Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Three's a crowd in two-and-a-half-party systems : how third parties have undermined their own policy… Karnazes, Alexander John Peter 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2014_spring_karnazes_alexander.pdf [ 1022.45kB ]
Metadata
JSON: 24-1.0072157.json
JSON-LD: 24-1.0072157-ld.json
RDF/XML (Pretty): 24-1.0072157-rdf.xml
RDF/JSON: 24-1.0072157-rdf.json
Turtle: 24-1.0072157-turtle.txt
N-Triples: 24-1.0072157-rdf-ntriples.txt
Original Record: 24-1.0072157-source.json
Full Text
24-1.0072157-fulltext.txt
Citation
24-1.0072157.ris

Full Text

  THREE?S A CROWD IN TWO-AND-A-HALF-PARTY SYSTEMS: How Third Parties Have Undermined Their Own Policy Objectives in Five Post-War Democracies   by  ALEXANDER JOHN PETER KARNAZES  B.A., University of Florida, 2009     A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF  THE REQUIREMENTS FOR THE DEGREE OF   MASTER OF ARTS   in   THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES  (Political Science)        THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)     February 2014   ? Alexander John Peter Karnazes, 2014   ii Abstract  This study examines the manners in which third parties? electoral results and shifts in policy have affected major parties? policy positioning. I respond to the work of Adams and Merrill (2006) and of Nagel and Wlezien (2010) by analyzing two-and-a-half-party systems that contain a centrist or a non-centrist third party. The cases include parties elected by a variety of voting systems and with various political traditions. Ultimately, I find that, over the past half-century, third parties in Austria, Canada, Germany, Ireland, and Luxembourg have regularly undermined the policy objectives most commonly associated with the these parties.  A modified version of Nagel and Wlezien?s occupied-centre effect, which I call the occupied-position effect, has been present in the five examined national party systems. This finding, however, is only applicable with respect to shifts in policies that have principally been associated with third parties, what I call ?key policies?, as opposed shifts in general left-right positions. The evidence presented in this study shows that the major parties in two-and-a-half-party systems have consistently responded to third-party electoral gains by becoming less supportive of third parties? key policies. Three such policy areas are examined: welfare spending, market liberalization, and ethnonationalism.  I also show that there are effects from third parties changing their own policy positions, independent of how well they do at the polls. Exacerbating the dilemma that the analyzed third parties have faced, a key-policy version of Adams and Merrill?s reverse-shift effect appears to have been present in the examined party systems. This   iii means that the major parties have followed shifts in third parties? policy positions by shifting their positions in the opposite directions. Thus, third parties have undermined their own policy objectives when they have expressed (and shifted to) strong key-policy positions during election campaigns. Though third parties do have a strategic option pertaining to this effect ? expressing insincere, moderated policy preferences ? the long-term applicability of this tactic appears limited, especially in conjunction with the problems third parties have faced regarding the occupied-position effect.    iv Preface  This thesis is the original, unpublished work of Alexander J. P. Karnazes. Copies of this paper can be obtained via cIRcle ? University of British Columbia's digital repository for materials created by the UBC community and its partners ? at https://circle.ubc.ca/.  Many of the ideas presented here were first explored in ?Party Competition and Responsiveness? (2012), an unpublished, original paper written by Alexander Karnazes and Nigel Kinney. The co-authored 2012 paper was written to complete part of the requirements for POLI 837: Parties and Parliaments, a graduate course taught by Dr. Steve Weldon of Simon Fraser University.  To write this thesis, I have relied extensively on StatCoprs? Stata Statistical Software: Release 13 as well as version 2012b of the Comparative Manifesto?s Project?s Full Manifesto Dataset.     v Table of Contents Abstract .......................................................................................................................  ii Preface .......................................................................................................................  iv Table of Contents ........................................................................................................  v List of Tables ..............................................................................................................  vi List of Figures ............................................................................................................  vii Dedication ................................................................................................................. viii I. Introduction ............................................................................................................... 1 II. Literature Review .................................................................................................... 6 III. Data ...................................................................................................................... 11 IV. Hypotheses .......................................................................................................... 14  V. Variables and Model Specification ........................................................................ 21 VI. Results and Analyses .......................................................................................... 27 VII. Discussion and Conclusion ................................................................................. 43 References ................................................................................................................ 48 Appendices ............................................................................................................... 51     vi List of Tables Table 1     Left-Right Policy Results .......................................................................... 28 Table 2     Key Policy Results .................................................................................... 33 Table 3     Results Less Kim?Fording Data ............................................................... 39     vii List of Figures Figure 1     Irish Parties on Welfare Dimension ......................................................... 19 Figure 2     Irish Parties on Left-Right Dimension ...................................................... 19     viii Dedication  To Drs. Petia Kostadinova, Amie Kreppel, Derek Reiners, and Robert Uttaro ? thank you for the years of friendship and guidance. Also, I am tremendously grateful for the academic instruction and support provided by my thesis supervisor, Dr. Fred Cutler.  1 I. Introduction  Political party systems are generally discussed in the context of a spectrum that has one-party and two-party systems at one end, multi-party systems at the other end, and numerous hybrid systems between the two poles. This paper addresses one particular hybrid pattern that is present across multiple continents and amongst multiple electoral systems: two-and-a-half-party systems in which there are two major parties and one especially significant minor party. Though the two largest parties usually make the headlines, such as Christian democratic and social democratic parties in Western Europe, recent developments highlight the significance of third (i.e., half) parties. Examples include Canada?s New Democratic Party, which, in the 2011 federal general election, after many decades with support around 20%, won 30.6% of the general vote and 33.4% of the parliamentary seats. The election marked the first time in the NDP?s history that it became the Official Opposition, surpassing the Liberal party as the leading party not in government. Similarly, in 2010, Liberal Democrats of the United Kingdom won 23% of the general vote, the party?s largest share of votes since 1983, and 8.8% of the parliamentary seats, the second largest share of seats in its history. This election resulted in the Liberal Democrats joining the Conservative party to form the first coalition government in British electoral history. And, in 2009, the Free Democratic Party of Germany won 14.6% of the general vote and 15.0% of the parliament?s seats, which are all-time highs for the ?kingmaker? party. But since then, the FDP has been in a ?vicious downward circle? (The Local 2012), which culminated with the party failing to win 5% of the vote in the 2013 election and thus failing to obtain   2 seats in the Bundestag for the first time in the party?s history. In this paper, I analyze the effects of this type of party?s fortunes and policy positions on the national political system in which it operates. Principally, this study examines the manners in which third parties? electoral results and shifts in policy have affected major parties? policy positioning. I respond to the work of Adams and Merrill (2006) and of Nagel and Wlezien (2010) by examining (two-and-a-half-party) systems that contain a centrist or a non-centrist third party. The cases include parties elected by a variety of voting systems and with various political traditions. My primary objective is to answer the following: Do third parties unwittingly undermine their primary policy objectives when they achieve electoral gains? If so, do third parties have any strategic means to counter such an effect? I address these questions via new empirical analyses that are built on the framework of Adams and Somer-Topcu?s (2009) Party Dynamics Model. Ultimately, I find that, over the past half-century, third parties in Austria, Canada, Germany, Ireland, and Luxembourg have regularly undermined the policy objectives most commonly associated with the these parties.  A modified version of Nagel and Wlezien?s occupied-centre effect, which I call the occupied-position effect, has been present in the five examined national party systems. This finding, however, is only applicable with respect to shifts in policies that have principally been associated with third parties, what I call ?key policies?, as opposed shifts in general left-right positions. The evidence presented in this study shows that the major parties in two-and-a-half-party systems have regularly responded to third-party   3 electoral gains by becoming less supportive of third parties? key policies.  Additionally, a near-policy effect has been present in conjunction with this election-related effect. That is to say, in each national party system, the occupied-position effect has been more pronounced for the one of the two major parties that has traditionally been located nearer to the third party along the key policy dimension. Three key policies areas are examined in this paper: welfare spending, market liberalization, and ethnonationalism.  I also show that there are effects from third parties changing their own policy positions, independent of how well they do at the polls. Exacerbating the dilemma that the analyzed third parties have faced, a key-policy version of Adams and Merrill?s reverse-shift effect appears to have been present in the examined party systems. This means that the major parties have followed shifts in third parties? policy positions by shifting their positions in the opposite direction. Thus, third parties have undermined their own policy objectives when they have expressed strong key-policy positions during election campaigns (i.e., when they have shifted further to the periphery of the key policy dimension). Only by moderating their key positions (i.e, shifting toward the centre of the key policy dimension) have third parties been able to engender desirable policy responses from major parties, but such responses have been fractional relative to the moderating movements of third parties. Though third parties do have this strategic option at their disposal, the long-term applicability of this tactic appears limited, especially in conjunction with the problems third parties have faced regarding the occupied-position   4 effect. If a third party were to consistently moderate its key policy positions, the party would likely lose credibility among voters, ownership of their key issue, and the participation and support of its most ardent activists.  Before moving on, I clarify the term ?two-and-a-half-party system,? which I use in a manner that is in line with Blondel's (1968) and Ware?s (1996) definitions.1 That is to say, for the purposes of this analysis, two-and-a-half-party systems include political landscapes that have two major parties and one smaller third party. It is important to avoid confusion with the idea of ?2.5 ENP?, a term based on Laakso and Taagepera?s (1979) concept of effective number of parties. The 2.5 ENP designation can include systems in which the 0.5 is the product of two or more minor parties, a concept which is explored in Siaroff?s (2003) discussion of Spanish and Portuguese political parties.2 Party systems such as those present in Spain and Portugal are not the focus of this paper; therefore, this study moves forward by referring to the ?half? party in two-and half party systems as a third party (a somewhat less demeaning term). This means that there is only one third-party in each of the examined systems. In                                                 1 Blondel describes his idea of a two-and-a-half-party system:  Three-party systems also have another, and converse, characteristic. They all have two major parties and a much smaller third party.?all the three-party systems are in the second group; all systems of the third group have more than three significant parties. It can thus be stated that, while theoretically possible, a genuine threeparty system is not a likely type of system among Western democracies: countries of the second group should therefore be more strictly labelled as 'two-and-a-half-party systems'. (1968) Ware describes his idea of a two-and-a-half-party system: Often no one party wins 50% of legislative seats. Two larger parties usually win at least 80% of the seats, while one of other party wins sufficient seats to hold the balance of power? As we have noted already, these are systems containing three parties, two of which are much larger than the other one. (1996, 162?63) 2 Siaroff also does not make a distinction between a two-and-a-half-party system and a 2.5-party system. Thus the reader should note that the ?two-and-a-half-party system? term has different meanings in this study and in Siaroff?s work.   5 the context of this analysis, it is important not to confuse a third party with a new, niche, or protest party. Authors such as Maurice Pinard (1975) may address multiple, so called, third parties in a given country, but parties such as the Social Credit and Bloc populaire canadien parties in Canada are beyond the scope of this paper. The third parties addressed in this analysis are the New Democrats (NDP) of Canada, the Labour Party of Ireland, the Freedom Party (FP?) of Austria, the Free Democrats (FDP) of Germany, and the Democratic Party (DP) of Luxembourg.3  Though there are case studies that address the parties listed above and a series of works that seek to classify typologies of party systems ? Duverger (1959), Blondel (1968), Sartori (1974), plus the more recent contributions of Ware (1996), Mair (1997, 2006), Siaroff (2003), Wolinetz (2006), etc. ? my research has uncovered no papers that systematically examine the impact of third parties in a longitudinal and cross-national manner. Alan Siaroff does address different outcomes related to so-called ?hinge? and ?wing? parties, which are third parties and others that comprise the 0.5 component of 2.5-party systems, in his 2003 paper; however, this discussion is far less developed than the one in which he systematically defines 2.5-party systems. The author only discusses probabilities of hinge and wing parties joining government; he does not directly examine the manner in which third parties affect public policy. As a starting point to do just that, I look to the time-series analyses presented and the questions raised by Adams and Merrill (2006) and Nagel and Wlezien (2010) in their case studies of the UK?s two-and-a-half-party system.                                                 3 Following the CMP?s lead, Canada?s Co-operative Commonwealth Federation (CCF) election results and policy positions are included in NDP data.   6 II. Literature Review  Adams and Merrill (2006), as well as Nagel and Wlezien (2010), find that the Liberal Democrats (and their predecessors), the UK?s centrist third party, has undermined its policy objectives by contesting elections. The two sets of authors propose a total of three mechanisms by which this has happened, two of which are pertinent to established two-and-a-half-party systems. Adams and Merrill present the policy-divergence and the reverse-shift effects, which are based on the assumption that parties are policy motivated, and Nagel and Wlezien present the ?occupied-centre? effect, which is based on the assumption that parties are vote motivated.4 The policy divergence effect theoretically occurs when a minor centrist third party is introduced to a two-party system, encouraging the two major parties to take up policy positions that are closer to the poles of the left-right scale. The reason for this is that major parties see half the return on shifts toward the centrist party; thus, the peripheral major parties diverge twice as much as would be found in a two party system because they can afford to take positions closer to their ideal (peripheral) policies without concern about losing electoral support. While the policy divergence effect may illustrate the consequences for a centrist party entering a two-party system, it does little to explain the (available) strategies and actions of actors in established two-and-a-half-party systems and it has no explanatory value for two-and-a-half-party systems that are absent a centrist third party and do not use first-past-the-post (FPTP) voting. However,                                                 4 The latter two authors also present a vacated centre effect:  This hypothesis helps explain the [centrist] third party?s vote. If the major parties move to the right and left for reasons of their own ? ideological convictions of leaders, internal caucus politics, influence of activists and funders ? the Liberals fill the vacated centre and flourish by attracting unhappy moderates with nowhere else to go. (2010)   7 the other two ? reverse shift and occupied centre ? mechanisms are pertinent to ongoing dynamics in two-and-a-half-party systems and it is worth exploring their generalizability. Adams and Merrill?s second finding is that the two major parties will remain dispersed to the same degree and shift their policy positions in the opposite direction of policy shifts made by a centrist third party ? this is the reverse-shift effect. The thinking behind this idea is that the major party on the right will move its policy positions further to the right, closer to its most desired policy position, in response to a leftward movement from the third party. The right-most major party can afford to do this without concern of losing electoral support because (a) it is able to pick up centre-right voters left behind by the third party while (b) its main competitor, the left-most party, loses centre-left votes as a result of the third party?s shift. Meanwhile, the major party on the left responds to the centrist party by also moving to the right. It does so to avoid losing significant centre-left voters, which would propel the other major party to electoral victory. And, assuming parties are policy motivated as Adams and Merrill do, the left-most party?s rightward movement does not take it past the centre party?s new position because the major party prefers to be close to its ideal policy preference. The opposite shifts occur if the third party moves to the right (i.e., the right-most party moves to the left and the left-most party moves further to the left.).  Thus, as a result of the policy divergence effect, the introduction of a centrist third party engenders a dispersal response from the major parties and the reverse-shift effect means that the third party has no strategic (a.k.a. ?instrumental? in Adams and Merrill?s   8 language) mechanism, a least not in terms of policy position, to combat the divergence effect. However, if the third party has, for example, centre-right policy preferences, it has a strategic incentive to put forth policies that are to the left of centre. While this may be an effective short-term strategy, such insincerity would likely strain voters? and activists? trust and support if employed too often. In another study, rather than looking at major parties? reactions to policy shifts of a centrist third party, Nagel and Wlezien (2010) examine major parties? reactions to prior third-party election results. Unlike the reverse-shift effect, this occupied-centre effect operates under the assumption that parties are primarily motivated to win votes. The authors propose that the major parties will further diverge from the centre at election t if the centrist third party increases its share of votes at election t ? 1. The third party?s success indicates that the major parties have been deserted by moderate activists, and these parties, pessimistic about attracting moderate voters away from the centrist party, respond by moving to please their remaining core supporters who are closer to the periphery. This vacating of the centre from within the major parties and by the major parties creates a cyclical effect that can last for years at a time, limited by the third partiy?s ability to maintain electoral success (Nagle and Wlezien 2010, 290). The major parties are able diverge from the centre because the third party has historically not been a threat to win a majority or plurality of seats. The authors follow the works of Budge (1994); Adams, Clark, Ezrow, and Glasgow (2004); and Somer-Topcu (2009) in testing the effects of past electoral results, but they differ by examining the impact of one party?s past electoral results on another party?s present policy position. Other works only   9 address the relationship between parties? own past election results on their own policy positions.  Nagel and Wlezien?s regression results show that a significant effect is related to past third-party electoral results, but they find that, in the UK, only the Conservatives have consistently moved closer to the periphery following increases in Liberal vote share. The authors suggest that the asymmetry in the major parties? responses to the third party?s electoral strength is a consequence of it having been easier for activists to transfer their support between the Conservatives and the Liberal Democrats rather than between Labour and the Liberal Democrats.5 In addition to this primary reason for asymmetry, Nagel and Wlezien suppose that the presence of extremist minor parties on the right, but not the left, has been producing a centrifugal pull only on the Conservatives. The authors conclude their paper by arguing that Liberal electoral strength is a proxy indicator for the number of constituencies the Liberal Democrats contest, which they infer to be an indicator of the third party?s ability to attract centrist activists ? some of whom might otherwise be involved with the Conservatives helping moderate the Tories? policies Adams and Merrill and Nagel and Wlezien present convincing evidence that the Liberal Democrats? presence and success in British elections motivate Labour and the Conservatives to take up policies anathema to Liberal voters? desires, so a natural next step is to look beyond the UK to see if other third parties have similarly undermined their policy ambitions. Regarding such an endeavor, Adams and Merrill actually conclude                                                 5 Nagle and Wlezien?s logic is based on the pronounced salience of social class that has long been present in British politics and the commonly held notion that the Liberal Democrats and Conservatives are both middle-class parties.   10 their work proposing that their two-and-a-half-party model can be extended to scenarios beyond British politics:  ? simple simulations suggest that the effect of a minor, peripheral party contesting the election is typically to shift the optimal strategies of both major parties in the direction opposite to that of the small party?a ?reverse shift? effect analogous to the one we identified in this paper, and one that is detrimental to the minor party?s immediate policy objectives. (2006, 414)  Nagel and Wlezien, on the other hand, correctly identify that, by definition, their occupied-centre effect has a far narrower application: ?Generalizing beyond Britain is difficult because only a few major [FPTP] systems still exist, and none of them exhibits the British pattern of a significant third party in the political centre? (2010, 302). Nonetheless, the latter authors? idea of exploring the relationship between third parties? electoral results and major parties? policy positions is one that warrants a cross-national investigation. I propose that such research involves the analysis of specific policies in addition to the conventional left-right dimension. The manner in which I investigate an occupied-position effect is explained in this paper?s Variables and Model Specification section. The current paper tests a total of four hypotheses: the aforementioned reverse-shift and occupied-position effects, the key-policy effect, and the near-party effect. To test theses hypotheses, I, like Merrill and Adams, as well as Nagel and Wlezien, utilize data produced by the Comparative Manifesto Project (CMP).     11 III. Data As part of the CMP, experts across 55 countries have coded quasi-sentences from the policy programmes of parties competing in elections. These codes have been grouped into 56 categories (e.g., welfare, environmental protection, defense, etc.) among 7 major policy areas, with each variable representing a percentage of emphasis within a given manifesto. Scholars have also developed indexes comprised of two or more categories among the 56 categories that CMP has coded. Budge and Laver?s (1992) RiLe ? Right-Left ? index, for example, is an indicator of parties? positions on a scale that ranges from -100 to 100, with lower scores denoting more leftist positions.6 Merrill and Adams, as well as Nagel and Wlezien, rely on RiLe measures for their examinations. CMP data is especially useful because it provides researchers with the ability to spatially orient and longitudinally compare party positions. It is, however, important to note that CMP data only addresses rhetorical policy positions, not the actual policy actions parties undertake while in office or opposition. Nonetheless, CMP scores over time correspond remarkably well to informed observers? judgments, and political scientists often use them to track the ideological movements of parties around the world (Nagle and Wlezien 2010). The long history of heated intra-party debates over the content of party programmes testifies to their and the CMP?s importance (Adams and Somer-Topcu 2009). In addition to the aforementioned RiLe scores, I use data from                                                 6 RiLe is comprised of 26 categories: the sum of 13 pro-right categories and 13 pro-left categories. RiLe = (per104 + per201 + per203 + per305 + per401+ per402 + per407 + per414 + per505 + per601+ per603 + per605) ? (per103 + per105 + per106 + per107 + per403 + per404 + per406 + per412 + per413 + per504 + per506 + per701 + per202)   12 three other indices in this analysis. The CMP-developed Welfare and Markeco indices represent support for the state provision for welfare and support for a market economy, respectively,7 and Gadjanova?s (2013) Ethnonat index represents support for a national way of life and traditional morality.8 Theses three indices, each ranging from 0 to 100, are especially pertinent to this study?s most original contribution: investigation of the key-policy effect.  Along with elections results data, the RiLe, Welfare, Markeco, and Ethnonat indices are used to examine the policies of political parties across a total of 87 elections. These elections have taken place from 1945 to 2011 in Canada, Ireland, Austria, Germany, and Luxembourg. I follow precedent in selecting the national party systems in theses countries as scholars have regularly described them in terms similar to this study?s definition of a two-and-a-half-party-system. For example, the above list of countries is a four-out-of-five match of Ware?s (1996) list of two-and-a-half-party systems. As our two lists demonstrate, Germany, Canada, Ireland, and Austria are commonly present in party system literature as examples of national two-and-a-half systems (see Siaroff 2003, Blondel 1968, Ware 1996, etc.). I follow Blondel?s lead in classifying Luxembourg as a two-and-a-half-party system and including it in this analysis. Over the years, Luxembourg national politics may have exhibited a higher ENP than the other party systems included in this study, but following all but one election since 1945, the same three parties have held the three largest shares of seats. Traditionally, the Christian Social People's Party (CSV) and the                                                 7 Welfare = per503 + per504. Markeco = per401 + per414. 8 Ethnonat = per601 + per603 + per608   13 Luxembourg Socialist Workers' Party (LSAP) have been the two major parties while the Democratic Party (DP) has been the smaller third party. While the DP has almost always won the third most seats in each election, there has been no regular fourth place challenger.9 Since Blondel first described Luxembourg as two-and-a-half-party system in 1968, four different parties have occupied the fourth-place position with no party able to occupy the fourth position on more than three occasions. The present study examines electoral and policy dynamics related to three-party systems, and it makes every sense to include Luxembourg.10 Having outlined the data that is used in this study, the following section presents four hypotheses that are tested on the data.                                                   9 The DP finished fourth in seat shares in 1948 and second in seat shares in 1979.  10 I also considered adding Greece?s national part system, with New Democracy (ND) and Panhellenic Socialist Movement (PASOK) as the two major parties and Communist Party of Greece (KKE) as the third party, but data limitations prevented its inclusion.    14 IV. Hypotheses The first hypothesis proposes that Adams and Merrill?s reverse-shift effect is present in two-and-a-half-party systems outside of Britain. The goal is to determine whether or not third parties have strategic options concerning their stated policy positions.  Hypothesis 1: Reverse-Shift Effect  H1: All things being equal, a previous change in a third party?s policy position engenders a change in major parties? present policy positions. The major parties shift their policies in the opposite direction of the third party?s shift.   Applying the above hypothesis to this study?s dataset, I expect to find, for example, that the Liberal and Conservative parties in Canada will move to the right from election t -1 to election t if the New Democratic Party (Canada?s third party) moves to the left from election t ? 2 to election t ? 1.11  Though their analysis is limited to the UK?s political system, Nagle and Wlezien also directly test for the presence of a reverse-shift effect via a lagged relationship.12 But, with an N of just 16, they find that correlations between third-party (Liberal Democrats) electoral results and major-party (Conservative and Labour) reverse shifts                                                 11 Adams and Somer-Topcu aptly describe the rationale for examining parties ?policy relationships, such as the reverse-shift, effect via lagged variables: Theoretically, [the] focus on parties? responses to their opponents? lagged policy shifts is justified by the time-consuming process of writing the party policy manifesto, which typically takes place over a two?three year period during which party-affiliated research departments and committees draft sections of this manuscript, which are then circulated for revisions and approval upward to party elites and downward to activists. (2009, 832) 12 I have also conducted cursory tests to see if there are significant relationships between this paper?s dependent variables and non-lagged versions the explanatory variables. I have found no convincing results that negated the effects presented in this paper or that strongly support the inclusion of non-lagged explanatory variables. Although, with more data in the future, it might prove fruitful to examine non-lagged versions of Adams and Somer-Topcu?s (2009) Party Dynamics Hypothesis and Somer-Topcu (2009)?s Time Moderation Hypothesis.    15 in Britain are only statistically significant at p < 0.17. By expanding the N via five national party systems rather than one and by testing actual shifts in policy rather than absolute positions, I have more statistical power to evaluate the reverse-shift effect (or lack thereof). The second hypothesis extends the idea of Nagel and Wlezien?s occupied-centre-effect to two-and-a-half-party systems beyond Britain. The goal is to determine whether or not third parties? electoral success undermine their policy objectives. Hypothesis 2: Occupied-Position Effect H2: All things being equal, major parties will move away from a third-party?s policy position if the third party experienced electoral gains in the previous election.   Applying the above hypothesis to this study?s dataset, I expect to find, for example, that Fine Gael and Fianna F?il in Ireland will move to the right from election t ? 1 to election t if Labour achieves electoral gains from election t ? 2 to election t ? 1. I test this second hypothesis by looking at vote shares, seat shares, and a seat-share dummy variable to see which might have the greatest influence. I examine vote shares, because of the three, it is the closest to representing the public?s preferences. I look at seat share because the underrepresentation attributed to electoral systems, e.g., Canada?s first-past-the-post system, may establish thresholds concerning vote share that must be met elicit major party responses. Finally, as third parties by definition win few seats, there may be little variance in positive results and thus a dummy variable for seat gains may be more appropriate than a continuous variable for vote or seat shares.  The third hypothesis proposes that focusing on specific policies will produce more robust results than only examining the general left-right scale.    16 Hypothesis 3: Key-Policy Effect H3: The effects described in the three other hypotheses are driven primarily by effects in specific policy areas as opposed to the left-right dimension.  The reasoning behind this is that third parties outside of Britain are often associated with a more specific issue focus than the major parties.13 I suspect that, in general, third parties are able to maintain their existence by managing ownership of a particular issue. As Russell and Fieldhouse note in their discussion about the Liberal Democrats lack of party ownership in the UK, ?In particular a crucial task facing the modern third party is to maintain an identity that is distinctive from that of the two major parties?  (2005, 114).14 Though I have found no cross-national studies of third parties that examine specific policies, key policies and issue ownership are ideas that are well discussed in niche party literature (see Meguid 2005; 2010; and Adams, et al. 2004) and in case studies of individual third parties. In Austria, for example, the FP??s electoral stagnation and internal conflict during the 1970s and early 80s is generally attributed to party leaders? attempt to shed the FP??s far-right image and establish the party as the country?s liberal/centrist party (Pelinka 2001; Meret 2009; Campbell 1995). Conversely, the FP??s electoral success                                                 13 One of the peculiar aspects of this study is the omission of the United Kingdom from the cross-national multivariate regression models. The reason for this is that the Liberal Democrats (the UK?s third party) and their predecessors have not been associated with a particular policy area (key policy) and thus it is not appropriate to include the party system in key-policy analyses, and including the UK in only the RiLe models makes for an unnecessarily uneven comparison between RiLe and key models. I have found no CMP category or index for which the Lib Dems have consistently held the highest or lowest scores in the United Kingdom. The UK?s party system may have inspired this study, but it is apparently the odd man out with regard to key-policy analyses and cross-national theories concerning the reverse-shift and occupied-centre effects in two-and-a-half-party systems. 14 While making this statement, Russell and Fieldhouse reference other works on the Lib Dems that highlight the importance of identity: Dunleavy (1993), Crewe and King (1995), and Russell and Fieldhouse (2000).   17 since 1983 is largely attributed to the party?s re-embrace of xenophobic and ethnonationalist policies. For the past three decades, the FP?, as it was prior to its liberal experiment, has undoubtedly been Austria?s principle ethnonationalist party.   Adams and Merrill, as well as Nagel and Wlezien, produced their findings based on RiLe data, but as the RiLe index consists numerous policy categories, a party?s policy shift in terms of this index may simply indicate a major shift in one or more of the policies included in RiLe. Five of the seven total policy categories within the Welfare, Markeco, and Ethnotnat indices are included in RiLe. Therefore, if major parties? responses to a third party?s shift in policy and/or electoral gains are based on welfare, market economy, or ethnonationalist policies, these responses would be present, in a diluted manner, in the major parties? RiLe shifts. To test the third hypothesis, I use (a) the Welfare index to measure the key policy positions related to the NDP in Canada and Labour in Ireland, (b) the Markeco index to measure the key policy positions related to the DP in Luxembourg and FDP in Germany, and (c) the Ethnonat index to measure the key policy positions related to the FP? in pre-1970 and post-1983 Austria.15 The fourth hypothesis addresses Nagel and Wlezien?s asymmetrical finding regarding their occupied-centre effect and tests for asymmetry concerning the reverse-shift effect. It does so by drawing from Adams and Somer-Topcu?s ?ideological families hypothesis,? which posits, ?Parties are more responsive to policy shifts by members of their ideological family than to the policy shifts of other parties in the system? (2009,                                                 15 FP??s liberal experiment years have been excluded because Ethnonat does not measure the party?s key variable during that time.   18 828).16 The term ?ideological family? simply refers to whether a party belongs to one of three groups: leftists and left leaning parties, centrist parties, or right-leaning and rightist parties.  I apply this idea by breaking down ideological families into two groups: major parties nearest to the third parties and major parties farthest from third parties on a policy scale.17 Hypothesis 4: Near-Party Effect H4: All things being equal, the major party that has historically been closest to the third party?s policy position will respond to the third-party?s policy shifts and electoral gains to a greater degree than the other major party.   By limiting the categories of Adams and Somer-Topcu?s ideological families hypothesis (from three categories to two), the essential principle can be applied to party systems that are the subject of this analysis. Regarding welfare policy in Ireland, for example, Fine Gael has historically been the major party closest to the Labour Party. In fact, Fine Gael has on occasion emphasized pro-welfare policies to greater extent than Labour in their manifestos. Thus, in applying this hypothesis to our dataset, I expect to find that Labour Party?s policy shifts and electoral gains have had a greater influence on Fine Gael?s welfare policy positions compared to its influence on the Fianna F?il?s welfare policy positions.                                                 16 Nagel and Wlezien find that, in Britain, the third party?s electoral results induced policy responses from the Conservatives and not Labour because (a) ?party activists would usually find it easier to transfer their support between Conservatives and Liberals than between Labour and Liberals? and (b) ?the frequent appearance in the last thirty years of extremist minor parties on the right but not the left, [produced] a centrifugal pull on the Conservatives but not on Labour.? Notice that point (a) and the ideological family hypothesis are based on a similar logic.  17 In their discussion of the reverse-shift effect, Adams and Merill?s propose that the major parties respond to third-party shifts in policy with symmetrical (not asymmetrical) policy shifts; however, I am not aware of any published empirical tests of this effect.    19 The fourth hypothesis is obviously only applicable if one or both of the first two hypotheses hold true, and I suspect that the near-party effect has been more pronounced when the key-party effect, the third hypothesis, has also been present. To this point, Figures 1 and 2 illustrate that there is a relatively consistent ordering to Irish parties? positions on welfare policy ? Labour has generally been the most supportive, Fianna F?il the least supportive, and Fine Gael has generally been in the middle ? but there is no such order to the parties? positions on a general left-right scale.18 Data indicates that parties outside of Ireland are also more ordered on key policy dimensions than they are on left-right dimensions, and that a clear ordering of parties is a necessary component of a robust near-party effect. Figure 1.     Figure 2.    The catch-all nature of the left-right dimension ? as measure by RiLe, an index composed of 26 CMP categories ? makes it more likely for party ordering to vary as the salience of social and economic policies fluctuate over the years. Take a classic                                                 18 Labour has traditionally been the most left-leaning party, but Fianna F?il and Fine Gael?s RiLe positions have not been consistent. The major parties? regularly crisscrossed positions until the 1990s at which time they started, and have since been, mirroring the median voter positions. 010203040CMP Welfare Score1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010YearFianna F?il Fine GaelLabourIrish Parties on Welfare Dimension-45-15154575CMP Rile Score1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010YearFianna F?il Fine GaelLabour Median VoterIrish Parties on Left-Right Dimension  20 libertarian party as an example. If economic issues are the primary concern in a given election, then the party is likely to emphasize its free market principles in its manifesto, which would lead to a higher RiLe scores (more rightward positions) relative to other elections. If, however, social issues are the primary concern in a given election, the libertarian party?s is likely to emphasize its liberal social principles in its manifesto, which would lead to lower RiLe scores (more leftward) relative to other elections. In a two-and-a-half-party system, this type of fluctuation could lead to the libertarian party being the rightmost party for one election and the centre party for another. On a market economy dimension ? as measure by Markeco, an index comprised of only two CMP categories ? the libertarian party would almost always, if not always, have the highest Markeco score. As Markeco represents a single policy issue, it is also hard to imagine the other two parties in the system (e.g., a Christian democratic party and a social democratic party) would regularly alter their ordering on this dimension. Details about testing the fourth hypothesis and testing the three other hypotheses are presented in the discussion pertaining to this analysis? primary framework.      21 V. Variables and Model Specification I utilize Adams and Somer-Topcu?s Party Dynamics Model (2009) as a framework to test this study?s four hypotheses. While the authors do not directly examine relationships specific to two-and-a-half-party systems, they do provide insight into testing lagged policy-to-policy and election-to-policy relationships on a cross-national level. Their findings encompass 193 parties in twenty-five post-war democracies, and their primary conclusion is that ?? parties tended to shift their policy positions in the same direction that their opponents had shifted their policies at the previous election; furthermore, parties were particularly responsive to policy shifts by other members of their ?ideological families??? (2009, 825). Though I utilize a framework based on Adams and Somer-Topcu?s Party Dynamics Model, I suspect negative relationships are present among major party policy shifts with respect to third-party electoral gains and policy shifts.  The Party Dynamics Model?s dependent variable is simply the change in left-right policy (RiLe) for a given party from election t ? 1 to election t. Unlike the dependent variable addressed in Nagel and Wlezien?s tests of the occupied-centre effect, Adams and Somer-Topcu?s dependent variable applies to a wide variety of parties, such as niche, third, major, etc. This paper?s dependent variables lie somewhere in between Nagel and Wlezien?s and Adams and Somer-Topcu?s dependent variables. They are limited to major parties, like Nagel and Wlezien?s variables, but, like Adams and Somer-Topcu?s variables, they address parties from across numerous national party systems. Attempting to predict the values of Adams and Somer-Topcu?s dependent   22 variable, the Party Dynamics Model employs three independent variables: two controls and one primary explanatory variable. Modified versions of these three independent variables serve as controls for the present examination. The dependent and control variables that comprise this study?s model specification, with brief justifications, are as follows:  Dependent  ? major Party?s shift in policy (t) o measured as change in a major party?s policy positions between election t ? 1 and election t  o derived from Adams and Somer-Topcu?s ?party shift (t)? as well as Nagel and Wlezien?s ?LRLab,t? and ?LRCon,t?  Control ? all other parties? shifts in policy (t ? 1) o measured as mean change in policy positions of all parties between election t ? 2 and election t ? 1, not including the major party in question o derived from Adams and Somer-Topcu?s ?average shift ? other parties (t ? 1)?, the primary explanatory variable in their models19 o accounts for the possibility that a major party is responding to policy positions of all other parties, not just the third party?s policies and electoral gains  ? major Party?s shift in policy (t ? 1) o measured as change in a major party?s policy positions between election t ? 2 and election t ? 1 o derived from Adams and Somer-Topcu?s ?party shift (t ? 1)? o accounts for Budge (1994) and by Adams? (2009) claims that party elites have electoral incentives to shift their party?s policies in the opposite direction from their shifts in previous election, and it eliminates autocorrelation that can otherwise be present in the data  ? public opinion shift in L-R (t) o measured as change in the median voter?s left-right position between election t ? 1 and election t. o derived from Adams and Somer-Topcu?s ?public opinion shift (t)?                                                 19 Adams and Somer-Topcu state the following about their primary explanatory variable:   To date, however, we are unaware of cross-national studies that address the question of whether real world political parties in fact systematically adjust their policy positions in response to rival parties? policy strategies. That is the issue we address here, and our party dynamics and ideological parties? findings suggests that the answer to this question is ?Yes? ? an answer that supports the central assumption used in spatial models of party competition. (2009, 842)   23 o accounts for the possibility that major parties are simply responding to desires of voters rather than third-party policies and electoral gains  Additionally, the next two variables allow for the testing of the reverse-shift and occupied-position effects: Explanatory  ? third party?s shift in policy (t ? 1) o measured as change in the third party?s policy positions between election t ? 2 and election t ? 1 o indicator of the reverse-shift effect  ? third party?s shift in elect. results (t ? 1) o measured as change in the third party?s electoral results between election t ? 2 and election t ? 1 o indicator of the occupied-policy effect  All of the operationalized models analyzed in this paper are built upon a model specification comprised of the above elements: major Party?s shift in policy (t) = + b1 [major Party?s shift in policy (t ? 1)]        + b2 [public opinion shift in L-R (t)]20        + b3 [all other parties? shifts in policy (t ? 1)]         + b4 [third party?s shift in policy (t ? 1)]        + b5 [third party?s shift in elect. results (t ? 1)]        + b6  Based on the work of Adams and Somer-Topcu, I expect negative signs to be present for b1 coefficients and expect positive signs to be present for b2 and b3 coefficients. Regarding the main explanatory variables, I expect negative signs to be present for b4 and b5 coefficients.  It is important to note that the variables used in this examination are attempts at operationalization of the variables that form the model specification above. For example, this study analyzes four dependent variables based on major Party?s shift in policy (t): Left shift in L-R (t), Right shift in L-R (t), Near shift in key (t), and Far shift in key (t). The                                                 20 This variable is only included if applicable (i.e., only pertains to L-R policy analyses).   24 term ?Left? refers to the major party that has generally been to the left of the third party based on historical RiLe scores while ?Right? refers to the major party to has generally been to the right of the third party.21 The term ?Near? refers to the major party closest to third party along a given policy dimension while ?Far? refers to the major party furthest away from third party along the a given policy dimension. The term ?key? refers to the primary policy of concern (e.g., market economy for Germany?s FDP, as measured by Markeco index) while ?L-R? simply refers to the left-right (RiLe) policy scale.22 Similar patterns apply to the model specification?s control variables, excluding public opinion shift in L-R (t), which is a single variable unto itself because of data limitations concerning specific policies. Following Adams and Somer-Topcu?s lead, I measure public opinion using median voter data developed by Kim and Fording (1998, 2003).23 Kim and Fording ?conceive of elections as large-scale opinion polls," and they calculate voter preferences on a left-right scale by linking electoral outcomes with CMP-derived party positions (2003, 96). Their data can be useful given the temporal limitations of survey data but the assumption of independence could be violated if the dependent variable is RiLe-related. Nonetheless, to my knowledge, there is no Kim-and-Fording-type measure for specific policies such as welfare expansion or free market, and there is no consistent national                                                 21 In some countries, such as Germany and Austria, the third party has historically had a major party to the left and right of it while in other countries, such as Canada, Ireland, and Luxembourg, the third party has only had major parties to the left or only to the right of it (see Appendix A). 22 In each country, the two major parties have historically had lower key policy scores than the third party (see Appendix).  23 For CMP and election data beyond 2003, I developed public opinion/median voter data via the method introduced by Kim and Fording.   25 survey that measures pre-election public preferences on specific issues prior to the mid-1980s.24 Just as Nagle and Wlezien and Adams and Merrill forgo the use of a public opinion measure in their models, this analysis proceeds by examining the third hypothesis without a control for public opinion. The third and fourth hypotheses ? the key-policy and near-party effects, respectively ? can be tested by comparing model results stemming from the above framework. Of the four aforementioned dependent variables that can be constructed from major Party?s shift in policy (t), two are applicable to left-right policy analyses and two to key policy analyses. I assess the third hypothesis by comparing the (lack of) robustness of key policy results to the (lack of) robustness of left-right results. If, for example, Third shift in key (t ? 1) and Third shift in elect. results (t ? 1) have substantially higher explained variances with respect to Near shift in key (t) and Far shift in key (t) than they do regarding Left shift in L-R (t) and Right shift in L-R (t), one could infer that the key-policy effect does indeed pertain to the reverse-shift and occupied-position effects. Furthermore, I use a similar form of analysis to test the fourth hypothesis. I compare the size of near-party effects to far-party effects. If, for example, Third shift in key (t ? 1) and Third shift in elect. results (t ? 1) have higher explained variances with respect Near shift in key (t) than they do regarding Far shift in L-R (t), this would indicate that the near-party effect does indeed pertain to the reverse-shift and occupied-position effects.  The above variables, model specification, and operationalizations illustrate the                                                 24 See Soroka and Wlezien (2004) for a good discussion on preferences for welfare spending in Canada.    26 manners in which CMP and electoral data can be used to test this paper?s four hypotheses. The strength of the reverse-shift effect (H1) is determined via results pertaining to Third shift in key (t ? 1), and the strength of the occupied-position effect (H2) is determined via Third shift in L-R (t ? 1). The key-policy effect (H3) is assessed via the comparison of key-policy and RiLe-policy models. Finally, the near-party hypothesis (H4) is assessed via the comparisons of near-party and far-party models. The next step is to actually test the hypotheses and analyze the results.     27 VI. Results and Analysis Table 1 displays results pertaining to shifts in policy on a left-right scale. Models 1?3 apply to policy shifts of major parties that have historically been to the left of third parties (e.g., policy shifts by SP? of Austria, SPD of Germany, LSAP of Luxembourg, and CSV of Luxembourg) and models 4?6 apply to policy shifts of major parties that have historically been the right of third parties (e.g., policy shifts by ?VP of Austria, Liberals and Conservatives of Canada, CDU/CSU of Germany, and Fianna F?il and Fine Gael of Ireland). The dependent variable in models 1?3 is Left shift in L-R (t) and the dependent variable in models 4?6 is Right shift in L-R (t).   28 Table 1: Left-Right Policy Results  Model 1 Model 2 Model 3 Model 4 Model 5 Model 6  Left shift in L-R (t) Left shift in  L-R (t) Left shift in  L-R (t) Right shift in L-R (t) Right shift in L-R (t) Right shift in L-R (t) Left shift in L-R (t ? 1) -0.229** ? (0.111) -0.246** ? (0.111) -0.234** ?(0.112)    Right shift in L-R (t ? 1)    -0.243*** ?(0.0775) -0.250*** ?(0.0784) -0.250*** ?(0.0787) public opinion shift in L-R (t) 0.347*** (0.0764) 0.349*** ? (0.0774) 0.350*** ? (0.0768) 0.470*** ? (0.0594) 0.475*** ?(0.0598) 0.473*** ? (0.0593) avg. all others? shifts in L-R (t ? 1) 0.216 ? (0.209) 0.209 ?(0.218) 0.255 ?(0.221) -0.234 ?(0.198) -0.248 ? (0.198) -0.239 ? (0.197) Third shift in L-R (t ? 1) -0.118 ?(0.153) -0.131 ?(0.160) -0.174 ? (0.164) 0.316** ? (0.139) 0.331** ? (0.140) 0.325** ? (0.139) Third shift in vote% (t ? 1) -0.313 ?(0.326) ? ? -0.126 ? (0.362) ? ? Third shift in seat% (t ? 1) ? -0.0769 ?(0.319) ? ? -0.238 ? (0.347) ? Third shift in seat%  ?    dummy (t ? 1) ? ? 1.757 ? (3.641) ? ? -1.968 ? (3.071) Constant -0.507 ?(1.640) -0.492 ?(1.654) -1.416 ? (2.531) -0.263 ? (1.446) -0.272 ? (1.443) 0.794 ?(2.200) R2 0.400 0.389 0.392 0.475 0.477 0.477 Adjusted R2 0.338 0.326 0.328 0.448 0.450 0.449 Observations 54 54 54 100 100 100   Standard errors in parentheses   ? p<.15, * p<.10, ** p<.05, *** p<.01   29 Two separate collections of models, 1?3 and 4?6, are presented because the occupied-position hypothesis states that third-party electoral gains are associated with different patterns of major-party policy shifts based on parties? spatial positioning: in a given system, if the third party experiences electoral gains at the previous election, the major party to the left of the third party will shift further to the left (negative RiLe scores) and the major party to the right will shift further to the right (positive RiLe scores) at the present election. Appendix Table A presents major and third parties? average RiLe positions from 1945?2011 in Austria, Canada, Germany, Ireland, and Luxembourg. The reverse-shift hypothesis, on the other hand, states that third-party policy shifts affect all major parties in the same fashion: in a given system, there is a negative relationship between (a) the direction of the third party?s left-right shift at the previous election and (b) the direction of major parties? left-right shifts in the present election. Furthermore, I present three models with Left shift in L-R (t) as the dependent variable and three models with Right shift in L-R (t) as the dependent variables ? instead of one model for each dependent variable ? to determine the measure of third-party electoral gains that best captures the proposed occupied-position effect. Models 1 and 4 test the second hypothesis using Third shift in vote% (t ? 1), models 2 and 5 use Third shift in seat% (t ? 1), and model 3 and 6 use Third shift in seat%  ? dummy (t ? 1).  Having explained the structure of Table 1, I now address the variables and coefficients within it. First, the expected signs and levels of significance are present for two of the three control variables derived from Adams and Somer-Topcu?s Party Dynamics Model. Coefficients indicate that public opinion shift in L-R (t) in models 1?6   30 has an expected positive effect on the dependent variables while Left shift in L-R (t ? 1) in models 1?3 and Far shift in L-R (t ? 1) in models 4?6 have expected negative effects.  Regarding public opinion shift in L-R (t), the numbers illustrate that the two major parties in each two-and-a-half-party system have been shifting their policies, not surprisingly, in the same directions that public opinion, as defined on a general left-right scale, has been shifting. In order to appropriately test this study?s hypotheses, I have account for this positive relationship. With respect to Left shift in L-R (t ? 1) and Right shift in L-R (t ? 1), the b1 variables in their respective models, the negative effects illustrate that major parties have put forth policies such that a zigzag, or regression-to-the-mean, pattern of policy alternation has been present (i.e., a shift to the right succeeds a prior shift to the left and vice versa). This is a commonly known equilibration process that must also be taken into account.25 While the above control variables are statistically significant at p < 0.01, avg all others? shifts in L-R (t ? 1), the third and last control variable in models 1?6, is not significant in any of the Table 1 models. This variable is the primary explanatory component of Adams and Somer-Topcu?s Party Dynamics Model, and given the authors? findings, the results here are surprising. It thus appears that the Party Dynamics Model in its original left-right application ? i.e., parties tend to shift their left-right policy positions in the same direction that their opponents had shifted their policies at the previous election ? does not apply to each of the two major parties in Austria,                                                 25 Budge (1994) argues that the policy alternation phenomena are a product of party leaders needing to appease both the radical and moderate factions of their parties. Adams (2001), on the other hand, argues that political parties zigzag their policy positions over time in attempts to please win voters, who have both policy and non-policy priorities (e.g., party identification).   31 Canada, Ireland, Germany, and Luxembourg.26 Regarding the main explanatory variables, Table 1?s results indicate that the reverse-shift and occupied-position hypotheses do not hold true in the five national party systems addressed in this study, at least with respect to left-right policy scales. This also means that a near-party hypothesis is not supported with respect to major parties? (lack of) left-right responses to third parties? electoral gains and shifts in left-right positions. The three variables that test for the occupied-position effect are not statistically significant in any of the Table 1 models. Moreover, Third shift in L-R (t ? 1), which tests for the reverse-shift effect, is not statistically significant in models 1?3 though it does have the expected negative sign in each of these models.  Interestingly, Third shift in L-R (t ? 1) does contribute to the explanatory power of models 4?6, but it does so in a manner opposite to that which the reverse-shift hypothesis posits. With positive coefficients in model 4?6 significant at p < 0.05, it appears that major parties to the right of third parties have followed left-right policy shifts made by third parties by shifting their own policies in the same direction as the third party had shifted. Thus for every 10-point shift to the left (right) made by a third party from election t ? 2 to election t?1, the rightward major party/ies have shifted, on average, 3.24 points to left (right) from election t ? 1 to election t.27 This preliminary result challenges the negative-relationship findings reported by Adams and Merrill and the                                                 26 Remember, I measure avg. all others? shifts in L-R (t ? 1), as Adams and Somer-Topcu measure ?average shift ? other parties (t ? 1),? by calculating an average based on each and every opposition party?s shift in policy at the previous election. 27 Remember, RiLe operates on a -100?100 scale, and negative scores indicate leftward positions while positive scores indicate rightward positions.     32 applicability of the reverse-shift hypothesis outside of Britain.  Given the results presented in Table 1, it appears that the key-policy effect (H3) needs to be rather potent if the reverse-shift and occupied-position effects are to offer any explanatory value in line with this paper?s first two hypotheses. Table 2 displays results pertaining to key-policy shifts and it does so in a format similar to that which is presented in Table 1. There are two sets of three models that comprise a total of six models. And each model contains three control variables; one variable that tests for the reverse-shift effect, Third shift in key (t ? 1); and one of three different variables that test for the occupied-position effect, either Third shift in vote% (t ? 1), Third shift in seat% (t ? 1), or Third shift in seat% ? dummy (t ? 1).   33 Table 2: Key Policy Results  Model 7 Model 8 Model 9 Model 10 Model 11 Model 12  Near shift in key (t) Near shift in key (t) Near shift in key (t) Far shift in key (t) Far shift in key (t) Far shift in key (t) Near shift in key (t ? 1) -0.364*** ? (0.107) -0.357*** ?(0.105) -0.320*** ? (0.104)    Far shift in key (t ? 1)    -0.343*** ? (0.103) -0.347*** ? (0.103) -0.372*** ? (0.103) public opinion shift in L-R (t) -0.00633 ? (0.0302) -0.00136 ? (0.0300) 0.00776 ?(0.0293) -0.0197 ?(0.0158) -0.0184 ?(0.0158) -0.0153 ? (0.0157) avg. all others? shifts in key (t ? 1) 0.655* ? (0.332) 0.588* ?(0.330) 0.484? ? (0.323) 0.368*** ?(0.131) 0.366*** ? (0.131) 0.389*** ?(0.129) Third shift in key (t ? 1) -0.410** ? (0.172) -0.379** ? (0.171) -0.365** ?(0.164) -0.309*** ? (0.0836) -0.306*** ?(0.0834) -0.320*** ?(0.0821) Third shift in vote% (t ? 1) -0.180 ? (0.162) ? ? -0.0441 ? (0.0840) ? ? Third shift in seat% (t ? 1) ? -0.273* ? (0.151) ? ? -0.0724 ?(0.0787) ? Third shift in seat%  ?   dummy (t ? 1) ? ? -4.153*** ? (1.454) ? ? -1.312* ?(0.764) Constant 0.0500 ? (0.720) 0.0323 ? (0.710) 2.246** ?(1.028) -0.282 ? (0.379) -0.286 (0.377) 0.413 ?(0.549) R2 0.258 0.278 0.323 0.298 0.303 0.323 Adjusted R2 0.205 0.227 0.275 0.248 0.254 0.275 Observations 77 77 77 77 77 77     Standard errors in parentheses   ? p<.15, * p<.10, ** p<.05, *** p<.01   34 The potency of the key-policy effect can be understood by comparing this table?s results with those in Table 1. The robustness of the near-party effect (H4) can be ascertained by respectively comparing the results in models 7?9 to the results in models 10?12. In each party system, I have classified one major party as near and one as far. The near party, in each system, has traditionally been closer the to third party?s key policy positions than the far party has been. These distinctions are based on the key policy positions of the three parties that comprise the five two-and-a-half-party systems addressed here. Appendix Table B presents major and third parties? average key positions, as measured by CMP data, from 1945?2011 in Austria, Canada, Germany, Ireland, and Luxembourg.  Apart from public opinion shift in L-R (t), the control variables have the expected signs and are statistically significant at traditionally accepted levels. First, Near shift in key (t ? 1) in models 7?9 and Far shift in key (t ? 1) in models 10-12 have p-values that are less than .001 in all of the Table 2 models, just like their b1 counterparts in Table 1. This means that, from election to election, each major party generally zigzags its positions on key policies just as they have their left-right positions (i.e., a higher key policy score follows a lower score and vise versa). Second, the lack of significance for public opinion shift in L-R (t) in Table 2 is not surprising. The robustness of b2 estimates in Table 1 compared to those in Table 2 highlights the fact that left-right and key policy positions are distinctly different from one another. In other words, it seems that public opinion on a left-right scale cannot be used as a proxy for public opinion on various key policies. Lastly, unlike the numbers related to its RiLe counterpart in Table 1, the figures   35 pertaining to avg. all others? shifts in key (t ? 1) in Table 2 are as expected and in line with the idea behind Adams and Somer-Topcu?s findings. This variable has a positive sign and is significant in all of the Table 2 models. Thus, at each election, major parties have shifted their key-policy positions to follow the direction of the average change in all other parties? positions.28  Looking at the explanatory variables, the b4 and b5 estimates in Table 2 indicate that major parties are much more likely to have responded to third-party policy shifts and electoral gains in terms of third parties? primary policy concerns rather than in terms of general left-right scales. That is, the reverse-shift and occupied-position effects are robust in Table 2 while they are nonexistent in Table 1. The results pertaining to Third shift in key (t ? 1) suggest that, all things being equal, a prior 10-point increase in a third party?s emphasis of its key-policy preferences will, on average, lead to major parties decreasing their present level of support for that policy by 3?4 points. Thus, when a third party became more (less) emphatics about its key policy from election t ? 2 to election t ? 1 in a given party system, the major parties became less (more) supportive of it from election t ?1 to election t ? 2, but the major parties? reverse shift responses were 30?40% less in magnitude than the third party?s initial shift. Remember, each CMP index (Welfare, Ethnonat, and Markeco) used to measure key policy positions in this study can have scores ranging from 0?100, and each point of those scores represents a percentage of emphasis in a given party manifesto. This                                                 28 For example, all things being equal, the SPD in Germany is more supportive of welfare expansion from election t ?1 to election t if all of the other parties in Germany (including the FDP and the CDU/CSU), on average, became more supportive of welfare expansion from election t ?2 to election t ? 1.   36 means that a decrease in a key score indicates a decrease in supportive statements for, not necessary statements against, a third party?s primary policy. Conversely, an increase in a key score indicates an increase in supportive statements for a third party?s primary policy. Regarding major parties? responses to past third-party election results, Third increase in seat% ? dummy (t ? 1), as opposed to Third increase in vote% and Third increase in seat%, produces the most robust results. All six b5 estimates in Table 2 have negative signs, as the second hypothesis proposes; however the dummy variables are the only ones that are statistically significant, one with Near shift in key (t ? 1) as the dependent variable (see Model 9) and one with Far shift in key (t ? 1) as the dependent variable (see Model 12). The increase in absolute size and statistical significance from Third shift in vote% (t ? 1) to Third shift in seat% (t ? 1) and from the latter variable to Third increase in seat% ? dummy (t ? 1) indicates that past third-party electoral gains must reach a certain threshold before major parties will consistently respond. Thus, in each party system, all things being equal, the major parties? responses to prior third-party election results have been binary in character: decreased support of the key policy if the third party increased its share of seats or no change in support if the third party lost or maintained its share of seats.  As Third increase in seat% ? dummy (t ? 1) has demonstrated itself to be the best measure of the occupied-position effect, I move forward with the analysis by focusing on models 9 and 12 in Table 2 and models 3 and 6 in Table 1. Model 9 illustrates that if a third party had increase its share of seats from election t ? 2 to   37 election t ? 1, the nearest major party in terms of key policy position will, on average, have responded by being 4.2 points less supportive of the third party?s primary policy. And Model 12 illustrates that the farthest major party will, on average, will have responded by being only 1.3 points less supportive of the third party?s primary policy.  The appreciable difference between these two coefficients suggests that there is indeed a near-party effect with respect to the occupied-position effect.29 In Model 9, the Third increase in seat% ? dummy (t ? 1)?s coefficient is more than three times the size of the corresponding coefficient in Model 12 as the ratio of theses estimates is 3.17:1.30 There is also a difference between the size of the b4 estimates in in model 9 and 12, but Third shift in seat% (t ? 1) is only 0.045 larger in Model 9 than in Model 12. The ratio of these estimates is just 1.14:1. Thus, while the near-party effect may apply to both the reverse-shift and occupied-position effects, the size and consistency of the effect are much greater regarding the occupied positions effect. All of this means that that, in each party system, both major parties have responded to third-party electoral gains by being less supportive of the third party?s primary policy ? and the negative responses of the major party closest to the third party along the key policy dimension have been about three times greater than the negative responses of the other major party. Like the first and second hypotheses, the fourth hypothesis is only applicable to key policy analyses. The contrast of the robust reverse-shift and occupied-position effects displayed in models 9 and 12 of Table 2  (key policy models that each use a                                                 29 A significant difference is also present in between models 8 and 11 and between models 7 and 10. 30 In fact the ratios are even larger regarding Third shift in vote% (t ? 1), models 7 and 10, and Third shift in seat% (t ? 1), models 8 and 11.     38 dummy variable for third-party seat share increases) to the lack of such effects displayed in models 3 and 6 of Table 1 (left-right policy models that each use a dummy variable for third-party seat share increases), respectively, is evidential support for the existence of a key-policy effect (H3). However, the two RiLe-related models have higher R2 values. How can this be? I suspect that public opinion shift in L-R (t) is responsible for this peculiarity. In models 3 and 6 of Table 1, only the b1 and b2 estimates (measures of the policy alternation effect and left-right public opinion, respectively) are associated with expected relationships and are statistically significant. While the b1 estimates in the RiLe-related models are similar to their counterparts in the key-related models 9 and 12, the same cannot be said for the b2 estimates. The b2 estimates (which pertain to public opinion shift in L-R in all of the Table 1 and Table 2 models) are robust in models 3 and 6, but they are not significant in any of the key-policy models. Given that the left-right and key models have similar have b1 estimates,31 it seems unlikely that the policy alternation variables are responsible for higher R2 values in the RiLe-related models. Thus, it is reasonable to assume that the b2 variable, public opinion shift in L-R (t), is the reason for higher R2 values in models 3 and 6, especially in Model 3.32 Below, I present Table 3 to illustrate the explanatory power of the reverse-shift and occupied-position effects in models without the Kim-Fording-derived public opinion                                                 31 Left shift in L-R (t ? 1) for Model 3, Right shift in L-R (t ? 1) for Model 6, Near shift in L-R (t ? 1) for Model 9, and Far shift in L-R (t ? 1) for Model 12 32 The b4 variable, Third shift in L-R (t ? 1), is also statistically significant in Model 6 and contributes to a higher R2 than in Model 3, but its positive relationship is directly opposite of the negative relationship presented in the first hypothesis.   39 variable.33 Not surprisingly, the figures in models 15 and 16 are very similar to those in models 9 and 12 of Table 2. As public opinion shift in L-R (t) has a negligible impact within the key-related models, the variable?s absence from models 15 and 16 is of little consequence. Table 3: Results Less Kim?Fording Data  Model 13 Model 14 Model 15 Model 16  Left shift in L-R (t) Right shift in L-R (t) Near shift in key (t) Far shift in key (t) Left shift in  L-R (t ? 1) -0.323** ?(0.131)  ? ? Right shift in  L-R (t ? 1)  -0.276*** ? (0.101)   avg. all others? shifts in L-R (t ? 1) 0.0498 ?(0.256) -0.424* ? (0.252) ? ? Third shift in L-R (t ? 1) 0.0135 ? (0.188) 0.377** ? (0.179) ? ? Near shift in key (t ? 1)   -0.323*** ?(0.102)  Far shift in key (t ? 1)    -0.385*** ?(0.102) avg. all others? shifts in key (t ? 1) ? ? 0.496? ? (0.318) 0.398*** ? (0.129) Third shift in key (t ? 1) ? ? -0.367** ? (0.163) -0.327*** (0.0817) Third shift in seat%  ? dummy (t ? 1) 2.182 ? (4.313) 1.922 ? (3.906) -4.072*** ?(1.412) -1.453* ? (0.750) Constant -1.586 ?(2.998) -1.537 ?(2.809) 2.201** ?(1.008) 0.491 ? (0.543) R2 0.128 0.123 0.322 0.314 Adjusted R2 0.057 0.086 0.284 0.276 Observations 54 100 77 77   Standard errors in parentheses   ? p<.15, * p<.10, ** p<.05, *** p<.01                                                 33 Models 13, 14, 15, and 16 of Table 3 include the same variables as models 3, 6, 9, and 12, respectively ? but public opinion shift in L-R (t) is excluded from these Table 3 models.   40 Looking at the RiLe-related results, models 13 and 14 of Table 3, respectively, have very different results than models 3 and 6 of Table 1. The very low adjusted R2 values in the Table 3 models ? 0.057 in Model 13 and 0.086 in Model 14 ? make it clear that public opinion shift in L-R (t) is responsible for nearly all of explanatory power presented in Model 3 and Model 6, which have adjusted R2 values 0.328 and 0.449, respectively. Interestingly, Third shift in L-R% (t ?1 ), as it does in Model 6 of Table 1, continues to have a significant positive relationship with Right shift in L-R (t) in Table 3.  All in all, the comparing of RiLe-related models to key-related models leads to three major findings. First, there have been no reverse-shift or occupied-position effects regarding major parties? shifts in general left-right positions. We see this in all of the Table 1 models and in models 13 and 14 of Table 3. Second, when taking into account the third parties? shifts in general left-right positions ? i.e., the sum of a wide variety of policy shifts (RiLe includes 26 CMP categories) ? major parties to the right of third parties appear to have moved in the same direction as the third parties, not the opposite direction as they do with respect to third parties? key policies.34 Major parties on the left have been unaffected by third parties? shifts in general left-right positions. Third shift in L-R (t ? 1)?s nonexistent effect and positive significant effect in RiLe-related models 13 and 14, respectively, relative to Third shift in key (t ? 1)?s negative significant effects in key-related models 15 and 16 highlights the distinctly different dynamics that have been                                                 34 If true, this would mean that, in a two-and-a-half-party system, the rightward major party shifts in the same direction as a third party with respect to one or more of the non-key policies. Tests, however, have show that no significant effect is produced when Third shift in L-R (t ? 1) is added to key policy models such as those displayed in Table 3. Thus, the manner in which third parties? adjust their non-key policies does not affect the manner in which major parties adjust their stances on key policies.    41 present in left-right policy relationships and in key policy relationships. Third, either public opinion shift in L-R (t) is remarkably effective at explaining major parties? left-right shifts in policy, or a variable derived from Kim and Fording?s median voter data may not be truly independent of a RiLe-based dependent variable. I suspect that there is more truth to the latter point than the first. In discussing a regression relationship similar to those presented here, Adams, Clark, Ezrow, and Glasglow warn, ?? the inclusion of a Kim?Fording measure would involve using party ideologies as both the dependent variable and a component of the independent variable? (2004, 595).35 Taking the differences between models 13 and 14 and models 15 and 16 (Table 3) into account, there is strong evidence of a key-policy effect. By examining the inclusion and exclusion of public opinion shift in L-R (t), I clarify that the reverse-shift and occupied-position effects are not responsible for the high R2 values in Table 1. Whether or not the questionable Kim?Fording median voter data is included in models, it is clear that the first and second hypotheses are not supported with respect to left-right policy analyses. Note that as the reverse-shift and occupied-position effects are not applicable to left-right analyses, then, by definition, neither is the fourth hypothesis (the near-party effect). Conversely, all three of these hypotheses are robust with respect to key policy analyses.  The three tables presented here illustrate that acknowledging a key-policy effect                                                 35 Interestingly, though Adams warns about the use of Kim?Fording data with his coauthors in 2004, he and Somer-Topcu argue for its applicability in 2009: This measure is the only cross- national public opinion measure that is available over the entire post-war period for the twenty-five party systems in our study. McDonald and Budge report analyses suggesting that the Kim?Fording measure closely tracks alternative cross-national measures of public opinion, such as those based on the Eurobarometer surveys. (830)   42 (H3) is essential for understanding the dynamics of two-and-a-half-party systems. The major parties in Austria, Canada, Germany, Ireland and Luxembourg have indeed shifted their positions based on third parties? key policies, and they have done so according to the patterns predicted by the reverse-shift and occupied-position hypotheses.36 They have also done so in partial accordance with the near-party hypothesis as the occupied-position effect is significantly stronger with respect to the major party in each country that is closest third party?s key policy position. The major parties, however, have not shifted their overall left-right positions in these ways. Along with illuminating that the reverse-shift and occupied-position effects are not unique to the United Kingdom but exist in two-and-a-half-party systems outside of Britain, this paper also highlights the importance examining parties and policies beyond the left-right paradigm.37                                                   36 There is some evidence that suggests, in each system, the key policy reverse-shift and occupied-position effects at election t are mildly mitigated if the third party was in government following election t ? 1. See coefficients that pertain to the dummy variable Third in gov?t ? dummy (t ? 1) and its interactions in Appendix Table C. Readers should note, however, that increasing the number of explanatory variables creates a bloated parameter space, which means that the results in this table are likely less tenable than those in Table 3. 37 In saying that the occupied-position effect exists in the UK, I am referring to the occupied-centre effect presented by Nagle and Wlezien.    43 VII. Discussion and Conclusion  In writing this paper, I have endeavoured to (a) determine if third parties outside of the United Kingdom have regularly undermined their immediate policy aspirations as the Liberal Democrats have done in Great Britain and, if they have done so, to (b) illustrate the mechanisms by which this has occurred. That is to say, I have sought to present the (a) ?what? and the (b) ?how? regarding third-party dynamics in two-and-a-half-party systems. Admittedly, more can said regarding the ?why.? Though this analysis presents strong evidence in support of the reverse-shift, occupied-centre, and key-policy hypotheses, along with partial support of the near-party hypotheses, it does not fully explain why these effects have been present in post-war Austria, Canada, Germany, Ireland, and Luxembourg. In their 2006 paper, Adams and Merrill present spatial models that explain, as well as empirical evidence that attempts to verify, the presence of a reverse-shift effect (and policy-divergence effect) in the UK party system. The authors, however, admittedly state that they have no tractable mathematical explanation for their simulation results that illustrate the presence of the reverse-shift effect in situations in which a third party?s ideological position is more extreme than the two major parties (414). Nagel and Wlezien, in their 2010 paper, dispute the notion of a reverse-shift effect having been present in British politics arguing instead that an occupied-position effect (and vacated centre-effect) has been present.38 However, these authors state that it is difficult to generalize their findings beyond the United Kingdom as there are no first-past-the-post                                                 38 For consistency?s sake, I use to term ?occupied-position? effect. In their paper?s UK-only context, Nagel and Wlezien use the term ?occupied-centre effect?.     44 systems that exhibit the British pattern of a significant third party in the political centre (302). Now, I am arguing that both the reverse-shift effect and occupied-position effect (in conjunction with the key-policy and near-party effects) are present in systems outside of the UK. How is it that both effects are present in non-British systems when the occupied-position effect is supposed to be an alternative theory to the reverse-shift effect? Is there an overarching explanation that ties the two effects together? Or is there some other yet fully developed explanation for the presence of this paper?s four hypothesized effects that have been present in the five examined party systems? Future, primarily theoretical, research needs to answer these questions. With this study?s introduction of the key-policy effect, one idea worth exploring is the application of Bonnie Meguid?s (2005; 2010) niche-focused Position, Salience, and Ownership (PSO) theory to third-party dynamics in two-and-a-half-party systems. Evidence presented here shows that issue ownership is an important aspect of third-party politics and that the same third-party dynamics exist across a variety of different institutional structures. Thus, there are reasons to investigate whether the four effects presented in this paper are a consequence of, or at least related to, major parties implementing dismissive, adversarial and/or accommodative tactics with respect to third parties and key policies.  Though there is not a clear understanding as to why the reverse-shift, occupied-position, key-policy, and near-party effects have persistently been present in post-war Austria, Canada, Germany, Ireland, and Luxembourg, the fact that they have been is a significant discovery with important implications. That major parties have indeed shifted   45 their positions on third parties? key policies according to patterns predicted by the occupied-position hypothesis raises serious concerns about the democratic nature of third parties. The concept of wasted votes in systems that use FPTP voting is well understood, and now it is clear that, regardless of the voting system, voters in two-and-a-half-party systems need to be aware of counterproductive voting. For example, a German voter whose primary concern is market liberalization likely does herself a disservice if she votes and/or rallies support for the Free Democrats.39 Current and potential supporters of third parties must look beyond policy congruence and electability to understand the policy benefits, if any, of voting for a third party. Though proximity and directional voting models (first popularized by Downs in 1957 and Rabinowitz and Macdonald in 1989, respectively) seemingly fall short of explaining why voters concerned about key policies would elect third-party parliamentarians when doing so runs counter to their national policy aspirations, perhaps an instrumental voting model could offer a better explanation. In federal and non-centralized systems, such as Canada and Germany, national electoral success may provide regional benefits for third parties and their supporters despite provoking undesired responses from major parties. National electoral gains usually bring funding and exposure that can help parties such the NDP and FDP win local elections and govern locally (e.g., cities, municipalities, states, etc.) and implement local policies that possibly impact voters? daily lives more so than national policies. If voters have long time horizons and intuitively understand the impact of institutions on the outputs of                                                 39 This same logic applies to (a) Austrian voters, ethnonationational policy, and the FP?; (b) Canadian voters, welfare policy, and the NDP; Irish voters, welfare policy, and Labour; and (c) Luxembourg voters, market liberalization policy, and the DP.    46 policy as Kedar (2009) argues, they may rationalize that engendering undesired response from major parties by voting for third parties in national elections is an acceptable price to pay for potentially greater regional policy benefits. Even if there is some validity to this conjecture regarding instrumental voting, the presence of the occupied-position effect is still troublesome in terms of representative democracy and is still a significant hurdle with which third parties must contend.   Further exacerbating the dilemma that third parties face regarding their key policies is the presence of the reverse-shift effect. In terms of major-party responses, third parties are punished if they profess strong key-policy positions and seem to be rewarded if they moderate and water down their rhetoric. Were it not for the reverse-shift effect, a third party could respond to the obstacle of the occupied-position effect by being a protest voice that highlights the primary concerns of its supporters and makes these concerns more politically salient, which would theoretically draw the major parties closer to the third parties? key positions. However, the significance of the reverse-shift effect means that such a protest position is not an effective option and moderation is the only programmatic means that third parties have to engender desired policy responses from the powers that be. The crucial element for discovering the perverse effects that pertain to third parties outside the UK has been the examination of specific policies. If this analysis had only been conducted in terms of the left-right spectrum and the CMP?s aggregated measurement of it, few, if any, substantial findings would have discovered. Ultimately, I have borrowed ideas from left-right analyses of three-party dynamics in the United   47 Kingdom and modified and applied them to specific-policy analyses of five national two-and-a-half-party systems that contain identifiable third parties. In Austria, Canada, Germany, Ireland, and Luxembourg, third parties have regularly undermined their most identifiable policy goals by winning parliamentary seats and by expressing strong key-policy preferences. These findings raise serious concerns about the quality of representative democracy in three-party systems, especially for policy-oriented supporters of the smaller third parties.     48 References Adams, James. Party Competition and Responsible Party Government: A Theory of Spatial Competition Based Upon Insights from Behavioral Voting Research. Ann Arbor: University of Michigan Press, 2001.  Adams, James, Michael Clark, Lawrence Ezrow and Garrett Glasgow. ?Are Niche Parties Fundamentally Different from Mainstream Parties? The Causes and the Electoral Consequences of Western European Parties' Policy Shifts, 1976?1998.? American Journal of Political Science L, no. 3 (July 2006): 513?529.  Adams, James and Samuel Merrill, III. ?Why Small Centrist Third Parties Motivate Policy Divergence by Major Parties.? American Journal of Political Science C, no. 3 (August 2006): 403?417.  Adams, James and Zeynep Somer?Topcu. ?Policy Adjustment by Parties in Response to Rival Parties? Policy Shifts: Spatial Theory and the Dynamics of Party Competition in Twenty-Five Post?War Democracies.? British Journal of Political Science XXXIX, no. 4 (October 2009): 825?846.   Blondel, Jean. ?Party Systems and Patterns of Government in Western Democracies.? Canadian Journal of Political Science 1 (1968): 180?203.  Budge, Ian. ?A New Spatial Theory of Party Competition: Uncertainty, Ideology and Policy Equilibria Viewed Comparatively and Temporally.? British Journal of Political Science XXIV (1994): 443?467.  Campbell, David F. J. "J?rg Haider." In Political Leaders of Contemporary Western Europe: A Biographical Dictionary, edited by David Wilsford, 183?188. Westport, CT: Greenwood Press, 1995.  Downs, Anthony. An Economic Theory of Democracy. New York: Harper & Row. 1957  Duverger, Maurice. Political Parties: Their Organization and Activity in the Modern State. New York: John Wiley, 1959.  Gadjanova, Elena. ?Measuring Parties' Ethnic Appeals in Democracies. Party Politics (12 February 2013): 1?19, accessed July 2013, doi:10.1177/1354068812472586.  Kim, HeeMin and Richard C. Fording. ?Voter Ideology in Western Democracies: An Update.? European Journal of Political Research XLII (2003): 95?105.  Kedar, Orit. Voting for Policy, Not parties: How Voters Compensate for Power Sharing. Cambridge: Cambridge University Press, 2009.    49   Laakso, Markku and Rein Taagepera. ??Effective? Number of Parties: A Measure with Application to West Europe.? Comparative Political Studies XII, no. 3 (1979): 3?27.   Laver, Michael and Ian Budge. ?Measuring Policy Distances and Modelling Coalition Formation.? In Party Policy and Government Coalitions, edited by Michael Laver and Ian Budge, 15?40. Basingstoke: Macmillan, 1992.   Mair, Peter. Party System Change: Approaches and Interpretations. Oxford: Oxford University Press, 1997.   Mair, Peter. ?Party System Change.? In Handbook of Party Politics, edited by Richard M. Katz and Crotty William, 63?74. London: Sage, 2006.  Meguid, Bonnie M. ?Competition Between Unequals: The Role of Mainstream Party Strategy on Niche Party Success.? The American Political Science Review XCIX, no. 3 (August 2005): 347?359.  Meguid, Bonnie M. Competition Between Unequals: Strategies and Electoral Fortunes in Western Europe. Cambridge: Cambridge University Press, 2010.  Meret, Susi. "The Danish People?s Party, the Italian Northern Lea gue and the Austrian Freedom Party in a Comparative Perspective : Party Ideology and Electoral Support." PhD Dissertation, SPIRIT Doctoral Programme, Aalborg University, Aalborg, 2009.  Nagel, Jack and Christopher Wlezien. "Centre-Party Strength and Major-Party Divergence in Britain, 1945?2005." British Journal of Political Science XL, no. 2 (2010): 279?304.  Pelinka, Anton. "The Haider Phenomenon in Austria: Examining the FP? in European Context." The Journal of the International Institute IX, no. 1 (2001), accessed August 2013, http://quod.lib.umich.edu/cgi/t/text/text-idx?c=jii;view=text;rgn=main;idno=4750978.0009.102  Pinard, Maurice. The Rise of a Third Party: A Study in Crisis Politics. Enlarged Edition. Manchester: Manchester University Press, 1975.   Rabinowitz, George, and Stuart Elaine Macdonald. ?A Directional Theory of Issue Voting.? American Political Science Review XXII, no. 4 (1989): 93?121.  Russell, Andrew and Edward Fieldhouse. Neither Left Nor Right: The Liberal Democrats   50 and the Electorate. Manchester: Manchester University Press, 2005.   Sartori, Giovanni. Parties and Party Systems: A Framework for Analysis. Cambridge: Cambridge University Press, 1976  Siaroff, Alan. "Two-and-a-Half-Party Systems and the Comparative Role of the `Half'." Party Politics IX, no. 3 (September 2003): 267?290.  Somer-Topcu, Zeynep. ?Timely Decisions: The Effects of Past National Elections on Party Policy Change.? The Journal of Politics LXXI, no. 1 (January 2009): 238?248.  Spiegel. "The Merkel Paradox: Chancellor Fares Well, But Government Tanks in Polls." Spiegel Online: International, November 19, 2012, accessed July 2013, http://www.spiegel.de/international/germany/germans-support-merkel-amid-doubts-about-government-performance-a-868034.html  The Local. "FDP Membership in 'Vicious Downward Spiral'." The Local: Germany's News in English, December 29, 2012, accessed July 2010, http://www.thelocal.de/politics/20121229-47037.html  Volkens, Andrea, Pola Lehmann, Nicolas Merz, Sven Regel, and Annika Werner. "The Manifesto Data Collection." Manifesto Project (MRG/CMP/MARPOR). 2012b. Berlin: Wissenschaftszentrum Berlin f?r Sozialforschung (WZB), 2012.  Ware, Alan. Political Parties and Party System. Oxford: Oxford University Press, 1996.   Wolinetz, Steven. ?Party Systems and Party System Types.? In Handbook of Party Politics, edited by Richard M. Katz and Crotty William, 51?62. London: Sage, 2006.      51 Appendices Appendix Table A: Historical Left-Right Policy Positions* Country Left Party/ies Third Party Right Party/ies Austria SP? (-14.76) FP? (3.00) ?VP (16.75) Canada ? NDP (-31.34) LIB (-3.56) CON (5.14) Germany SPD (-13.89) FDP (1.20) CDU/CSU (12.33) Ireland ? LAB (-21.29) FF (4.64) FG (8.40) Luxembourg LSAP (-29.01) CSV (-6.50) DPP (-4.74) ? *Parties? average RiLe scores from 1945?2011 are in parentheses. RiLe scores from 1970?1983 in Austria are excluded.   Appendix Table B: Historical Key Policy Positions* Country Policy [CMP Index] Far Party Near Party Third Party Austria Ethnonationalism [Ethnonat] SP? (3.20) ?VP (5.03) FP? (7.50) Canada Welfare [Welfare] CPN (9.34) LIB (13.58) NDP (20.27) Germany Free Market [Markeco] SPD (2.51) CDU/CSU (7.79) FDP (7.98) Ireland Welfare [Welfare] FF (8.70) FG (12.63) LAB (21.17) Luxembourg Free Market [Markeco] LSAP (1.13) CSV (2.72) DPP (4.96) *Parties? average key scores from 1945?2011 are in parentheses. Key policy scores from 1970?1983 in Austria are excluded.     52 Appendix Table C: Third Parties in Government ? Key Policies  Model 17 Model 18  Near shift in key (t) Far shift in key (t) Near shift in key (t ? 1) -0.329*** ?(0.0979)  Far shift in key (t ? 1)  -0.389*** ? (0.111) avg. all others? shifts in key (t ? 1) 0.513? ?(0.309) 0.400*** ?(0.134) Third shift in key (t ? 1) -0.426** ?(0.170) -0.325*** ? (0.0862) Third shift in seat% ? dummy (t ? 1) -5.990*** ? (1.647) -1.390? ? (0.921) Third in gov?t ? dummy (t ? 1) -5.595** ?(2.149) 0.456 ? (1.339) Third shift in key (t ? 1) X  Third in gov?t ? dummy (t ? 1) 0.403* ?(0.219) -0.00602 ?(0.163) Third shift in seat%  ? dummy (t ? 1) X  Third in gov?t  ? dummy (t ? 1) 6.738** ?(2.828) -0.317 ?(1.772) Constant 3.560*** ? (1.166) 0.374 ? (0.642) R2 0.403 0.316 Adjusted R2 0.343 0.246 Observations 77 77            Standard errors in parentheses            ? p<.15, * p<.10, ** p<.05, *** p<.01 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0072157/manifest

Comment

Related Items