Перейти на главную страницу





главная страница | наши сотрудники | фотобанк | контакт
 



  Цели и задачи Центра  
  Текущий комментарий  
  Тема  
  Автор дня  
  Социология и политика  

  Аналитика  
  Социологические исследования  
  Публикации и интервью  
  Новости  


A Spatial Analysis of Ukraine`s 1998 Parliamentary Elections

[30.10.03]

Spatial Analysis of Ukraine`s 1998 Parliamentary Elections1

Melvin J, Hinich

Dept. Of Government, University of Texas at Austin

Valeri Khmelko

Kiev International Institute of Sociology, Kiev-Mohyla Academy

Peter C. Ordeshook

Humanities and Social Sciences, California Institute of Technology

June 1998

Abstract

Ukraine`s March 1998 parliamentary elections ushered in a potentially new era for its transition to democracy. By implementing a national proportional representation system, stronger incentives than had existed previously were established for the formation and consolidation of political parties. Nevertheless, questions remain as to the coherence of the electorate`s political perceptions, the severity of Ukraine`s regional conflicts, and the opportunities for the formation of truly national parties. Using data from a national pre-election survey, this essay applies a statistical methodology designed to assess the coherence of the policy space that voters perceive when evaluating parties and to test the general relevance of a specific model of electoral processes - the spatial model of electoral competition - that has been applied extensively in a variety of other contexts and countries. Our general conclusion is that, despite the brief time parties had to prepare for these elections, voter perceptions are remarkably coherent. And although the "traditional" left-right dimension dominates those perceptions, there is evidence of a second evaluative criteria that corresponds to a preference for reform, but a shunning of more radical nationalist views. However, our analysis suggests that rather than conceptualize the parties and candidates that attempt to appeal to this segment of the electorate as "centrists" or as a compromise between, say, communists and national democrats, this segment is itself a third force in Ukrainian politics that divides pro-reform voters, and it is unclear whether parties can take advantage of this third force to shape a truly national party system and whether the forthcoming presidential election will be anything more than a replay, in one form or another, of the 1994 election contest.

A Spatial Analysis of Ukraine`s 1998 Parliamentary Elections

Melvin J, Hinich

Dept. Of Government, University of Texas at Austin

Valeri Khmelko

Kiev International Institute of Sociology, Kiev-Mohyla Academy

Peter C. Ordeshook

Humanities and Social Sciences, California Institute of Technology

Ukraine`s March 1998 parliamentary elections - the first held under a dual system of representation in which half of the Rada`s 450 seats were filled in single-mandate elections and half by national party-list proportional representation (PR) - presented voters with a confusing array of party lists. Many of the thirty parties and blocks that qualified to have lists placed on the ballot did not exist a few months prior to the balloting and some shared labels that would confuse even the most politically astute voter. Hoping to encourage a coherent party system (and also, doubtlessly, to aid the reelection of those already holding seats in the Rada), Ukraine adopted an electoral procedure that was a near carbon-copy of Russia`s method for filling seats in its State Duma. 2

Encouraged perhaps by the advantages those procedures seemed to confer on Russia`s Communists and other opponents of Yeltsin, the Rada committed to the new election law only months before the actual balloting, thus giving existing parties and political formations a distinct advantage over potential opponents.

Of course, such manipulations are a time-honored democratic tradition, and rather than fret over who is advantaged or disadvantaged by a change in electoral rules, the more important question is whether such change will in fact encourage a coherent political party system and whether it will play any role in ameliorating the conflicts that currently bedevil Ukraine`s transition to democracy. The most evident conflict, in fact, is not only between opponents and proponents of reform, but also the East-West cultural (ethnic, religious, and linguistic) divide that permeates Ukraine`s politics, that threatens its political stability, and that undermines coherent national policy, both domestic and international (Arel and Wilson 1994, Kravchuk and Chudowsky 1996, Khmelko 1997, Wilson 1997).

This essay, though, does not seek to provide a full analysis of the March elections or even of their overall consequences. Instead, our purpose is to test the applicability of a methodology for the analysis of electoral processes that is widely used elsewhere and holds the advantage of being linked directly to a well-developed theoretical paradigm for studying elections - the spatial model of voting. Described extensively elsewhere (Enelow and Hinich 1984, Ordeshook 1987) and applied in a variety of contexts (see, for example, Dow 1997 in Chile, Myagkov and Ordeshook 1998 in Russia, Enelow and Hinich 1994 in the United States,, Hoadley 1986 to the process of political party formation, Lin, et al 1996 in Taiwan, and, for its application to legislative voting, Poole and Rosenthal 1997), this model assumes that:

(1) each criterion a voter uses to evaluate the alternatives he or she confronts can be represented by a segment of the real line;

(2) the policy "issues" that are normally associated with an election campaign can be "mapped" into these criteria in a coherent way - in much the same way as ideology summarizes preferences and perceptions on a vast menu of policies;

(3) voters have "ideal" points on these criteria which, in turn, tell us their preferences on the substantive actionable public issues;

(4) the alternatives over which voters must choose - candidates or parties - can also be represented by points on each dimension;

(5) the preferences of voters over these alternatives is some monotonically decreasing function of the distance between their ideal and the positions of the alternatives;

(6) an individual choice rule, presumably one that derives from the election law that translates votes into seats and thereby, ultimately, into policy, dictates each voter`s ultimate decision in the election (e.g., "vote for the alternative spatially closest to one`s ideal").

This model, originally formulated to bridge the theoretical gap between economics and politics (Downs 1957) has given rise to a considerable theoretical literature that seeks a coherent understanding of such things as the sources of information for voters, the strategic manipulation of uncertainty by candidates, coalitions, nonvoting, strategic voting, the entry and exit of parties, alternative objective functions for candidates and parties, and the consequences of alternative rules for translating votes into seats (e.g., plurality rule, majority rule with runoffs, proportional representation, etc.). However, as simple and general as this structure might seem, we confront a number of methodological problems in its empirical application. Specifically, although expert commentators might have good guesses about the substantive issues that are relevant in a campaign and the approximate positions of the candidates or parties on them, we cannot be certain that these perceptions are shared by the electorate or that they are uniform across all parts of the electorate. More specifically, we know that voters in even mature democracies simplify their choices by simplifying the way they view issues, candidates and parties. Thus, while commentators and those active in day-to-day politics might see a complex nexus of issues and alternative policies, voters typically operate in a far simpler conceptual environment in which many or all of the substantive issues discussed in the mass media map into some smaller set of basic evaluative criteria, such as when voters vote ideologically on the basis of criteria we label "left-right," "liberal-conservative", or."nationalist-internationalist". We can reasonably suppose, in fact, that election campaigns, if they have any effect at all, merely cause them to make minor adjustments in these schemes - moving one candidate or party to the "right", another to the "left," etc. Thus, while journalists and politicians may concentrate on substantive "issues" such as taxes, investment schemes, social welfare policy, a government"s pension and retirement policies, and international relations, voters typically make decisions on the basis of something that simplifies the political discourse that swirls around them, something that simplifies their political universe and helps them makes sense of the often contradictory claims of competing political elites (Hinich and Munger 1994). That "something" is frequently a thing we call ideology, but rather than assume that we know the precise structure of this scheme, its dimensionality, how the different issues of the campaign map into it, where voters place candidates and parties in it, and how voters use this scheme to estimate the policies of candidates and parties on the substantive issues that most concern them, these things require statistical estimation through the application of appropriate formal methodologies.

Identifying whether voters in Ukraine can be described as if they possess such a scheme - ascertaining whether there is coherence to their perceptions of parties, candidates, and campaign issues - is important for a number of reasons. First, we want to learn whether voters there are much different than their counterparts in other democracies. Did the Soviet "experience" leave them in a political vacuum so that they can be easily swayed by a candidate"s or party`s appeal on a single issue and that makes them especially vulnerable to political demagoguery, or do they posses a conceptual scheme that acts as a buffer between vote choice and campaign rhetoric? Given, moreover, the unsettled nature of Ukraine`s party system (if it can be called a "system"), are voters there able to form a coherent map of the policy predispositions of the parties that vie for their support? And perhaps most importantly, is the East-West divide in Ukraine so deep that voters from one geographic region see a different political universe than voters from the other?

Any number of methods have, in fact, been used to estimate the structure of voter preferences and beliefs, including various adaptations of discriminant analysis, factor analysis, and so on. However, to the extent that the formal structure of these methods is not directly linked to a formal mathematical model of preference and choice (like the one used, for instance, in contemporary economic theory), they are at best stopgap measures in the pursuit of statistically justifiable estimates. Fortunately, the requisite methodology exists, and its application is this essay"s focus. Briefly, then, Section 1 outlines that methodology and its connections to the underlying theory of spatial voting. Section 2 describes te data to which this method will be applied, and Section 3 offers an initial spatial assessment of the Ukranian electorate prior to the March 1998 balloting. Section 4 then pursues a number of related matters, including an assessment of whether there are any detectable differences in the perceptions of voters in the Eastern versus the Western halves of the country, and a comparison between final election returns and the predictions derived from our analysis.

1: Model, Methods and Data

The formal presentation of a spatial voting model begins necessarily with some notation:

k: - the number of "criteria" voters use to evaluate the candidates or parties and their positions on the actionable, salient policy issues of the day;

X*i = (x*i1,xi2,...,xik): voter i's ideal position on criteria 1 through k;

a = (a1,a2,...,ak):candidate or party a's position on each of the k criteria, with similar notation for candidate or parties b, c, etc.

ui (a) = f(xi,a):the utility voter i associates with candidate or party a, where the function f is monotonically decreasing with the distance between x*i and a and where i prefers a to b if ui (a) > ui (b).

There are, of course, a great many alternative metrics for defining the distance between two points, but much of the theoretical literature assumes that distance here is the usual simple Euclidean metric, so that in matrix notation

distance(x* i , a) = x* i - a2 = (x* i - a)'A(x* i - a)

where A is a k-by-k matrix that represents the weights (salience) of the different criteria as well as the possible interdependence of criteria in a voter`s mind. Notice, though, that if A is the same for all voters, we can transform the issue space so as to make A the identity matrix, I, in which case our expression for utility becomes:3

ui (a) = f(distance(x*i, a)) = f(x*i - a2) = f(x*i'x*i - 2x*i'a + a'a)

This is the model of voters that serves as the basis for much of the formal theorizing about election processes. Nevertheless, statistical estimation of its parameters - the ideal points of voters, the positions of the candidates, and the dimensionality of the space - requires additional simplification, chief among them being an assumption about the functional form of f (in much the same was as regression analysis cannot proceed without a specification of the functional form of the relationship between independent and dependent variables). And again, such estimation proceeds most straightforwardly if we let f take a specific linear form, namely:

ui (a) = Lia - x*i'x*i + 2x*i'a - a'a

where L is a constant that scales each voter`s utility function, and, for purposes of this presentation, can be assumed to be equal to zero. To this equation we can add one other consideration that often plays an important role in elections; namely, the possibility that there are criteria for which all voters have the same preferences - the same ideal points - but along which the candidates or parties are perceived to vary. These valence issues correspond to things like a candidate's reputation for honesty or, in the event that all voters prefer to avoid uncertainty, the ambiguity associated with a candidate or party`s policy positions. Letting q denote the importance of such an issue relative to the other criteria of choice, and Va be candidate or party a's perceived position on a relevant valence issue, then ui(a) becomes

ui (a) = - x*i'x*i + 2x*i'a - a'a - qVa

The final step in generating an estimatable statistical model is to notice that if we arbitrarily choose one candidate or party - say o - and set the origin of the issue space at that candidate's position (i.e., let o = (0,0,...,0)), then after subtracting ui(o) from ui(a), we have an equation that is strictly linear in the ideal points of each voter.4 That is,

ui (a) - ui(o) = 2x*i'a - a'a - q(Va - Vo)

This is the model we use for estimating the dimensionality of the criteria space, the x*i's, the candidate or party positions a, b, c, and so on, and, if necessary, the relative salience of the valence dimension, q. Before proceeding, however, notice the unstated assumptions of this model. First, we assume that all respondents used in the data analysis share a common space. This is an assumption we later want to reconsider by examining the possibility that respondents in Eastern Ukraine do not evaluate parties using the same criteria as respondents in Western Ukraine - the possibility that there are fundamental differences between these subpopulations in terms of how they perceive the arena of party electoral conflict. Second, notice that we do not subscript party or candidate positions - all respondents are assumed to hold the same perception of these positions on each criterion. Again, the viability of this assumption when looking at Ukraine's different regions is one we want to assess.

There are other purely statistical assumptions that must be imposed in order to proceed that concern the nature of stochastic variables, but here it is best to simply refer the reader to the original sources that elaborate on the design of this methodology and to proceed instead to the data to which this model is applied.5 Briefly, that data, taken from the All Ukranian Survey conducted by the Kiev International Institute of Sociology (KIIS) approximately one month prior to the actual balloting, focuses on a specific set of questions designed to secure an estimate of each voters cardinal evaluation of parties - each respondent's utility for party a, b, c, etc. Taking the ten parties that were assumed to be the strongest contenders, each respondent was asked the following question:

I will name some parties and I will ask you to tell me how the election of this party will affect your family's welfare, positively or negatively, and to what extent. For responses, please use the following criteria:

7:very good

6:mostly good

5:more good than bad

4:partly good and partly bad

3:more bad than good

2:mostly bad

1:very bad

where the parties each respondent was asked to evaluate was limited to the following list:

Agrarian Party (Ag)

Block of Democratic Parties (NEP)

Socialist and Peasant Block (SP)

Communist Party (C)

People's Democratic Party (PDP)

Rukh (R)

Hromada (H)

The Green Party (G)

Social Democratic Party (SDP)

United Social Democratic Party (USDP)

Even if we accept all of the assumptions necessary to justify theoretically our estimation procedures, there are a number of practical issues that need to be addressed. First, as we note earlier, many of the parties and blocks that ultimately competed were unlikely to be known to many voters, having formed only in time to register for inclusion on the ballot. As a consequence, many respondents failed to grade all the parties or, to avoid not responding to the questionnaire, simply gave all or nearly all parties the same score. Since the corresponding data contains little or no information, our analysis proceeds thus: First, rather than simply eliminate a respondent if he or she failed to grade one or more parties, unscaled parties are first assigned the "neutral" score of 4. The second step is to eliminate a respondent if he or she failed to distinguish (give different scores) to at least three parties. For example, then, if a respondent gave, say, the Communists a score of 6, gave Social Democrats a score of 4 and failed to grade all other parties, then that respondent is eliminated from the analysis (since ungraded parties are also assigned a score of 4, in which case this respondent is coded as having distinguished only two parties). Of course, this procedure biases the estimated location of parties that are relatively unknown in the direction of the center of the issue space - near the mean population preference. However, even though we prefer to avoid deliberately introducing any biases into the analysis, the alternative is to severely limit the size of the sample to be analyzed, thereby increasing the instability of our other parameter estimates.6

Even with this procedure, our sample population is reduced considerably. First, of the 2923 respondents interviewed, 716 (24.4%) stated that they would not vote, and these respondents were not asked to evaluate the parties. Of the remaining respondents, 1058 were eliminated as having failed to grade or distinguish among a sufficient number of parties, leaving us with an analyzable sample of 1149 respondents. There is, though, a partial external check on our procedures, methodology, and even the general relevance of the spatial model to the Ukranian context. Specifically, of the 2207 respondents who did not rule out the possibility of voting, approximately 28% failed to indicate a party preference when asked "if the election were held tomorrow, for which party or block would you vote?" One application of our methodology, then, is to take these potential voters and, on the basis of our estimates of their ideal points and the candidate positions, identify the party closest to their ideals. Adding these respondents to those who stated a party preference allows an overall estimate of the likely vote for each of the ten parties respondents graded, and comparing these estimates to the actual election returns gives us a sense of the nature of the error, if any, in our methodology.

2. A First Analysis

The first step in our analysis is to simply recover the spatial positions of the ten parties graded by our respondents. Briefly, without attempting to accommodate the possible existence of a valence dimension, Figure 1a presents the two-dimensional recovery of those positions (notice that here, and throughout this essay, we center the coordinate axes at the mean preference of the respondents), while Figure 1b portrays the distribution of estimated ideal points in this space.7 Looking a Figure 1a, there is little doubt that insofar as representing party positions is concerned, this recovery appears to be largely one-dimensional (depending on details, the variance explained by the first eigenvalue generally ranges between 80 and 90%). This dimension, moreover, appears to correspond to the classical left-right, liberal-conservative, pro-reform versus anti-reform axis that distinguishes opposing political camps in the early stages of democratization throughout the former USSR.8 At one extreme (the right in the orientation of our figure) we find the Communist Party, whereas at the other extreme we find the nationalist and pro-reform Rukh. The anti-reform Socialist-Peasants Block is near the Communists, whereas the center is occupied by a cluster of pro-reform parties with arguably less distinctive characters than Rukh or the Communists.

However, despite the predominance of this first dimension, the issue space is not unidimensional. First, the second dimension distinguishes the two social democratic parties from the other centrist ones. Second, notice in Figure 1b that ideal points are not unidimensionally distributed - a second dimension is required to capture the dispersion of preferences, especially in the "po-reform" half of the issue space. Finally, we note that if we in fact try only a 1-dimensional recovery with our methodology, the overall variance explained in the estimation of parameters drops from 59% to 7%. Thus, two dimensions are required to represent the evaluations of parties and blocks.

To check our interpretation of the first dimension and to get a better understanding of the second (vertical) dimension, respondents were asked about their attitudes on a number of policy matters ranging from language to property rights. More specifically, respondents were asked the following:

Please look at the following list of goals of the different parties and tell me whether, if you had to vote, for what goals would you rather vote "for" and what you would rather vote "against":

Q1.Renaissance of the Ukranian nation

Q2.Restoration of the USSR

Q3.Restoration of central planning

Q4.State support of the main spheres of industry

Q5.Defense of human rights, political freedoms, etc.

Q6.Order in the country and fighting corruption by the firmest means

Q7.Unification of Ukraine with Russia and Belarus

Q8.Speedy market reforms

Q9.Social security for low wage earners, but with market reform

Q10.Maintenance of the Ukranian state"s independence

Calculating the mean ideal points of respondents who answered "vote for" and "vote against" for each of these questions, Figures 2 and 3 graph the locations of these means using the same coordinates and scale as Figure 1a.9 Notice that with the exception of questions pertaining to human rights (#5 and, arguably, #6), all responses correlate highly with the horizontal dimension. But in addition, respondents were also asked several specific questions concerning reform policy and the matter of Ukraine`s relationship to Russia:10

There are different ideas concerning how to improve the situation in the economy and the welfare of the people. I will read several statements and ask you to indicate the extent to which you agree or disagree with them:

Q11.To improve the welfare of a majority of the people there is a need to renew the state socialist economy without private entrepreneurship

Q12.The primary thing in the development of the welfare of the people is the development of private entrepreneurship

Q13.Citizens must have the right to private property of land - the right to own, buy, and sell land.

In the last question we examine, respondents were asked:

Q14.What kind of relationship would you prefer to see between Russia and Ukraine?

where the allowable answers to this last question (in addition to "don"t know") were: (1) Same as with all other states, with closed borders; (2) Ukraine and Russia must be independent states but with open borders; and (3) Ukraine and Russia should be united as one state.

Figure 4 graphs the mean preferences of respondents for the various answers to these questions and shows, as in Figures 2 and 3, the nearly perfect left-right orientation of those means. Figures 2, 3 and 4 together, though, do give us a hint as to the possible substantive content of the second dimension. Notice, in particular, that aside from the two questions that concern rights - #5 and #6 - the responses with the greatest positive slope are those that pertain, one way or another, to Ukranian nationalism. The questions that focus exclusively on market reform and state control of enterprises line up essentially perfectly horizontally. This is most readily seen in Figures 2 and 3 and the responses to questions Q3, Q4, Q8 and Q9, where the aveage slopes of the four lines is zero. On the other hand, looking at questions Q1, Q2, Q7, and Q10 - those that pertain to relations between Ukraine and Russia or to the independence of Ukraine - we see that although responses here correlate predominantly with the first dimension, they also load on the vertical one as demonstrated by the decidedly positive slopes of the lines in those figures. Thus, although the issues of reform, Ukranian sovereignty, and relations with Russia correlate strongly, there is evidence of their separability.

It is interesting, then, to speculate about the meaning and importance of the second (vertical) dimension, and in this respect notice the greater dispersion of estimated ideal points to the left of the vertical coordinate axis in Figure 1b (that is, notice the approximate triangular shape of the distribution of estimated ideals). This fact suggests that although the "anti-reform" respondents to the right of the distribution do not have preferences that vary much on the second dimension, those to the left are divided on this second criterion. If we suppose, for the moment, that political competition in Ukraine to this point has combined two issues - with pro-reform attitudes and nationalism combining at one extreme, and anti-reform and pro-Soviet attitudes at the other - Figure 1b suggests that the door is open for parties and candidates that support reform in some fashion but that are identified with a more moderate stance on nationalist issues. That is, the variance in the distribution of ideal points suggests that there is room for pro-reform parties that do not adopt an extreme nationalist platform, that are internationalist in their outlook, and that accept the possibility that Ukraine`s relations with Russia must be somehow different than its relations with other countries, including open borders so as to allow for the free flow of people between the two states.

Of course, the existence of three "forces" in Ukrainian politics has been noted in earlier elections (see, for instance, Arel and Wilson 1994, Bojcun 1995) - leftists (Communists and others opposed to reform), national democrats (Rukh and its allies), and "liberals" (those who are not opposed to reform in the context of a strategic economic partnership with Russia and who, being Russophones themselves, oppose any severe policies limiting use of the Russian language). However, rather than view liberals as a group that lies between nationalists and communists - as a political force of the "center" - our analysis suggests that they are more properly placed as the third vertex of a triangle in a two-dimensional space. Note, in fact, that if we continue to adhere to the view of parties such as SDP, USDP, Greens, Hromada, and NEP as constituting some sort of "center", this center is itself divided, with the SDP and USDP distinct from the rest.

To see things differently, consider Figures 5a-5d, which graph the estimated ideal points of respondents after filtering for various responses to several questions pertaining to nationalism and economic reform. Notice first that all four distributions are essentially equivalent, with the mass located to the left of center on the first dimension, but around the center on the second. The questions used as filters here, though, are distinct, with Figure 5a looking only at respondents opposed to making Russian an official state language, Figure 5b pertaining only to those respondents who oppose Ukraine`s unification with Russia and Belarus, Figure 5c graphing the ideal points of respondents opposed to the resurrection of a state socialist economy, and Figure 5d graphing the ideal points of respondents strongly in favor of private land ownership. But while these figures are quite similar, there are some potentially important differences. Specifically, notice that there are many more ideal points to the right of the vertical coordinate axis in Figure 5d compared to the rest. Thus, although voters to the left of the vertical axis are more likely to oppose Russian as an official language, the renaissance of the USSR, and a resurrection of the socialist economy than are voters at the opposite end of the ideological dimension, the issue of private property less cleanly divides the population. Indeed, although those favoring private land ownership are, on average, somewhat further to the left than the population as a whole (their mean preference on the first dimension is -.22 versus 0 in Figure 1b), they are not quite as far left as those graphed in Figures 5a-5c (the respective means on the first dimension are -.54, -.54, and -.43). Notice also the slightly greater concentration of ideal points in the upper left quadrant of Figures 5b-5d as compared to Figure 5a, which suggests the existence of a subset of voters who support reform, oppose reunification with Russia and Belarus, but who are not opposed to letting Russian be used on an equal basis with Ukranian. Now look at Figures 6a and 6b, which graph the ideal points of respondents who ranked Rukh as their first electoral choice and those who ranked Social Democrats (SDP) or the United Social Democrats (USDP) first. These parties, then, divide the pro-reform vote. But notice also in Figure 6b that some of the support for SDP and USDP comes from the upper right quadrant, which, as we show later, lies in that half of the space from which the Communists garner their support.

This analysis suggests, then, that what we and others label "liberals" or "centrists" and are identified here as supporters of the SDP and USDP should not be viewed simply as a compromise between Communists and nationalists or national democrats, but rather as a semi-independent third force in Ukrainian politics. Indeed, as we see later (see Figure 12b), a significant share of voters with ideal points in this upper left quadrant state that they would never vote for Rukh or the Communists. A true center, occupied by parties such as the Greens, NEP, Hromada, and PDP remains highly fractured, and it remains to be seen whether any such party can successfully negotiate a compromise of the three primary contending electoral forces. Nevertheless, in light of the fact that the Ukrainian electorate had so little time in which to make clear distinctions between parties, our analysis does point to the potential for such parties, especially if they can find a balance between the desire for reform and the desire for maintenance of some minimal social safety net for those who fail to prosper under the new political-economic regime. Before we speculate further, though, we need to consider directly the political divisions within Ukraine that correlate so strongly with geography.

2. Regional Differences

The issue that concerns any student of Ukranian politics is that of the country"s east-west divide, and the question that concerns us here is whether these two regions or populations hold different perceptions of the political world - of the issue space that serves as the context for political competition. The existence of such differences would be an ominous note with respect to the potential stability of the Ukranian state since it would mean that there is little opportunity for parties to form and compete nationally in such a way as to help bridge the existing divide. Parties would necessarily and permanently be regionally based. Of course, there already is a bias that favors nationalist parties in the west and anti-reform (communist) ones in the east, but any perceptual divide in the population would mean that centrist parties seeking to compete nationally by advocating policies that might appeal across Ukraine`s geographic divide would have little hope of success.

However, before we look at this east-west divide, we should note first that whatever differences in perceptions, preferences and opinions exist, they are not related to any urban-rural distinction. Figures 7a and 7b plot the estimated ideal points of "urban" and "rural" respondents (those living in cities with a population greater than 500,000 versus those living in villages or settlements), and we see here little difference in these distributions. In fact, rural or urban, Ukraine`s citizens look similar in terms of their spatial position on the first dimension; but interestingly, this distinction only applies to the first dimension, and only to the urban-rural distinction. To see what we mean, Table 1 reports the average ideal point on both dimensions for several standard socio-economic measures, and the first thing to notice is that although there is no systematic variation on the first dimension as we move from rural villagers to "big-city" residents, there is discernable shift in average ideal point positions on the second dimension. Moreover, it is only the rural-urban distinction and language that yield any separation on the second dimension, with urban residents and Russian-speakers scoring highest on that dimension.11

The urban rural distinction, though, is the only one that fails to yield a meaningful separation of estimated ideals on the first dimension; all other traditional socio-economic variables yield a pattern that accords with initial expectations. Specifically, the young, the educated, those who speak Ukranian and who identify themselves a Ukrainians, or who do not

Table 1: Estimated Mean Ideal Points for Various Subpopulations

 issue 1issue 2n
urban-rural   
Village-.11 -.12343
Settlement and City < 20,000 .06 -.15158
City < 250,000 .04 .06150
City < 1,000,000 .08 .03298
City > 1,000,000 .00 .23196
age   
18-39 years-.23 .03365
40-49-.09-.07233
50-59 .09 .07167
>59 .25-.01372
Education   
< 10 years .24-.17172
Some additional .01 .01684
Some college or more-.16 .09289
Self-reported national identity   
Ukranian-.09-.04893
Russian .40 .15204
Respondent’s Language used in survey   
Predominantly Ukranian-.22-.18564
Predominantly Russian .21 .18503
Self-described financial status   
not enough for food .26-.04459
enough for food but not clothes-.14 .03543
enough for food and clothes-.31 .01144

describe their financial status as impoverished, scale to the left on the first dimension; whereas the older subpopulation, the less well educated, those who answered our questionnaire in Russian, who describe themselves as Russian, or who characterize their economic status as impoverished, scale to the right. And although the sharpest distinction seems to be offered by self-reported economic status, the differences here are too slight (a total of.57 units versus.49 for both age and self-reported national identity) to be significant. At the very least, we can say only that although economic circumstances may be an important determinant of political party preference and ideological orientation, language and self-reported ethnic identity are equally important, and any attempt to disentangle this causal nexus requires data other that what is available to us in our survey (for evidence of the relatively greater importance of economic issues see Kravchuk and Chudowsky 1996).

Perhaps no classification of Ukraine`s population, though, generates the greatest variation in ideal points than that of geography. But before we consider the differences here, we must make certain that the same spatial structure applies to both parts of Ukraine. We begin, then, with Figures 8a and 8b, which graph the recovered candidate positions when we look only at respondents living in the nine western-most (n = 391) and eastern-most (n = 563) regions.12 The first thing to notice is that, aside from the location of the parties relative to the origin and, thus, relative to the mean preferences of the populations, we estimate essentially the same two-dimensional maps in both cases, with the parties" relative positions nearly the same in both figures. Thus, although east and west hold different preferences, with the west as expected more favorably disposed to Rukh and the east more favorably disposed to the Communists and Socialist-Peasant"s parties, both halves of Ukraine perceive their political world in identical terms - largely one dimensional with the Communists anchoring down one extreme and Rukh the other, but with the necessity for inclusion of a second dimension in order to secure reasonably accurate estimates of ideal points and party positions.

But while the eastern and western halves of the country share the same "ideological" space, their preferences vary greatly within this space. Figure 9 graphs the mean ideal point of respondents for each of Ukraine`s 26 regions (oblasts), and offers a map that is in fact not much different from that of Ukraine itself - with western oblasts falling largely to the left of the center and eastern oblasts to the right (notice that the scale of this figure is different from all others in order to distinguish among regions).13 For example, observations #11, 14, 18 and 20, which anchor the left-most part of the distribution, correspond to the western regions of Irano-Frankivsk, Lviv, Rivne, and Ternopil, whereas observations #1, 12, and 13, which anchor the right-most part of the distribution, correspond to Crimea, Kirovohrad, and Luhansk. Somewhat less extreme in their rightist positions, but nevertheless firmly in the pro-Communist, pro-Socialist-Peasant quadrants of the issue space are the eastern regions of Donetsk (#7), Zaporizhzya (#10), Kharkiv (#21), and Dnepropetrovsk (#6). The city of Kiev (#2) and Kiev oblast (#3) are more centrally located, although, in keeping with their recent political history, still to the left of center. Only two regions seem misplaced in this figure - Khmelnytsy (#23) and Chernivtsi (#25) - and it both instances it is reasonable to suppose that the relatively small samples from those regions yield less reliable estimates than the others (the average sample sizes are, respectively, 23 and 16, versus an average of 46 per region for all other regions).14

Of course, Ukraine`s regional political differences are well documented (Kuzio 1997, Kravchuk and Chudowsky 1996, Arel and Wilson 1994, and Khmelko 1996, Wilson 1997), and it is in fact possible to find pairs of regions whose populations show little or no overlap in the distribution of ideal points. Figures 10a and 10b illustrate two extreme cases - Crimea and, in the west, Ivano-Frankivsk (the respective means of these populations are (.62,.58) and (-.91,-.73)). These two regions, though, appear to be special cases: Although we can find western regions with the vast majority of estimated ideals in the lower-left quadrant (e.g., Ternopil and Rivne), most western regions and essentially all eastern regions have a "reasonable" share of ideal points in the upper-left quadrant. Moreover, we should not place undue emphasis on the ideal point distributions estimated for individual regions. We are generally dealing with small samples (< 40) and, owing to the few degrees of freedom available when estimating individual ideals, our estimates are subject to considerable error. Figures 11a and 11b, then, graph the estimated ideals for the nine eastern-most and nine western-most oblasts and we see first, and as expected, that the density of ideal point to the right of the vertical coordinate axis is far greater in the eastern regions than in the west. Second and also as expected, the density of ideals in the lower left quadrant is greatest for the western regions. Finally, though, both east and west regions offer approximately equivalent proportions of ideals in the upper left quadrant - in that quadrant we interpret as belonging to those voters who are not opposed to reform, but who reject or are at least uncomfortable with extreme nationalist appeals. Thus, although Figures 10a, 10b, 11a, and 11b show the regional disparities that exist in Ukraine, Figures 11a and 11b also reveal the existence of an electorate that is not wholly divided into two "warring" camps. Thus, rather than suppose that the failure of "centrist" parties to successfully garner support from all regions of the country is due to some bipolarity in the electorate, we might speculate that this failure derives instead from some failures on the part of these parties to mount effective campaigns that realize the full potential of this "compromise" vote.

3. Patterns of Party Support

To better assess this latter possibility, let us turn our attention to the practical matter of predicting the election outcome. Of course, it is by now well-understood that if one wants to predict how a person will vote, the simplest and most accurate approach is to ask the question "If the election were held today, for whom would you vote." The survey from which our data is taken asked such a question, and the results are not radically different from final vote tallies. Table 2 contrasts the final outcome against the percentages yielded by our survey from among those who answered this question with a definitive party choice for the ten parties that are the focus of this study. Thus, although the Communist vote is underestimated and SDP and USDP"s vote over-estimated, the poll does remarkably well, especially if we keep in mind that the question was asked nearly a month before the actual balloting.

Table 2: Actual and Predicted Vote Totals

 Actual percentagePercentage in Poll
Agrarians 3.67% 0.9%
NEP 1.23 3.3
Socialist-Peasants 8.64 5.8
Communist24.6819.4
PDP 4.99 2.9
Rukh 9.40 9.3
Hromada 4.68 2.1
Greens 5.46 6.0
SDP 0.32 1.5
USDP 4.02 7.9

One difficulty, though, with relying on polls here is that a large part of the sample refused or could not identify a preferred party - approximately 28% - and given how close many parties are to meeting the 4% threshold for representation, it is important to try to see if we can use the methods and perspectives of spatial analysis to predict how such respondents might actually cast their ballots if they in fact choose to vote. We realize that there are other more accurate means of making these predictions, but our objective here is to examine the performance of our spatial model and its relevance to the Ukrainian electorate. To that end, recall that earlier, in Figures 6a and 6b, we show the near perfect separation between voters who support Rukh and either the SDP or USDP. Figure 12a looks at the voters who intend to vote Communist and shows their separation from both of these groups. Of course, this separation is expected since the questions upon which our spatial maps are built in effect ask respondents to identify their most preferred party. Figure 12b, though, is perhaps more interesting in that it shows that voters who would never support either the Communists or Rukh lie in that quadrant where SDP and USDP"s support is concentrated, thereby reenforcing our earlier inference that the multi-dimensionality we recover is the result of voters who have rejected a return to the past, but also appear to reject Rukh"s nationalist appeals. The issue that remains, though, is understanding why SDP and USDP failed to garner the support suggested by our poll.

First, to see if we can refine the predictions in Table 1 using our spatial methodology, recall that of the 1149 "scalable" respondents, 172 (15%) did not reveal a party preference. Our first step, then, is to predict the votes of these respondents using our estimates of their ideal points and the parties" locations. One difficulty here, though, is that we can estimate the party preference of respondents only for the ten parties for which we have data, so since these ten parties account for 68.7% of the choices in our sample, we proceed with the assumption that the same percentage of voters among those indicating no party preference will vote for one of these parties, with the remaining 31.3% voting for other party lists. Using that as the basis of an initial prediction, Table 3 reports the percentage breakdown for the respondents in our subsample of scalable responses. Briefly, column (A) reports the final election outcome, column (B) gives the percentage breakdown for the full sample of respondents, and column (C) gives the percentage breakdown among those voters with scalable preferences - those in our sample of 1149 respondents who stated a party preference. Immediately we can see here that our full sample as well as our scalable subsample overestimate the vote of the SDP and USDP but underestimate that of the Communist Party. Column (D) now uses the estimated ideals and party positions to predict the likely votes of those 172 scalable respondents who failed to state a party preference and column (E) gives the final percentage breakdown if we add these estimates to those in column (C).

The differences in percentages between the first column of numbers (the actual percentage of the vote received by the ten parties in question) and column E (our estimates based on the spatial map) all seem within the range of normal statistical sampling errors, but with two apparent exceptions - the SDP and USDP. The relatively large over-estimate for these two parties derives from the fact that in our spatial maps at least, they alone occupy the upper left quadrant of the issue space and, thereby, are awarded the lion"s share of voters with estimated ideal points in this quadrant, and the only explanation left open to us at this point for this discrepancy - and explanation that can only be judged by expert commentators - is that of these parties ran ineffective campaigns that failed to garner the support otherwise available to them.15

We should not, though, wholly overlook the discrepancies in our other predictions, and to that end notice that thus far we have not made any use of the idea of a valence issue

Table 3: Estimating the Vote of Respondents Who Failed to Give a Party Preference

 (A) Actual % received(B) % vote among those stating a prefe-rence, full sample16(C) % vote among scalable respond. with a party prefe-rence (n=977)(D) estimated vote among scalable respond. without a prefe-rence (n = 172)(E) overall estimate of % of the vote looking only at scalable voters17 (F) est. vote among scalable resp. without a prefe-rence using valence issue (n=172)(G) overall estimate of % of vote looking only at scalable voters, using valence issue
Agrarian 3.67% 0.8 0.8%13.3% 2.1%15.1% 2.2
NEP 1.23 3.3 3.8 6.4 3.9 4.7 3.7
Socialist-Peasant 8.64 5.8 6.8 6.4 6.5 7.6 6.6
Communist24.6819.520.921.520.022.120.1
PDP 4.99 2.9 3.5 3.5 3.3 5.8 3.6
Rukh 9.40 9.312.110.511.4 9.311.2
Hromada 4.68 2.1 2.4 3.5 2.4 3.5 2.4
Greens 5.46 6.0 6.0 4.7 5.6 7.6 5.9
SDP 0.32 1.5 2.217.4 3.715.7 3.5
USDP 4.02 7.910.212.810.1 8.7 9.6
Other32.9140.931.9 -31.9 31.9

an issue for which there is no variation in voter ideal points, but only in the perceptions of individual parties.18 The difficulty here, though, is that the presence and identity of such an issue must be come from sources other than our spatial methodology - from the analyst"s understanding of the substantive context of the election. In this case, though, a reasonable guess as to what might constitute a valence dimension is the uncertainty voters associate with the different parties. Not all of the ten parties in our sample are well-known to voters, and we might suppose that, ceteris paribus, voters would prefer a known quantity to an unknown one.

It is possible, of course, that some voters, weary of the conflicts between established parties and political figures, and weary as well of Ukraine`s seeming inability to foster a productive economy, might prefer to "toss the coin" and give an unknown party the chance to lead. Nevertheless, to assess the possibility that voters prefer less uncertainty, we can construct an "index of uncertainty" by counting the proportion of times respondents were unable to grade a party on the questions used to generate our spatial maps. Normalizing these scores to the range 0 to 1, we can then reestimate individual ideal points and party locations assuming the existence of this valence issue.

Rather than report the full results of this exercise (with which, admittedly, we are not wholly satisfied, given the ad hoc nature of our index of uncertainty), column (F) in Table 3 gives the revised estimates of the share of the vote each party would now receive after we incorporate the valence issue into our analysis, and one thing in particular is worthy of note when this column is compared to our original estimate without a valence dimension, column (D). Specifically, notice that in every instance but one (the Green Party, although the magnitude of the difference here is too negligible to be significant), the vote of each party is increased or decreased in the direction of the actual returns. Thus, it would appear that uncertainty played a role in the March parliamentary elections, with voters preferring greater certainty to less. That is, our analysis is consistent with the hypothesis that, ceteris paribus, voters, like voters in nearly all other electorates with which we are familiar, preferred "the devil they knew to the devil they did not know".

It is tempting, now, to extrapolate these results to all respondents in our sample, including those we cannot scale. However, there is little reason to suppose these two sub-samples (scalable and non-scalable respondents) are comparable. Indeed, there are reasons for supposing they are not comparable: specifically, all respondents were asked "which political forces are best able to drag Ukraine from crisis", and 35.7% of scalable respondents answered "difficult to say" (16.4%) or "no one" (19.3%), whereas more than half of all nonscalable respondents - fully 60.5% - gave one of these two responses (32.9% and 27.6% respectively). A more theoretically satisfying approach, then, is estimate the econometric relationship between the ideal points of scalable respondents and their socio-economic characteristics (SEC)- their age, incomes, education, and so on - and to use this relationship to predict the ideal points of nonscalable respondents. These predictions, in combination with our estimates of party spatial positions, can then be used to predict party preferences. The advantage of this approach, of course, is that it allows us to predict the vote shares of parties knowing only the party`s positions and the socio-economic characteristics of the population. Unfortunately, the problem here is that the few SEC measures in our poll predict ideal points rather poorly. For the 1149 scalable respondents, the variables Sex, Age, Education, self-identification of Nationality (Russian or Ukranian), self-reported household financial status, a classification of residence into three region (east, central, west), and size of city or village of residence predict a paltry 20% of the variance of ideal points along the first dimension and only 4% along the second. It is impossible, then, to have any confidence in the resulting estimated parameters.

A variant of this approach is to add some opinion questions from our survey that were asked of all respondents, and here we somewhat arbitrarily (for purposes of an initial analysis) select three: (1) whether respondents agree or disagree (strongly, weakly, etc.) with the statement "the primary thing in the development of the welfare of the people is the development of private entrepreneurship"; (2) the "kind of relationship [the respondent] prefers to see between Russia and Ukraine"; and (3) "he political forces that are best able to drag Ukraine from crisis" Communist-socialist, social democrats, centrists, national democrats, radical nationalists).

One methodological problem, here, is that not all respondents answered all the questions or gave codeable responses, and if any respondent who failed to answer any one of these questions - SES or opinion - is dropped from the analysis, we are left with an uncomfortably small sample. Our approach, then, is to take this limited sample and form a preliminary estimate of the relationship between the variables and our estimated ideal points. This preliminary estimate, though, is used only to identify relevant variables so that we can filter on a smaller subset of independent variables and, thereby, drop fewer respondents from the analysis. This procedure, applied to a subsample of 1064 respondents out of 1149 suggests that we drop all SES variables except age, self-identified nationality, self-identified family financial status, size of city of residence, and region of residence. Sex, level of education, and the language used by respondents when interviewed had insignificant coefficients in our initial estimates. Moreover, not all variables were relevant in estimating the ideal points on both dimensions, and our subsequent estimates look only at the variables that were relevant in this initial pass through the data. Table 4, then, gives the estimated relationship between ideal points and these variables for that part of our full sample of scalable respondents for whom we had a full menu of data (t-statistics are in parenthesis).

Notice the considerably greater success at predicting ideal points on the first dimension as compared to the second. This is not surprising, of course, given that it is preferences on the first dimension that correlate most strongly with policy preferences (see Figures 2-4). In general, though, the variables in this analysis all have signs that we might anticipate: older respondents and those preferring closer relations with Russia, along with respondents living in the East and identifying themselves as Russian are more "conservative", while those reporting improved economic status for their families are more "liberal".

Table 4: Ideal Point Estimation

 1st dimension2nd dimension
constant .76 (4.0)-.005 (.03)
entrepreneurship-.13 (7.5)-
Russian-Uk relations .17 (4.4)-
forces to drag Uk from crisis-.38 (15.9)-.21 (7.6)
age (in years) .003 (1.8)-
nat. Identity .17 (2.8)-
family economic status-.11 (3.1) .11 (2.4)
size of city of residence - .03 (2.7)
region (W,C,E) .06 (2.0) .09 (2.1)
adjusted R20,560,11

Turning now to our full sample, the difficulty here is that fewer that 1200 respondents answered all of the relevant questions. Nevertheless, examining this limited subsample does give us some insight into the value of our methodology and potential modifications of subsequent surveys. First, Figure 13 portrays the distribution of estimated ideal points for this subsample using the regression equations in Table 4, while Table 5 contrasts the predicted vote shares of the ten parties considered (assuming again that they account for 69% of the overall vote) against the actual shares they received in the election.

Table 5: Predicted versus Actual Vote Shares

 ActualPredicted
Agrarians3.67%10.79
NEP1.233.19
Socialist-Peasant8.643.30
Communist24.6829.36
PDP4.993.00
Rukh9.4011.55
Hromada4.684.47
Greens5.460.96
SDP0.320.57
USDP4.020.06

The predicted vote shares here are not, of course, as accurate as the ones we get from the direct question "for whom would you vote if the election were held today", but the deviations from actual shares are instructive. First, with the Socialist-Peasants falling between the Communists and Agrarians spatially, it is the later two parties that are predicted to get the "conservative" vote, which suggests that considerations (valence issues?) other that the one`s represented in our spatial structure account for some of the variance in voting patterns. Rukh"s vote share is also over-estimated while that of the Greens and USDP is considerably underestimated. The explanation for some of these deviations lies in Figure 13. Specifically, notice that the distribution of ideal points there, rather than resembling the triangle we see in Figure 1b, is now "cigar shaped". We are, to put matters simply, unable with the variables at hand, to sort out those "non-anti-reform" respondents (those with ideal points to the left of the vertical axis) who are attracted by nationalist appeals and those who are not. That is, too small a share of estimated ideal points lie in the upper-left quadrant of the coordinate space, which is where some of the centrist parties get their votes and where the USDP gets its support. Thus, since the Communists and Socialist-Peasant"s parties also get a small share of their support from this quadrant (owing to their rejection of pure nationalist appeals), the vote of Rukh, and the Communist"s are over-estimated and that of the USDP underestimated.

This is not to say, of course, that a different coding of variables or even a different combination of the variety of independent policy-oriented attitude measures might not yield more accurate predictions, but only that as things stand right now it is far easier to sort "liberals" from "conservatives" than it is to differentiate the various types of pro-reform voters.19 We could, of course, anticipate this problem from Table 4, which shows the relatively low variance explained by our variables with respect to variation of ideals on the second dimension, but we suspect that this problem concerns more than an inappropriate selection of independent variables or coding of those variables. Earlier we conjecture that the second dimension has not yet emerged in Ukrainian politics to "compete" equally with the more traditional left-right one. However, now that independence is an established fact, but the economy continues to languish, an increasing share of the electorate might reasonably be expected to be searching not for a route back to the past, but to reforms "that work" and that are somehow different from the policy menus offered by established political entities and politicians. And if these issues are only newly emergent, then "classical" measures of opinion will not necessarily be good predictors of their appeal.

The primary purpose of our methodology, of course, is not prediction - there are, as we note, better tools for that purpose. Rather, we want to assess the general coherence of the electorate`s attitudes and perceptions and the establishment of a baseline whereby we can detect the early emergence of new issues and there relationship to the general evaluative crityeria of the electorate. Thus, as a final check on the validity of our analysis, our survey presented respondents with a list of 23 prominent political figures, and asked, for each of them, whether they would consider votting for or against that person if given the opportunity to do so in their election district. Table 6, then, reports the average ideal point of those scalable respondents who indicated either that they would vote for or against the thirteen most recognizable names on this list after ordering the candidates from left-to-right on the basis of these scores. The results, as we can see, are consistent with initial expectations. Specifically, notice that the "supporters" of V. Chornovil - the leader of Rukh - anchor the lower right (nationalist) quadrant of the issue space (with V. Yavorivskyi of the Democratic Nationalist Party and S. Khmara of the Ukrainian Conservative Republican Party only slightly closer to the center), while those of P. Symonenko and O. Moroz, perhaps the most prominent figures in the Communist and Socialist camps, anchor the right half of the issue space. In fact, if we graph the mean ideal of respondents who indicate they would consider voting "for" a candidate, then with but a single exception, the support bases form an almost perfect upward sloping line - the principle deviation here being that of Kuchma"s prime minister Y. Marchuk, whose potential supporters lie in the upper-right quadrant that we show earlier is the primary source of support for the SDP and USDP.

We emphasize that the spatial positions reported in Table 6 are not those of the personalities themselves, but of respondents who indicated they would vote "for" or "against" them. Nevertheless, this table does suggests that the Ukrainian electorate (or at least that part of it that constitute our scalable sub-sample) hold consistent perceptions of candidates and parties. But that table also reveals the problem that "reformers" confront in the future, which is the same problem they confronted in the most recent parliamentary contest - namely, the crowded field of potential "centrist" competitors. Thinking ahead to the upcoming presidential contest and beyond, at least among those political figures included in our survey,

Table 6: Mean Ideal Points of Scalable Respondents Suggesting they would Vote "For" or "Against" Specific Political Figures

 1st / 2nd Dimension Fornumber
 1st / 2nd Dimension AgainstFor/Against/Total
V.Chornovil-.55 / -.49284 / 688 / 972
  .27 / .21 
V.Yavorivskyi-.47 / -.21301 / 347 / 648
  .31 / .01 
S.Khmara-.45 / -.33208 / 638 / 846
  .17 / .09 
V. Pynzenyk-.41 / -.19236 / 499 / 735
  .16 / .01 
V.Lanovyi-.28 / -.08305 / 315 / 620
  .19 / -.08 
Y.Marchuk-.26 / .22220 / 458 / 678
  .08 / -.19 
L.Kravchuk-.26 / .06294 / 670 / 964
  .14 / -.06 
V.Yushchenko-.25 / -.04315 / 273 / 588
  .26 / -.03 
I.Pliushch-.15 / -.17166 / 533 / 699
  .06 / -.03 
V.Grynev-.12 / .10146 / 327 / 473
 -.07 / -.21 
P.Lazarenko-.03 / -.10133 / 760 / 893
  .01 / -.03 
O.Moroz .35 / .13386 / 531 / 917
 -.26 / -.13 
P.Symonenko .82 / .17260 / 514 / 774
 -.43 / -.12 

ex-parliamentary speaker Moroz appears to have few competitors, while the nationalist camp has a small handful. Instead, most personalities are clustered around a pro-refom center, and the question remains whether a candidate such as Marchuk can capture this center while at the same time consolidating the upper-left quadrant of voters - those who support reform but reject any extreme nationalist appeal. However, as Table 6 shows, with the exception of ex-president Kravchuk, non-centrist personalities remain the most recognizible figures in Ukrainian politics, and, with the exception of National Bank chairman Viktor Yushchenko, every personality generates more opposition than support - doubtlessly reflecting a general sense of dissatisfaction with the progress of economic refom.

4. Conclusions

Much has beeen written about Ukraine`s somewhat tourtoured transition to democracy - a transition that has thus far taken place within the context of a fragmented or even non-existent party system, dangerous regional conflicts, changing electoral rules, and an imperfect national constitution (for an overview of these problems see Prizel 1997). Nevertheless, the picture painted of the Ukrainian electorate by our analysis here is a remarkably coherent one. To a considerable extent, that picture conforms to the one drawn by analysts using other methodological tools - the predominance of a single ideological dimension that summarizes preferences on a wide range of issues, including attitudes toward reform, Russia, and the sovereignty of Ukraine itself. However, our anaysis also uncovers a second dimension that serves largely to differentiate a more internationalist subpopulation of pro-refom voters who reject any strong or srident nationalist appeal. Moreover, rather than identify thee voters as centrists - as a natural compromise between national democrats such as supporters of Rukh and Communists and Socialists, our analysis suggests that these voters are best seen as the third vertex of a triangle of preferences in a two-dimensional evaluative criteria space.

In lieu, though, of any extensive discussion of Ukrainian politics and specualtions about the probable implications of our findings for future elections, we note here simply that, despite its short history as a democracy (and given at least one peaceful electoral replacement of its chief executive, and two successive and peaceful parliamentary contests, Ukraine does now qualify as a democracy, albiet a fragile one), the methodological tools and theoretical perspectives we apply here, which were developed to study more established democracies and more "experienced" electorates, seem suprisingly aplicable to Ukraine. That most, or even all, of our findings comport with accepted opinion about Ukrainian politics is itself a hopeful sign - a sign that electoral competition there is evolving in a direction that is not much different than what we find elsewhere. Regionalism and the bifurcation of attitudes between pro-and anti-reform forces remains problematical for th development of a national party sysem, but these features of Ukraine are not necessarily exceptional - indeed, we suspect that if the same methods were applied to the United States in its early years as a republic, we would, if anything, have found an even more sgarply divided electorate. The fact remains that both the eastern and western parts of the Ukrainin electorate perceive things in similar ways and evaluate the alternatives that confront them using equivalent criteria. Preferences differ, but there remains a vast middle ground that can be nurtured in search for a national comproimise if not consensus.

References

Arel, Dominique and Andrew Wilson. 1994. "The Ukranian Parlaimentary Elections," RFE/RL Research Report. v.3, #26 (July 1, 1994), 6-17.

Bojcun, Marko. 1995. "The Ukranian Parlaimentary Elections in March-April 1994," Europe-Asia Studies, 47.2: 229-49.

Cahoon, Lawrence, Melvin J. Hinich, and Peter C. Ordeshook. 1978. "A Statistical Multidimensional Scaling Method Based on the Spatial Theory of Voting," in P.C. Wang, Graphical Representation of Multidimensional Data, N.Y.: Academic Press.

Cahoon, Lawrence S. and Melvin J. Hinich. 1976. "A Method for Locating Tagets Using Range Only," IEEE Transactions on Information Theory. 22.2: 217-25.

Dow, Jay. 1998. "A Spatial Aalysis of the 1989 Chilean Presidential Election," Electoral Studies, forthcoming.

Downs, Anthony. 1957. An Economic Theory of Democracy. N.Y.: Harper an Row

Enelow, James and Melvin J. Hinich. 1984. The Spatial Theory of Voting, N.Y.:Cambridge University Press.

Hinich, Melvin J. and Michael C. Munger. 1994. Ideology and the Theory of Political Choice, Ann Arbor: University of Michigan Press.

Hoadley, John F. 1986. The Origins of American Political Parties. Lexington: Univ. Press of Kentuckey.

Khmelko, Valeri. 1996. "The Political Orientations of Different Regions nd Ethno-Linguistic Groups in Ukraine Since Independence, mimeo, KIIS, Kiev-Mohyla Academy.

Kravchuk, Robert S. and Victor Chudowsky. 1996. "The Political Geography of Ukraine`s 1994 Presidential and Parliamentary Elections," paper presented at the 1996 Annual Meeting of the N.E. Slavic Association, Worcester, Mass (April 19-20, 1996).

Kuzio, Taras. 1997. Ukraine Under Kuchma. N.Y.: St. Martin"s Press

Lin, Tse-Min, Yun-Han Chu, and Melvin J. Hinich. 1996. "Conflict Displacement and Regime Tansition in aiwan: Spatial Analysis," World Politics, 48.4: 43-81.

Marples, David R. 1994. "Ukraine After the Presidential Election," RFE/RL Research Report, v. 3, #31 (August 12, 1994), 7-10.

Myagkov, Misha and Peter C. Ordeshook. 1997. "The Spaial Character of Russia`s New Democracy," forthcoming, Public Choice

Myagkov, Mikhail, Peter C. Ordeshook, and Alexander Sobyanin. 1997. "The Russian Electorate, 1991-1996.," Post-Soviet Affairs, 13.2: 30-62.

Ordeshook, Peter C. 1997. "The Spatial Analysis of Elections and Committees," in Dennis C. Mueller, ed., Perspectives on Public Choice. Cambridge: Cambridge University Press.

Poole, Keith and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting. N.Y.: Oxford University Press

Prizel, Ilya. 1997. "Ukraine Between Proto-democracy and "soft" Authoritarianism," in Karen Dawisha and Bruce Parrott, eds. Democratic Changes and Authoritarian Reactions. Cambridge: Cambridge University Pess.

Remington, Thomas F., Steven S. Smith, D. Roderick Keiwiet, and Moshe Haspel, "Transitional Institutions and Parliamentary Alignments in Russia, 1990-1993," in T. F. Remington, ed,. Parliaments in Transition, Boulder (Colorado): Westview Press.

Tomenko, Mykola. 1995. "The Political System and Political Parties," in Volodymyr Polokhalo, ed., The Political Analysis of Postcommunism: Understanding Postcommunist Ukraine. College Station (Texas): Texas A+M University Press.

Wilson, Andrew. 1997. Ukrainian Nationalism in the 1990's. Cambridge: Cambridge Univ. Press.


Endnotes

1This research was supported by a grant to The California Institute of Technology by the National Council for Eurasian and East European Research.

2The primary difference between the two election systems is that Ukraine`s threshold for representation is set at 4% as against Russia`s 5%. Thus, eight parties with a combined total of 65.8% of the vote secured seats under PR in March, whereas if the 5% threshold had been used, only 5 parties with 53.1% of the vote would have won seats.

3It is, of course, heroic to assume that A is identical for all voters. But if there are reasons for supposing this is not true, any estimation of parameters should proceed by analyzing the separate subpopulations across which A is assumed to vary. Put differently, the absence of sufficient degrees of freedom preclude the possibility of independent estimates of A for each and every voter (respondent) represented in our data and, thus, we must rely on exogenous information to examine this possibility.

4There is, in general, some sensitivity of estimated parameters to the actual selection of the candidate whose scores are subtracted from the rest. Most commonly this sensitivity manifests itself in the rotation of the axes, and there is in fact an inherent ambiguity in that rotation. Throughout this essay the party subtracted from the rest is the Communist Party, and the most common effect of changing the identity of party "o" is the relative location of the Social Democratic Party (SDP) and the Socialist Democratic Party Block (USDP). Looking ahead to Figure 1a, both the SDP and USDP, rather than being centered above such parties as the Greens, NEP, PDP and Hromada, may, after an appropriate rotational adjustment, be shifted somewhat to the right.

5For the most detailed description of this statistical methodology see the methodological appendix in Enelow and Hinich 1984, but see also Cahoon and Hinich 1976 and Cahoon, Hinich and Ordeshook 1978.

6Admittedly, the decision to eliminate respondents who failed to give at least three parties distinctive grades is somewhat arbitrary, since there is some information in the responses of respondents who distinguish at least one party from the others or who distinguish one set of parties from the rest. Therefore, as a partial check on the sensitivity of our analysis to our procedures, the appendix to this essay offers the distribution of estimated ideals (Figure A1) and the recovered party positions (Figure A2) when we eliminate only those respondents who failed to distinguish any party from the rest. In this instance our sample of respondents increases from 1149 to 1557, and comparing Figures A1 and 1b and Figures A2 and 1a, we see that there is little change in our results. The primary difference can be seen when comparing Figures A2 and 1a. Specifically, the recovered candidate positions, most notably those of Rukh and the Communists are less dispersed in Figure A2 than in Figure 1a. This, however, is to be expected. If a respondent distinguished only one party from the rest and if that party is, as it tends to be, either the Communists or Rukh, all other parties will be "pushed" together, and this is precisely what we observe in the data. Throughout the remainder of this essay, however, we rely on the analysis of our smaller refined sample of 1149 respondents, if only because we ave somewhat greater confidence that these respondents have made a slightly greater effort at distinguishing one party from another, rather than merely identify their preferred party and treat all the others as equivalent.

7Our estimates here subtract the scores of the Communist party so as to eliminate the quadratic term in the expression for utility (see the previous section). The choice of party is somewhat arbitrary, although our results - most notably, the rotation of the recovered issue space - is sensitive to the party we choose. In general, the relative locations of the parties themselves is not sensitive to this choice.

8See, for example, Myagkov, Ordeshook and Sobyanin (1997) for a spatial analysis of early Russian elections, and Remington, et al. (1994) for a spatial analysis of voting in the early Russian Congress of People"s Deputies.

9There is little need to complicate these figures with any indication of which response corresponds to "for" and which "against" since all responses have precisely the anticipated orientation. For instance, voters to the left are more opposed to the unification of Ukraine with Russia or Belarus (Q7) than are voters to the right; respondents saying they would vote "for" speedy market reforms (Q8) have mean preferences to the left on the horizontal dimension whereas those saying they would vote "against" have mean preferences to the right; and so on.

10The allowable responses here were: (1) fully disagree; (2) disagree more than agree; (3) difficult to say; (4) agree more than disagree; and (5) fully agree.

11That both of these variables correlate with preferences on the second dimension is unsurprising given that approximately two-thirds of Ukraine`s rural population is Ukranian-speaking.

12There is some evidence that the second dimension plays a slightly greater role in the west than the east. The percent variance explained by the first eigenvalue of the recovery is 85% in the east and 75% in the west, whereas the variance explained by the second dimension is 5.3% in the east and 8.9% in the west. These differences, though, are too slight for any definitive conclusions.

13The coding for this figure is as follows: (1) Crimea; (2) Kiev city; (3) Kiev oblast; (4) Vinnytsya; (5) Volyn; (6) Dnepropetrovsk; (7) Donetsk; (8) Zhytomyr; (9) Transcarpathia; (10) Zaporizhzya; (11) Irano-Frankivsk; (12) Kirovohrad; (13) Luhansk; (14) Lviv; (15) Mykolaiv; (16) Odessa; (17) Poltava; (18) Rivne; (19) Sumy; (20) Ternopil; (21) Kharkiv; (22) Kherson; (23) Khmelnytsy; (24) Cherkasy; (25) Cherniytsi; (26) Chernihiv.

14Another potential explanation for Khmelnytsy`s "deviation" from expectations here is that this oblast holds a heavy concentration of military bases and, thereby, a somewhat higher proportion of Russian speakers or ethnic Russians than we might otherwise expect from its simple geographic position.

15Another possibility, of course, is that there are other parties, not included in our study, who siphoned off votes from SDP and USDP. And although we cannot exclude this explanation, this hypothesis leaves unanswered the question of why such "siphoning" did not occur with respect to the other parties in our study. So again we are left with the hypothesis of ineffective campaigning.

16These numbers are calculated thus: Of the full sample, 14.0% indicated an intention to vote for the Communist Party list. But only 71.9% indicated any party preference, leaving 28.1% with no stated preference. The number in this column for Communist, then, is 14.0/.719 = 19.5%.

17Since there are 172 scalable respondents who failed to state a party preference and 977 who did, and since we assume that of the 172 estimated votes, 68.7% will vote for one of the ten parties for which we have data, this cell is calculated as follows:

final % = 100[977(column B) +.687(172(column C))]/1149

18For an example of an "issue" that does not correspond to a valence dimension, respondents were asked to rank the parties in terms of their ability to bring order to the country. But of the scalable respondents, 291 ranked the Communists "very low" while 167 ranked them "very high", thereby indicating radically different perceptions of this party on this issue.

19Examining the set of all possible independent policy-preference variables in any simple straightforward way is made difficult, even impossible, by the considerable multicolinearity among these variables. Thus, our specific selection of independent variables is somewhat ad hoc.


Социология и политика

Рубрики: 





Рейтинги партий - [25.10.2008]



Подрубрика: Реакция на текущие события. - [17.08.2007]



Результаты эксперимента с техникой secret ballot в изучении электорального поведения украинских граждан - [26.05.2004]



Проблемы "рейтингологии" - [26.05.2004]



Sociological Survey as a Tool for Control of the Results of Elections and Referendums - [31.10.2003]



A Spatial Analysis of Ukraine`s 1998 Parliamentary Elections - [30.10.2003]



Загадка рейтингов - [11.10.2003]



Что могут и чего не могут социологи - [30.05.2002]















Copyright © 2002-2012 Киевский центр политических исследований и конфликтологии
Copyright © 2002-2012 Центр эффективной политики

При использовании материалов сайта ссылка на источник обязательна.






bigmir)net TOP 100