the group to which individuals were attached (Fitton, 1973: 471).

The number of studies employing some variation of the neighborhood effects approach is suggestive of the intriguing character of the phenomenon and supportive of the hypothesis that local political environments structure in some way the direction and magnitude of the behavior of electorates. Nonethe- less, a number of critical comments seem warranted.

First, it seems clear that friends and neighbors voting, in the sense suggested by Key, is found in rather restricted circum- stances, namely those elections in which strong political parties or factions are lacking, issues are not dominant, and candidates are contesting their first election (Tatalovich, 1975: 47047 1).

Second, the neighborhood effects models, to the extent that they are based on interpersonal communication flows and conversions through personal interaction, are based on assump- tions negated by more broadly based empirical works. The argument presented in The American Voter (1960: 295-333), for instance, clearly denies the efficacy of personal interaction processes with respect to voting behavior. To the extent that party identification, candidate orientation, or issue orientation ema- nate from or are constrained by forces above and beyond the local community, the influence of personal relationships must be diluted. Party identification, in particular, can be regarded as providing a filter to extraneous messages, whether they come from the media or the local neighborhood.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

550 POLITICS AND GEOGRAPHY

David Butler and Donald Stokes reported in their Political Change in Britain (1969: 216-220) that two-fifths of respondents did not talk about politics with anyone and the overwhelming majority followed the 1964 and 1966 elections on radio and television. Those with jobs were less likely than those without jobs to engage in political discussions, and even in the latter case the proportion of respondents was barely one-quarter.

Robert Dahl, in his Who Governs? (1961: 261-264), presented data that indicate an important cautionary qualification to models based on interpersonal communications. In New Haven only 17% of survey respondents reported that they got "more" information from discussions with people as opposed to the mass media. Moreover, those people who are relatively less susceptible to the influence of others-namely, those who are active in political life-rely to a great extent on personal communications. Dahl further noted that it is in those areas where people have direct contacts and experiences that interpersonal communica- tions are important; for those issues with which people have only indirect experience, which must constitute the bulk of political life, the reliance on the mass media is much greater.

Fitton (1973: 456-470) presented evidence that the- neigh- borhood is not necessarily benign. Less than a majority of his respondents reported street-based friendships, and people on one street reported no perception of friendship relationships on the street. While there were variations between the streets, the general pattern was one of low neighborhood discussions of politics, which resulted from either apathy or a desire to avoid potentially threatening situations. Toomey (1970: 266) argued on the basis of his study of British traditional and new working-class estates that there was no evidence to support the view "that physical mobility and contact with kin exert important influences on working-class norms and values."

The impact of the local political environment need not necessarily be hypothesized as a function of interpersonal communication networks. The long stretch of studies extending as far back as Tingsten's work have presented persuasive evidence that social structures and institutions to some extent either deflect

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 551

or amplify patterns of electoral behavior. For instance, Miller, in his Electoral Dynamics (1977: 65), argued that the "partisanship of individuals is influenced more by where they live than what they do." These formulations of neighborhood effects, it must be noted, contain nothing peculiarly geographical in character other than a variegated pattern of sectoral location of pertinent actors. Indeed, these formulations can be expressed in terms of reference group theory, in which the attitudes, opinions, and behaviors of citizens are influenced strongly by their organizational alle- giances (and to the extent that other allegiances or influences are lacking, the probability for behavior to be amplified in the direc- tion of the dominant group is greater). The same formulations can also be expressed by the pattern of political competition with- in constituencies. Crewe and Payne (1976: 63-68), for instance, despite the fact that their data do not include religion and educa- tion, are able to account for 89.3% of the variance in the percent- age of Labour vote, with six predictors comprised of structural and political variables.

Furthermore, it is rather remarkable-although understand- able, in practical terms-that so many of the studies of neigh- borhood effects commit the very thing which survey analysts have been accused of many times: the wrenching of the individual out of his or her social context (Galtung, 1968: 15). The trauma may not be as great, but nonetheless it still is present when there is a failure to take into account higher-order systems. Many of the studies which have stressed the importance of establishing the contexts of behavior overlook or downplay the fact that in the overwhelming majority of instances of electoral behavior, in- dividuals are subject to a great variety of contextual influences, which arise from the many environments in which electors find themselves. It seems imperative, then, if the discussion about neighborhood effects is to proceed in a productive manner, that note be taken of the possibility of the intrusion of higher-order effects. In light of these remarks attention is now given to an examination of the "nationalization" of electoral responses approach.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

552 POLITICS AND GEOGRAPHY

THE NATIONALIZATION APPROACH

Political scientists have been concerned for some time with the question of regional deviations from national patterns of elec- toral behavior, particularly party vote. The commonplace meth- od of ascertaining such deviations has been the utilization of swing, the measurement of which has taken a number of forms. ' The logic of all forms of swing is the same: subnational deviations from higher-order movements, normally the national trend itself, are taken as evidence of subnational effects; the greater the deviation, the greater the presence of subnational political effects. Two characteristics of swing hamper seriously the analytical ends to which it can be applied. First, the measure is restricted to pairs of elections (although not necessarily consecutive ones), so it is at best only crudely sensitive to cyclical movements or secular changes. Second, although it is intended to provide information about the spatial sources of subnational movements, it is in fact very imprecise about the character of those movements.

Most calculations of swing have been at the level of con- stituencies, although it is not clear why the constituency should necessarily be the site of the greatest subnational variations in behavior. Voters can be conceptualized as being located in a number of environments, at levels lower or higher than the constituency. Even if, say, party vote movements across some group of constituencies are aggregated to create regional values, the problem still remains that the analyst is only comparing one set of scores and their deviations from the overall national pattern. Political forces related to a number of different levels of the system are potential sources of influence on the behavior of electorates, but with swing analyses the relative impact of other levels cannot be ascertained. Indeed, what may be said to be a "regional" or "local" effect may be a function of the impact of higher or lower levels of the political system. The inability of swing measures to be able to handle simultaneously a number of different levels seriously weakens estimates of subnational politi- cal effects.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 553

Stokes's (1965, 1967) partially nested, mixed variance com- ponents model, although initially developed in the course of research into linkages between electors and representatives, is particularly useful in two respects. First, it provides for the impact of both random or time-specific and fixed or timeless effects which, respectively, can be conceptualized as being equivalent to short-term influences, such as issues, the appeal of particular politicians, or the impact of extraordinary cam- paigning, and long-term influences, of which a prominent example would be party identification. Second, it simultaneously provides for the measurement of a number of levels of the political system, normally the national level, an intermediary level (represented by provinces, states, or "regions," and the local level (represented by constituencies or congressional districts).

This model, with variations, has been used to examine electoral behavior in a number of different systems (Aitkin, 1968; Barnett, 1973; Jackman, 1972; Mughan, 1978; Wilson, 1979; Wools- tencroft, 1975). Interpretations of the model's results have revolved around the development from "denationalized" systems to "nationalized" systems. The logic of the model is that uniform movements in party vote and voter turnout across the nation are evidence of the impact of national political forces and will be expressed in relatively high national variance component scores. The remaining variance component scores will be disaggregated in the same fashion. The intermediate variance score provides a measure of the impact of political forces specific to that level, and the local variance component score provides a similar estimate for constituencies. A high intermediate variance component score, after the removal of the national effect, would indicate similar movements within regions and interregional differences. High constituency scores would mean from one constituency to the next, after the removal of higher-order effects, that con- siderable variation remained.

Nationalization of electoral responses is found in the cor- respondence from one constituency to the next of the time- specific movements. In a nationalized system, a gain for a party of 6% in its national vote share would be paralleled by similar

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

554 POLITICS AND GEOGRAPHY

movements for the constituencies, in which case the political system would be interpreted as being dominated by issues, events, personalities, and political organizations which have a national focus and orientation.

In the contrary case, that of a denationalized political system, a national 6% party gain would represent an arithmetic mean of many diverse movements, with a high probability that just as many constituencies registered gains as losses, with considerable variation as well in the magnitudes of the various movements. In such a case the political system would be interpreted as being dominated by issues, events, personalities, and political organi- zations would have strong local roots and orientations.

Richard Katz (1973: 819) raised a number of objections to the Stokes model. First, Katz argued that it is invalid to assume that a national (or, for that matter, regional or local) factor will have a constant impact on some set of voters. Instead,

since different groups may react in opposite ways to the same stimulus, the direction as well as the magnitude of the shift caused by a single factor may vary across [constituencies] because of varying distributions of ethnic, social, political or economic groups.

Furthermore, Katz argued that constituencies vary in the strengths of the parties: In some constituencies a party will do very poorly, while in others it will do very well. Katz in this regard makes an argument similar to that put forward by Berrington (1965: 19) about swing measures of electoral change:

The swing is said to be a swing of x percent of the electorate, or, more precisely of those voting: it is not related to the popula- tion at risk. The electorate, as such, however, cannot swing; it is only former Conservative or Labour voters who can swing by changing their party. (Although, of course, new voters or ab- stainers can create the equivalent of a swing.) A two percent swing to Labour in Rhondda East is not the same as a two percent swing in South Kensington. In Rhondda East, two percent constitutes a fairly high proportion of the former Conservative vote, in South Kensington a small fraction.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 555

Katz, however, in the development of an alternate model of electoral nationalization, did not address himself to the resolu- tion of this problem. Instead he concentrated on the problem of variations in the character of constituencies. His model is similar to the Stokes model inasmuch as it is assumed that political forces associated with different levels of the system determine the nature of movements around the timeless constituency trends. Unlike the Stokes model, however, Katz focuses not on the direction and magnitude of movements but rather on the pattern of covaria- tion. Since Katz had argued that similar forces may result in different reactions from different groups, it followed that it is necessary to measure electoral nationalization in terms of consis- tent patterns of covariation, and denationalization in terms of inconsistent covariation patterns. In other words, time-specific movements contrary to the national time-specific movement may reflect the impact of national political forces just as much as parallel movements do.

Katz has identified a crucial problem with the Stokes model. However, his alternative model introduces additional analytical complications by interpreting dissimilar time-specific movements from similar and dissimilar constituencies in similar terms, while the problem with the Stokes model is that similar movements- which may or may not arise from similar constituencies-are interpreted in similar terms.

During the post-World War II elections, the United Kingdom has been commonly interpreted as a political system charac- terized by national political forces and relatively uniform elec- toral movements. However, for this electoral period the con- stituency-level variance components score for the party vote movements reported by Stokes (1967: 188), although not as strong as the national variance component, is at the level of .40. Clearly, although the total amount of variation was not very large, a considerable portion of that variation was attributable to local political forces.

The explanation for this anomaly may lie within either the structure of the model or the method of aggregating regions. In

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

556 POLITICS AND GEOGRAPHY

the latter case, if the regions that have been created contain too

many heterogeneous movements the model will apportion their

impact to the level of constituencies. In the former case,

attribution of local variance component scores may reflect the

fact that while there are spatial qualities to the location of

constituencies (so that to a certain extent similar constituencies are contiguous), there are still elements of spatial dispersion. To

the extent that this is the case, incongruent movements-which are taken as evidence of the presence of local political factors- may in fact be the response of dissimilar constituencies to similar stimuli.

The argument here is that the Stokes model is preferable to the

Katz model. While the question of whether or not swing or

variance components models should be based on percentage

points or proportions is still inconclusive, refinement of the

Stokes model should provide for the inclusion of measures of

constituency similarity and dissimilarity, particularly in terms of

socioeconomic characteristics. Two alternative strategies are

apparent. One is to control for similarities in the socioeconomic characteristics of constituencies and examine constituency-level time-specific movements. If nonisomorphic patterns are dis-

cerned, then we can attribute the residual variation to the impact of other political forces. The other approach is to divide

constituencies in a system along the lines of their socioeconomic characteristics and treat each set-a number of combinations are

possible-as a "nation," so that it would be possible to control for

the intrusion of spatially separated but otherwise similar con-

stituencies into the interpretation of the intermediary-level variance component scores. In the event that high constituen- cy-or, for that matter, regional-variance component scores are

found, then, again, we would have evidence that political forces

associated with that level are important influences on the ebb and flow of the vote. Persistent patterns of nonisomorphic time-

specific movements in particular will attract the attention of

analysts interested in the processes underlying deviations from nationwide patterns of behavior.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 557

CONCLUSION

In this article some recent contributions to the field of electoral geography have been examined. This examination of a number of different conceptual approaches and analytical techniques has identified some crucial differences, but it is also apparent that there is an underlying continuity of concern: the character and delineation of spatially demarcated subnational political effects and regional patterns of party domination and subordination. While it may not be the case that further work in electoral geography will satisfy those, such as Muir, who are looking for a clear spatial quality to electoral processes, work in the area of electoral nationalization, with its overriding concern for the spatial connections between levels of the system and the various units within each level, should satisfy those, such as McPhail (1974: 191), who perceive much traditional work in electoral geography as aspatial and "ignoring the reality of inter-unit association or spatial auto-correlation."

NOTE

1. There are a number of different ways of calculating swing. The most common is found in the various Nuffield College election studies; see David Butler et al., The British General Election (see References for years). Or what is -known as the "Steed Swing," see Michael Steed (1965). See also Berrington (1965), Hawkes (1969), and Miller (1972). For a criticism of the concept of swing, see Rasmussen (1965). For a discussion of the paradox of swing, see Butler and Stokes (1969: 303-312).

REFERENCES

AITKIN, D. (1968) "Electoral forces in federal politics." Presented to the Tenth Annual Conference of the Australasian Political Studies Association, University of Tasmania.

BARNETT, J. R. (1973) "Scale components in the diffusion of the Danish Communist Party vote, 1920-1964." Geographical Analysis 5 (January): 35-44.

BERRINGTON, H. (1965) "The general election of 1964." J. of the Royal Stat. Society Series A, 128: 17-66.

BERRELSON, B. (1954) Voting. Chicago: Univ. of Chicago Press. BERRY, B.J.L. (1969) Review of B. M. Russett, International Regions and the Inter-

national System. Geographical Rev. 59: 450-451.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

558 POLITICS AND GEOGRAPHY

BLAKE, D. E. (1978) "Constituency contexts and Canadian elections: an exploratory study." Canadian J. of Pol. Sci. 11, 2: 279-307.

BUSTEED, M. A. (1975) Geography and Voting Behaviour. London: Oxford Univ. Press.

BUTLER, D. and D. E. STOKES (1969) Political Change in Britain. London: Mac- millan.

BUTLER, D. et al. (1952, 1955, 1960, 1965, 1966, 1971, 1975) The British General Elec- tion. London: Macmillan.

CAMPBELL, A. et al. (1960) The American Voter. New York: John Wiley. COX, K. R. (1972) "The neighborhood effect in voting response surfaces," pp. 159-

176 in D. C. Sweet (ed.) Models of Urban Structure. Lexington, MA: D. C. Heath. --- (1970) "Residential relocation and political behavior: conceptual model and

empirical tests." Acta Sociologica 13: 40-53. ---(1969) "The voting decision in a spatial context," pp. 81-117 in C. Board et al. (eds.) Progress in Geography, Volume 1. London: Arnold.

---D. R. REYNOLDS and S. ROKKAN (1974) Locational Approaches to Power and Conflict. New York: John Wiley.

CREWE, I. and C. PAYNE (1976) "Another game with nature: an ecological regression model of the British two-party vote ratio in 1970." British J. of Pol. Sci. 6, part I (January): 43-8 1.

DAHL, R. A. (1961) Who Governs? New Haven: Yale Univ. Press. DOGAN, M. and S. ROKKAN (1969) "Introduction," pp. 1-15 in M. Dogan and S.

Rokkan (eds.) Quantitative Ecological Analysis in the Social Sciences. Cambridge, MA: MIT Press.

FITTON, M. (1973) "Neighbourhood and voting: a sociometric examination." British J. of Pol. Sci. 3, part 4 (October): 445-472.

GALTUNG, J. (1968) Theory and Methods of Social Research. London: Heinemann. GOGUEL, F. (1951) Geographie des Elections Francaises de 1890 a 1951. Paris: Colin. HARTSHORNE, R. (1954) "Political geography," pp. 167-225 in P. E. James and C. F.

James (eds.) American Geography: Inventory and Prospect. Syracuse, NY: Syra- cuse Univ. Press.

HAWKES, A. G. (1969) "An approach to the analysis of electoral swing." J. of the Royal Stat. Society Series A, 132: 68-79.

JACKMAN, R. W. (1972) "Political parties, voting, and national integration: the Cana- dian case." Comparative Politics 4 (July): 511-536.

JOHNSTON, R. J. (1977) "People, places and votes: essays in the electoral geography of Australia and New Zealand." University of New England Department of Geography.

KATZ, R. S. (1973) "The attribution of variance in election returns: an alternative mea- surement technique." Amer. Pol. Sci. Rev. 67 (September): 817-828.

KEY, V.O., Jr. (1949) Southern Politics. New York: Random House. KINNEAR, M. (1968) The British Voter. London: B. T. Batsford. KREHBIEL, E. (1916) "Geographic influences in British elections." Geographical Rev.

2 (March): 419-432. KUHN, T. S. (1970) The Structure of Scientific Revolutions. Chicago: Univ. of Chicago

Press. LEWIS, P. F. (1965) "Impact of Negro migration on the electoral geography of Flint,

Michigan, 1932-1962: a cartographic analysis." Annals of the Assn. of Amer. Geog- raphers 55 (March): 1-25.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

Woolstencroft/ELECTORAL GEOGRAPHY 559

McPHAIL, l. R. (1974) "The spatial approach to electoral analysis." Politics (Sydney) (November): 189-194.

MILLER, W. L. (1977) Electoral Dynamics. London: Macmillan. ---(1972) "Measures of electoral change using aggregate data." J. of the royal Stat.

Society Series A, 132: 122-142. MUGHAN, A. (1978) "On measuring national integration in regionally divided socie-

ties." Quality and Quantity 12, 1: 91-98. MUIR, R. (1975) Modern Political Geography. London: Macmillan. PRESCOTT, J.R.V. (1972) Political Geography. London: Methuen.

---(1969) "Electoral studies in political geography," pp. 376-383 in R. E. Kasperson and J. V. Minghi (eds.) The Structure of Political Geography. Chicago: AVC.

---(1959) "The function and methods of electoral geography." Annals of the Assn. of Amer. Geographers 40 (September): 296-304.

PRZEWORSKI, A. and H. TEUNE (1970) The Logic of Comparative Social Inquiry. New York: John Wiley.

RASMUSSEN, J. (1965) "Disutility of the swing Concept in British psephology." Parlia- mentary Affairs (August): 442-457.

REYNOLDS, D. R. (1974) "Spatial contagion in political influence processes," pp. 233-273 in K. R. Cox et al. (eds.) Locational Approaches to Power and Conflict. New York: John Wiley.

--and J. C. ARCHER (1969) "An inquiry into the spatial bases of electoral geo- graphy." Discussion Paper 11, University of Iowa Department of Geography.

ROBERTS, M. C. and K. W. RUMAGE (1965) "The spatial variations in urban left- wing voting in England and Wales, 1951." Annals of the Assn. of Amer. Geographers 55 (March): 161-178.

ROBINSON, W. C. (1950) "Ecological correlations and the behavior of individuals." Amer. Soc. Rev. 15 (June): 351-357.

SIEGFRIED, A. (1947) Geographie Electorale de l'Ardeche. Paris: Colin. STEED, M. (1965) "The results analyzed," pp. 337-359 in D. Butler and A. King, The

British General Election of 1965. London: Macmillan. STOKES, D. E. (1973) "Comment: on the measurement of electoral dynamics." Amer.

Pol. Sci. Rev. 67 (September): 829-831. --- (1967) "Parties and the nationalization of electoral forces," pp. 182-202 in W. N.

Chambers and W. D. Burnham (eds.) The American Party Systems: Stages of Poli- tical Development. New York: Oxford Univ. Press.

--- (1965) "A variance components model of political effects," pp. 61-95 in J. M. Claunch (ed.) Mathematical Applications in Political Science. Dallas: Southern Methodist Press.

TATALOVICH, R. (1975) "'Friends and neighbors' voting: Mississippi, 1943-1973." J. of Politics 37: 807-814.

TAYLOR, P. J. and R. J. JOHNSTON (1979) Geography of Elections. Harmondsworth, England: Penguin.

TINGSTEN, H. (1963) Political Behavior. Totowa, NJ: Bedminster. TOOMEY, D. M. (1970) "The importance of social networks in working class areas."

Urban Studies 7 (October): 259-270.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions

560 POLITICS AND GEOGRAPHY

WILSON, R. J. (1979) "The impact of modernization on B. C. electoral patterns." Pre- sented to the annual meeting of the Canadian Political Science Association, Saska- toon, Saskatchewan, June.

WOOLSTENCROFT, R. P. (1975) "The interplay between geography and politics: the case of Canada, 1953 to 1965." Presented to the annual meeting of the Canadian Political Science Association, Edmonton, Alberta, June.

WRIGHT, J. K. (1932) "Voting habits in the United States. Geographical Rev. 22 (Oc- tober): 66-72.

R. P. Woolstencroft is an Assistant Professor in the Department of Political Science, University of Waterloo, Waterloo, Ontario, Canada, N2L 3GL. His research interests include spatial dimensions of political behavior. Papers which discuss spatial and temporal aspects of Canadian voting behavior have been presented at the annual meetings of the Canadian Political Science Association (1975) and International Political Science Association (1976). Address: Department of Political Science, University of Waterloo, Waterloo, Ontario.

This content downloaded from 130.127.238.233 on Mon, 23 Sep 2013 06:05:04 AMAll use subject to JSTOR Terms and Conditions