Different Measures of the Ideological Positions of Political Parties: a Research Note from the Spanish Case

Different Measures of the Ideological Positions of Political Parties: a Research Note from the Spanish Case Luis Ramiro Fernández Researcher. Social and Political Sciences Department. European University Institute (Florence) Paper presented to the 27 th ECPR Joint Sessions of Workshops 26-31 March 1999 Workshop n 5: Estimating the Policy Positions of Political Actors English not revised by a native speaker. Do not cite without the permission of the author

The goal of this research note is simply to give a general view of the differences and similarities in the ideological position of Spanish parties using different techniques to locate them. Therefore, it is expected to show to which degree the results produced by these different techniques are equivalent and whether or not the movements of the parties along time are reflected in a consistent way. The results of three very common techniques used to establish the ideological or political positions of parties will be revisited: expert surveys, mass surveys, and content analysis of party manifestoes. The research note is divided in three main parts. First, party positions in the traditional left-right scale obtained through expert surveys will be compared with those derived from mass surveys. The expert surveys that will be cited in this note are those of Castles and Mair 1, Huber and Inglehart 2, and one expert survey carried on recently by myself. Second, different ways of establishing the position of parties in the left-right scale through mass surveys (party placement in the scale by their voters, party placement by the entire sample, and party voters self- placement) will be analyzed discussing whether or not they may be interchangeable. The goal here is to discover if the results of these different methods show differences small enough to make possible their use as interchangeable measures. Third, it will be revisited the analysis of party manifestoes. This is a partial approach, given that only one Spanish party is analyzed. The political party selected is Izquierda Unida (third Spanish political force in number of votes). In this last section the goal is to determine whether party movements along time in the left-right dimension, established through the analysis of its electoral platforms, match with its movements in the traditional left-right scale shown by mass and expert surveys. The next section deals with the comparison between the results of mass and expert surveys, begining with a brief description of the survey conducted a few months ago among Spanish experts. 1 F. Castles and P. Mair, Left-Right Political Scales: Some Expert Judgements, European Journal of Political Research, 12 (1984) 73-88. The exact date in which the survey was conducted is not detailed in this article. I have considered it to be 1983. 2 J. Huber and R. Inglehart, Expert Interpretations of Party Space and Party Locations in 42 Societies, Party Politics, vol. 1 No. 1 (1995) 73-111. The survey was conducted in 1993. 2

1. Expert and Mass Surveys I carried out an expert survey in the first weeks of 1999. The questionnaires were sent by post the 22nd of January and the last response was received the 19th February 3. Given that the survey was made just in the begining of 1999 it will be considered that the experts judgements refer to the parties positions as of in 1998 (or in the last months of 1998) and this date will be used in the following pages to name this expert survey. The questionnaire only included one question, asking the experts to locate a list of parties in the left-right scale. Two things should be specified. First, a list of parties was offered to the respondent asking him/her about the location of those parties. Therefore, the respondent had an indication of what parties he should include in his/her answer. Second, the questionnaire included a scale with ten cells explaining to the respondent that he/she could write the name of the parties inside the cells chosen or that if he/she prefered could express party location using decimals. The sample to which the questionnaire was sent was formed by 28 experts. Therefore, the N of the sample was high with a view to reduce the danger of high nonresponse in postal surveys and of the leverage of extreme cases in the calculus of the mean. The experts were selected among Spanish scholars specialized (i. e., having developed research or publications) in one of the following fields: political sociology, political parties, electoral sociology, and electoral competition. The experts belong to eight Spanish universities and research centres. The list of parties was formed by 14 nation-wide and non-nation-wide parties. The parties included were those that gained parliamentary representation in the last 1996 general election (except one, Partido Andalucista, that was included for reasons of comparability with other surveys). The response rate was high in comparison to normal standards of postal surveys and given the fact that a follow-up mailing was not used: 19 questionnaires over the 28 sent were returned (67,8%). This survey is different from the others cited in this note (Castles and Mair, and, Huber and Inglehart) at least in two important features. The main difference is the questionnaire. The inclusion of a list of parties that should be located is debatable and it is not the path followed in the two previous surveys. I prefered to adopt this strategy 3 One more questionnaire was received when this research note was already written: it will be incorporated in next jobs. 3

trying to obtain information not only on the main nation-wide and non-nation-wide parties but also on other less important (in national parliamentary terms) non-nationwide parties that are regionally relevant (and that are included in other surveys). It is clear that adopting this strategy, the respondent may feel an obligation to respond about parties on which he/she has not enough information to give a judgement or that does not consider relevant parties. Given the fact that the sample was formed by academic experts I thought that the risk was relatively low in following this strategy and I would be able to avoid the omission of the location of minor parties. Since two possible ways of answering were available (writing down party positions using decimals or writing down the party s name into the chosen cell of the scale) the intention was to let the respondents use, if necesary, a more sensitive mode of response (with decimals). The survey was also different in the bigger sample of experts who received the questionnaire (for example, Inglehart and Huber used a sample of 20 persons per country). As has been said, the goal was to minimize (given the limits of expert surveys) the influence of more extreme scorings in the calculus of the mean and assure an acceptable response rate. The possibility of appearance of extreme scores could be favoured by the presence of minor parties in the list, and therefore, the bigger sample should reduce their leverage. The results of the expert survey are presented in the next table (Table 1) including also the data from the surveys of Castles and Mair, and Huber and Inglehart. As it can be observed, the N of respondents was much higher than in the survey conducted by Huber and Inglehart. In several cases the respondents do not locate the 14 parties listed in the questionnaire (it may well be that respondents judged they could not answer apropriately or that they forgot these parties in their answer). Six parties were located by 19 experts, six parties by 18 and only two parties by 17 (in Huber and Inglehart s survey one party received three responses, the minimum and three parties received eight responses, the maximum). As a consequence, the variation in the location of each party is higher in my survey that in the previous ones. One half of the parties had a range of three or more scores. As expected, the increase in the number of respondents had as a consequence a parallel increase in the values of the standard deviations. It is interesting to note that, as it could also be anticipated, the higher standard deviations are shown by the minor parties (being the Catalonian nationalists of Convergència i Unió the main exception). 4

Table 1. Expert Surveys Castles and Mair 1983 Huber and Inglehart 1993 Own Survey 1998 Party Abbrev Range Mean Range Mean SD N Range Mean SD N Herri Batasuna HB 0-1.5 0.5 1 1.00 0.00 4 1-3 1.56 0.68 18 Unión Pueblo UPC 0-1 0.7 Canario Euskadiko Ezkerra EE 2-3 2.4 Partido Comunista PCE/IU 2-4 2.7 2-3 2.38 0.48 8 1.5-4 2.48 0.63 19 España/Izquierda Unida Partido Socialista PSOE 2.5-5 3.6 3-5 4.00 0.50 8 3-5 4,20 0.61 19 Obrero Español Esquerra Republicana de Catalunya ERC 2-6.5 4.1 2-5 3.55 0.93 18 Partido Socialista PSA/PA 2.5-6.5 4.5 4-7 5.52 0.99 17 de Andalucía/Partido Andalucista Convergència i Unió CiU 6.6 6-7 6.17 0.37 6 5-8 6.21 0.75 19 Partido Nacionalista PNV 6-8 6.7 6-7 6.14 0.35 7 6-8 6.55 0.59 19 Vasco Unión del Centro UCD 6.5-7.5 7.1 Democrático Partido Aragonés PAR 7.5-8.5 8.2 6-9 7.33 0.95 18 Regionalista Alianza AP/PP 7.5-9 8.4 7-8 7.50 0.50 8 6-8 7.47 0.56 19 Popular/Partido Popular Unión UN/FN 9-10 9.8 Nacional/Fuerza Nueva Coalición Canaria CC 4-8 6.05 0.90 18 Bloque Nacionalista BNG 2-4 2.91 0.71 18 Galego Unión Valenciana UV 6-9 7.60 1.03 19 Eusko Alkartasuna EA 3-6 4.67 1.25 3 4-7 5.10 0.80 18 Centro Democrático CDS 5-6 5.40 0.49 5 y Social Iniciativa Per Catalunya IC 2-4 3.05 0.63 17 In the next table (Table 2) it is possible to follow the evolution of Spanish party positions according to the averages derived from the three expert surveys cited. The main feature of the evolution of party positions that arise from the data is their moderation between 1983 and 1998. It seems that there is a tendency to moderation that makes parties adopt a more centered stance. However, this tendency has some 5

exceptions and it is not, according to these data, a phenomenon implying big movements. This aspect will be treated later contrasting these results with the results from mass surveys. Table 2. Expert Surveys: Means. Party Means 1983* 1993 1998 Herri Batasuna 1.45 1.00 1.56 Partido Comunista de España/Izquierda Unida 3.43 2.38 2.48 Partido Socialista Obrero Español 4.24 4.00 4,20 Alianza Popular/Partido Popular 8.56 7.50 7.47 Convergència i Unió 6.94 6.17 6.21 Partido Nacionalista Vasco 7.03 6.14 6.55 Esquerra Republicana de Catalunya 4.69 3.55 Partido Socialista de Andalucía/Partido 5.05 5.52 Andalucista Partido Aragonés Regionalista 8.38 7.33 Eusko Alkartasuna 4.67 5.10 Unión del Centro Democrático 7.39 Unión Nacional/Fuerza Nueva 9.82 Unión Pueblo Canario 1.63 Euskadiko Ezkerra 3.16 Coalición Canaria 6.05 Bloque Nacionalista Galego 2.91 Unión Valenciana 7.60 Iniciativa per Catalunya 3.05 Centro Democrático y Social 5.40 *The original scale was a 0-10 scale. Here, the means have been transformed into a 1-10 scale. The data from expert surveys are now going to be compared with those from mass surveys. The 1983 expert survey will be compared with the Data-1982 survey extracting the results from Montero 4. The 1982 post-electoral survey conducted by the CIS (Centro de Investigaciones Sociológicas) does not include the question about the position of parties in the left-right scales and because of this I do not use it. There is a survey done in 1980 (CIS 1980) but due to the time gap with the expert survey I have prefered not to use it. The data for 1993 came from the 1993 CIS post-electoral survey. The comparison with the expert survey of 1998 should be made only with the 1996 CIS post-electoral survey given that there are no more recent surveys including the question about party positions. However, this is only a provisional problem because the referred 4 J.R. Montero, Sobre las preferencias electorales en España: fragmentación y polarización (1977-1993), in P. Del Castillo (ed.), Comportamiento Político y Electoral, Madrid, CIS, 1994. 6

question will be surely included in the 1999 CIS post-electoral survey (European, regional, and local elections) and the analysis can be improved subsequently. Table 3. Mean Party Position. Mass Surveys. Party Means 1982 1993 1996 Herri Batasuna 1.69 2.25 Partido Comunista de España/Izquierda Unida 1.8 2.53 2.52 Partido Socialista Obrero Español 3.5 4.50 4.52 Alianza Popular/Partido Popular 8.5 7.88 7.93 Convergència i Unió 5.76 (6.36) 6.29 (6.26) Partido Nacionalista Vasco 5.51 (4.69) 6.05 (6.00) Esquerra Republicana de Catalunya 3.04 2.60 Partido Socialista de Andalucía/Partido Andalucista 5.13 Partido Aragonés Regionalista 6.17 Eusko Alkartasuna 3.50 4.06 Unión de Centro Democrático 6.2 Coalición Canaria 5.64 6.34 Bloque Nacionalista Galego 3.05 2.77 Unión Valenciana 6.65 6.99 Centro Democrático y Social 5.7 5.37 Sources: own elaboration from the following surveys; Data 1982 survey, 1993 CIS post-electoral survey, and 1996 CIS post-electoral survey. The numbers in brackets are the average scores obtained from the responses of the respective subsamples of Catalonia and Basque Country, respectively. This particular aspect will be treated with more detail later in the paper. In the next table (Table 4) party positions according to expert and mass surveys are compared. The comparison between 1983 and 1982 surveys produces contradictory evidences. On the one hand, in three cases (UCD, PSOE and PCE) over four the differences among party scores in expert and mass survey are relatively high (1.19, 0.74, and 1.63, respectively). On the other hand, the position of AP is identical in both surveys. Taking this into account, the mean difference among party positions measured with expert and mass surveys is 0.90. When comparing the average party positions in the 1993 surveys a lesser number of parties appear with big differences in their scores. In three cases over eight the differences between expert and mass survey positions are higher than 0.5 (HB: 0.69, CiU: 0.60, and EA: 1.17). In two more cases the difference is extremely low (IU: 0.15 and CDS: 0.03). The average difference of party locations measured with expert and mass surveys in 1993 is 0.48. Table 4. Mass and Expert Surveys. Comparison of Mean Party Positions by Pairs of Surveys 7

Party 1983/82 1993/93 1998/96 ES MS ES MS ES MS Herri Batasuna 1.00 1.69 1.56 2.25 Partido Comunista de España/Izquierda Unida 3.43 1.8 2.38 2.53 2.48 2.52 Partido Socialista Obrero Español 4.24 3.5 4.00 4.50 4,20 4.52 Alianza Popular/Partido Popular 8.56 8.5 7.50 7.88 7.47 7.93 Convergència i Unió 6.17 5.76 (6.36) 6.21 6.29 (6.26) Partido Nacionalista Vasco 6.14 5.51 (4.69) 6.55 6.05 (6.00) Esquerra Republicana de Catalunya 3.55 2.60 Partido Socialista de Andalucía/Partido 5.52 5.13 Andalucista Eusko Alkartasuna 4.67 3.50 5.10 4.06 Unión del Centro Democrático 7.39 6.2 Coalición Canaria 6.05 6.34 Bloque Nacionalista Galego 2.91 2.77 Unión Valenciana 7.60 6.99 Centro Democrático y Social 5.40 5.37 The numbers in brackets are the average scores obtained from the responses of the respective subsamples of Catalonia and Basque Country, respectively. This particular aspect will be treated with more detail later in the paper. Finally, in the 1998/1996 comparison, four parties over twelve show differences in their position measured by expert and mass surveys higher than 0.5 (HB: 0.69, EA: 1.04, ERC: 0.95, and UV: 0.61). Party scores are practically the same in another three cases (IU: 0.04, CiU: 0.08, and BNG: 0.14). The mean difference of party locations measured through both kinds of surveys remains in this case very similar to that of 1993: 0.45. It is difficult to discover some regularity or tendency in the bias of the measures. Perhaps, according to 1993 and 1996 data, mass surveys seem to locate the PSOE in a more moderate or centered position than expert surveys, while they locate the PP in more extreme positions. However, it is risky to extract any clear pattern. It should be remembered that there are no data from 1996 onwards, the year in which the PP won the elections and entered the national government. It remains not clear if the PP 8

Left-Right Scale (1-10) governmental policies have changed decissively mass perceptions on its left-right position. In the next graphs we can see more easily the similarities and disparities between the results of expert and mass surveys and whether or not party movements resulting from both techniques produce similar trends. I have chosen the five main parties in the current Spanish party system (PP, PSOE, IU, CIU, and PNV). They are also parties that, with more or less intense organizational and political changes, have hade a continuous presence since the begining of the democratic transition in the 1970s. Graph 1. AP/CP/PP: Party position (Expert and Mass surveys) 10 9 8 7 6 5 Expert Mass 4 3 2 1 1980 1982 1983 1986 1989 1993 1996 1998 Year of survey 9

As shown in Graph 1 there are scarce differences between the ideological positions of Partido Popular (PP) following expert and mass surveys. The party has moved towards more moderate and centered positions, although in more centered locations for experts than for the respondents of mass surveys. In the case of the Partido Socialista Obrero Español (PSOE), presented in Graph 2, the differences between the locations given in expert and mass surveys are also limited. However, the party s movements appear slightly different, although the disparities have a reduced magnitude. While, following mass surveys, the PSOE has experimented a process of moderation since 1982, the result from expert surveys shows a constant position very near the score four in the left-right scale. The trajectory of Izquierda Unida (IU) shows a significant difference between its position according to the 1983 expert survey and the 1982 mass survey (Graph 3). From this point of departure, the trajectory of IU appears different following expert surveys or mass surveys (slight moderation in mass surveys or radicalization in expert surveys). However, leaving aside the different location of IU in the expert and mass surveys of 1983/1982, the close position of the party in subsequent years using the different surveys is striking. 10

Left-Right Scale (1-10) Graph 2. PSOE: Party positions (Expert and Mass surveys) 10 9 8 7 6 5 Expert Mass 4 3 2 1 1980 1982 1983 1986 1989 1993 1996 1998 Year of survey 11

Left-Right Scale (1-10) Graph 3. PCE/IU Party positions (Expert and Mass surveys) 10 9 8 7 6 5 Expert Mass 4 3 2 1 1980 1982 1983 1986 1989 1993 1996 1998 Year of survey There are two more parties to be observed, Convergència i Unió (CiU) and Partido Nacionalista Vasco (PNV). For these two parties it was not given their position in 1982/1982 because I do not have data for that date from mass surveys. This is the reason why they are not present in the comparison between the results of expert and mass surveys for 1983/1982 done in previous pages, and the reason why there is no score in 1982 for both parties in the next graphs. Clarified this, the Graph 4 remarks the limited difference in the trajectory of CiU according to the results of expert surveys or mass surveys. The party seems to follow a similar movement with a slight moderation of 12

Left-Right Scale Left-Right (1-10) Scale (1-10) its ideological position, although the mass surveys reflect certain oscillations in 1989 and 1993. Finally, the case of PNV, shown in Graph 5, seems to produce a less clear party movement and, to a certain degree, a greater difference between expert and mass surveys results. The analysis would benefit from a greater availability of data but it is clear that the different results of expert and mass surveys, even if existent, are not very high and they show a consisten trend in reflecting a party movement towards the right. Graph 4. CiU: Party Positions (Expert and Mass Surveys) 10 9 8 10 7 9 6 8 5 7 4 6 3 5 2 4 1 3 2 1 Graph 5. PNV: Party Positions (Expert and Mass Surveys) 1980 1982 1983 1986 1989 1993 1996 1998 Year of Survey 1980 1982 1983 1986 1989 1993 1996 1998 Year of Survey Expert Mass Expert Mass 13

Left-Right Scale (1-10) The two next graphs (Graph 6 and Graph 7) put together the trajectories of the five main parties analysed according to expert surveys and mass surveys. The observation of these two graphs highlight the scarce differences in the trajectories of parties following the results of expert surveys or mass surveys. It is also visible the slow convergence towards the centre that PSOE and PP follow, the parallel movements of PSOE and IU in the left (especially clear in Graph 7), and the also parallel trajectory of PNV and CiU. 10 Graph 6. Party Positions: Expert Surveys 9 8 7 6 5 4 PCE/IU PSOE AP/CP/PP CiU PNV 3 2 1 1983 1986 1989 1993 1996 1998 Year of Survey 14

Left-Right Scale (1-10) 10 Graph 7. Party Positions: Mass Surveys 9 8 7 6 5 4 PCE/IU PSOE AP/CP/PP CiU PNV 3 2 1 1980 1982 1983 1986 1989 1993 1996 Year of Survey 15

A peculiarity, originated in the strong presence of non-nation-wide parties in the Spanish party system, emerges from tables 3 and 4. In some cases, the location of certain parties is only asked in the regions in which they act, while in other cases the entire sample is asked about the location of some non-nation-wide parties. This fact enables to discover whether o not party positions are significantly different when the entire sample responds and when only the inhabitants of these regions respond. The cases for which it is possible to find this peculiarity are Convergència i Unió (the main Catalonian nationalist party) -in 1993 and 1996- and Partido Nacionalista Vasco (the main Basque nationalist party) -only in 1996. In tables 3 and 4 party positions according to the answer of the inhabitants of Catalonia and Basque Country were written in brackets. As shown in Table 5, the differences between party positions considering the entire sample or only the respective regional inhabitants are only significant for CiU in 1993. In the other cases there is no difference at all. Table 5. Party Placement. The cases of Convergència i Unió and Partido Nacionalista Vasco Partido Convergència i Unió Nacionalista Vasco 1993 1996 1996 Party Placement by Population (nation-wide) 5.76 (2454) 6.29 (2966) 6.05 (2636) Party Placement by Population (Catalonia) 6.36 (629) 6.26 (691) Party Placement by Population (Basque Country) 6.00 (227) In brackets, the number of cases upon which the average has been calculated 2. Party placement by their voters, by the population and party voters self-placement In the previous pages party positions obtained through mass surveys were extracted from the scores that the entire sample gave to the parties in the left-right scale. 16

There are, at least, two other ways, to a certain degree supplementary, for obtaining party positions: 1) the party location assigned by its own voters, and 2) the selfplacement of the party s voters 5. These three possible measures have different meanings and, as it is well known, there is usually no total match between their results. In fact, the spatial theory of voting and the spatial theory of party competition are based in the existence of these differences or spatial gaps between parties and voters. Therefore, in this section the analysis will consider not whether these differences exist, but which is their magnitude and in which cases their size is relevant. The goal is to try to clarify whether or not the three measures considered (party placement by the population, party placement by their voters, and party voters self-placement) are interchangeable for the study of party positions in the Spanish case. This analysis can be useful to determine if in the absence of one of the above mentioned measures (especially, party placement by population 6 ) the others can be used alternatively (specially, party voters self-placement) without introducing important biases. In this section four post-electoral CIS surveys -1986, 1989, 1993 and 1996- will be used. In the next table (Table 6) the data on party placement by the population, by their voters, and party voters self-placement are shown for the parties included in the surveys cited let the analysis 7. 5 See, G. Sani and G. Sartori, Polarization, Fragmentation and Competition in Western Democracies, in H. Daalder and P. Mair (eds.), Western European Party Systems, London, Sage, 1983. 6 This question is absent in a big number of CIS surveys. 7 The parties included have a minimun of 15 voters in the sample. This criterium rendered it impossible to include some regional parties. 17

Table 6. Placement of parties by their voters, by the entire population, and self-placement of the voters of different parties Parties 1986 1989 Party Placement by their Voters Party Placement by Population Party Voters Self-placement Party Placement by their Voters Party Placement by Population Party Voters Self-placement Partido Socialista Obrero Español 3.65 3.73 3.67 3.71 4.29 3.71 Izquierda Unida 2.18 1.94 2.53 2.28 2.25 2.64 Coalición Popular/Partido Popular 8.00 8.44 7.19 7.71 8.19 7.17 Centro Democrático y Social 5.14 5.36 5.12 5.39 6.03 5.12 Convergència i Unió 6.06 6.44 5.58 6.04 6.63 5.63 Partido Nacionalista Vasco 5.10 5.50 4.66 5.63 5.96 4.66 Herri Batasuna 1.90 1.59 2.08 Partido Andalucista 4.54 4.69 4.50 Unión Valenciana 5.87 6.21 5.75 18

Table 6. (Continued) Parties 1993 1996 Party Placement by their Voters Party Placement by Population Party Voters Self-placement Party Placement by their Voters Party Placement by Population Party Voters Self-placement Partido Socialista Obrero Español 4.15 4.50 3.64 4.00 4.52 3.67 Izquierda Unida 2.65 2.53 2.90 2.55 2.52 2.93 Coalición Popular/Partido Popular 7.55 7.88 6.85 7.30 7.93 6.50 Centro Democrático y Social 5.55 5.37 5.20 Convergència i Unió 6.31 6.36 5.68 6.21 6.26 5.29 Partido Nacionalista Vasco 4.48 4.69 4.43 5.82 6.00 5.00 Herri Batasuna 1.56 1.69 1.95 1.57 2.25 2.16 Esquerra Republicana de Catalunya 3.04 3.04 3.42 2.61 2.60 3.42 Coalición Canaria 5.60 5.64 4.67 5.69 6.34 5.43 Bloque Nacionalista Galego 5.14 3.05 3.20 2.98 2.77 3.24 Eusko Alkartasuna 4.38 3.50 3.92 Partido Andalucista 4.72 5.13 4.26 Unión Valenciana 6.00 6.99 5.22 19

According to the results, the mean difference in 1986 between party placement by population and party voters self-placement was 0.64, the mean difference between party placement by the population and party placement by their voters was 0.29, and the mean difference between party placement by their voters and party voters self-placement was 0.35. For 1989 the mean difference between party placement by population and party voters self-placement was 0.70, the mean difference between party placement by the population and party placement by their voters was 0.38, and the mean difference between party placement by their voters and party voters self-placement was 0.32. For 1993 the mean difference between party placement by population and party voters self-placement was 0.50, the mean difference between party placement by the population and party placement by their voters was 0.39, and the mean difference between party placement by their voters and party voters self-placement was 0.59. However, this survey shows an absolutely surprising value (5.14) in the party placement by its voters for the case of Bloque Nacionalista Galego (BNG), a left nationalist party from Galicia. Given the fact that this value may be very influenced by the presence of extreme cases in the limited sample of BNG voters present in the 1993 survey, I prefer to exclude the BNG of the calculus, so the differences in 1993 are as follows. Mean difference between party placement by population and party voters self-placement: 0.53. Mean difference between party placement by the population and party placement by their voters: 0.22. Mean difference between party placement by their voters and party voters self-placement: 0.45. Finally, for 1993 the mean difference between party placement by population and party voters self-placement was 0.87, the mean difference between party placement by the population and party placement by their voters was 0.39, and the mean difference between party placement by their voters and party voters self-placement was 0.58. Therefore, as expected, in the four years analyzed the bigger difference was that between party placement by population and party voters self-placement: 0.64, 0.70, 0.53 and 0.87. In three cases the smaller difference was found in the comparison between party placement by their voters and party placement by the population (0.29 in 1986, 0.38 in 1989, 0.22 in 1993, and 0.39 in 1996). The mean differences between party placement by their voters and party voters self-placement were: 0.35 (1986), 0.32 20

(1989), 0.45 (1993), and 0.58 (1996). Being party placement by the population the standard measure for party location, and understanding party voters self-placement as subsidiary or secondary measures, the data indicate that there is a difference between them that could be considered relevant. The mean values cited in previous paragraphs hide important differences among parties and years. While the mean difference between party placement by population and party voters self-placement in the 1986-1996 period was 0.44 for IU, it was 0.58 for PSOE, and 1.18 for PP. In relation with the differences between party placement by their voters and party voters self-placement the mean values for the period were: 0.21 for PSOE, 0.33 for IU, and 0.71 for PP. Finally, the mean differences between party placement by their voters and party placement by the population were: 0.38 for PSOE, 0.47 for PP, and 0.10 for IU. Therefore, these measures are more interchangeable in the case of IU and less in that of PP. 3. The analysis of party manifestoes: the case of Izquierda Unida In the next pages another technique will be used to establish the political position of parties: the analysis of party manifestoes or electoral programmes. The aim of this section is to delineate the political position of IU through the emphasis reflected in their electoral programmes, to locate IU in the left-right scale using these programmes, and to determine whether or not the evolution along time of the positions of IU in the left-right scale according to the analysis of electoral programmes is consistent with the trajectory of the IU in the left-right space according to the expert and mass surveys shown in previous pages. It should be recognized that the choice of IU for this analysis is based in pragmatic reasons. I am conducting a more wider research on IU for my Ph. D., and, therefore, I know with detail not only its platforms but also the processes by which the manifestoes are elaborated. Five electoral programmes corresponding to the last five Spanish general elections will be analyzed. The first one is that of 1982. This electoral programme corresponds to the Partido Comunista de España and not to IU. IU was created before 21

the 1986 elections but the inclusion of the 1982 Communist electoral programme has been judged important with the aim of comparing IU stances with that of its antecedent electoral organization. IU electoral manifestoes correspond to the 1986, 1989, 1993 and 1996 general elections. They are extremely different in size. While the 1982 programme was a document of 48 pages, the last programme (1996) had 280 pages. Obviously, these differences have a direct translation in the number of sentences and quasisentences to be analyzed. There is a clear tendency to produce a longer and more detailed programme, as in other countries, and this fact has its most evident example in the impresive 1996 electoral programme. The technique of content analysis used was that explained and used in Budge, Robertson and Hearl 8. The coding categories were also those proposed in that comparative research. Therefore, the analysis of IU electoral manifestoes was made coding each sentence of the programmes in one of the categories previously established in the international analysis of party manifestoes. The unit of analysis was the sentence, in spite of the long programmes, but also the quasi-sentences were considered as coding units when the length and nature of the sentence made adviseable its division. As it has been said, the general coding frame used was the same of Budge, Robertson and Hearl, but a few sub-categories were added. The main sub-categories added were three also included in the 1987 analysis of Italian manifestoes: Administrative and State Reform, and Judicial, Legal and Police Reform (included in the category Government Efficiency); and Support of Employment (included in Keynesian Demand Management). Other sub-categories were used but in this case are exclusively related to the particular analysis of IU. An important number of categories were not applicable or showed a very low use. A total of 27 categories obtained more than 1% of the total number of coding units in some of the five electoral programmes. It should be considered, in this respect, that IU represents the most left-wing party among the three or five more relevant Spanish parties, and this fact produces a natural low presence, or even absence, of certain issues in its platforms. The coding procedure was carried out only by myself and, unfortunately, that rendered impossible to test inter-coder reliability. Besides, there is an important difference in the proportion of coding units that could not be coded. The 8 I. Budge, D. Robertson, and D. Hearl (eds.), Ideology, Strategy and Party Change: Spatial Analyses of Post-war Election Programmes in 19 Democracies, Cambridge & New York, Cambridge University 22

average, considering the five programmes was 12.05 but in one case there was a relatively higher proportion of non-codifianle sentences compared to the total coding units (slightly more than 20% in 1996). Besides this, the proportion of uncodifiable sentences was clearly related with the length of the programme: they increased with the length of the document (longer platforms have more general declarations and, therefore, more uncodifiable units). Given that there is a big difference in the proportion of coding units that could not be coded among the five programmes I have preferred to include both the percentages of the categories over the total number of coding units, and over the total number of coding units effectively coded. In the next table I present the percentages corresponding to the categories that obtained more than 1% in any of the five electoral programmes. Press, 1987. 23

Table 7. Categories with more than 1% Category % over total number of sentences % over total number of codifiable sentences 1996 1993 1989 1986 1982 1996 1993 1989 1986 1982 Foreign Special 0.08 0.12 1.22 0.63 0.36 0.12 0.16 1.50 0.75 0.43 Relations: Positive Military: Negative 1.55 2.09 2.37 2.77 4.52 2.19 2.72 2.91 3.3 5.33 Peace 1.34 0.51 2.76 1.89 0.72 1.89 0.67 3.38 2.25 0.85 Internationalism: 1.49 1.54 2.68 0.50 1.44 2.10 2.01 3.28 0.6 1.71 Positive European Community: 0.46 0.32 0.92 0.25 1.08 0.66 0.41 1.12 0.3 1.28 Positive Democracy 4.33 4.64 3.22 7.57 3.98 6.09 6.03 3.94 9 4.69 Constitutionalism: 0.34 0.35 0.30 0.63 0.90 0.48 0.46 0.37 0.75 1.07 Positive Decentralization: Positive 3.86 3.51 3.68 2.77 3.62 5.43 4.56 4.50 3.3 4.26 Government Efficiency 9.94 8.06 5.21 9.84 7.60 13.99 10.46 6.38 11.7 8.96 Regulation of 0.89 1.41 2.91 1.26 1.26 1.26 1.84 3.56 1.5 1.49 Capitalism Economic Planning 0.06 0.64 0.69 2.02 0.54 0.09 0.83 0.84 2.4 0.64 Economic Goals 0.06 0.03 0.46 1.13 0 0.09 0.04 0.56 1.35 0 Keynesian Demand 1.19 1.89 1.99 2.39 4.16 1.68 2.47 2.44 2.58 4.9 Management Productivity 1.13 1.48 0.61 0.75 1.26 1.59 1.92 0.75 0.9 1.49 Technology 4.22 5.61 4.98 3.91 2.89 5.94 7.28 6.10 4.65 3.41 Controlled Economy 1.49 1.80 2.60 1.01 3.44 2.10 2.34 3.19 1.2 4.05 Nationalization 1.28 0.74 2.53 0.88 2.71 1.80 0.96 3.09 1.05 3.2 Environment 10.24 11.06 6.44 6.43 3.80 14.41 14.36 7.88 7.65 4.48 Art, Sport, Leisure, 2.53 3.87 1.76 4.92 4.34 3.57 5.02 2.16 5.85 5.12 and Media Social Justice 1.13 0.96 2.60 2.27 0 1.59 1.25 3.19 2.7 0 Social Services 5.29 5.80 7.59 2.77 5.79 7.44 7.53 9.29 3.3 6.82 Expansion: Positive Education: 3.03 7.93 8.43 6.06 5.43 4.26 10.3 10.33 7.2 6.4 pro-expansion Traditional Morality: 0.49 1.03 0.23 0 0 0.69 1.34 0.28 0 0 Negative Labor Groups: 2.26 4.41 5.82 4.67 12.8 3.18 5.73 7.13 5.55 15.17 Positive Agriculture and 1.66 1.09 2.53 1.01 3.26 2.34 1.42 3.09 1.2 3.84 Farmers Underprivileged Minority Groups 2.36 0.16 0.61 1.76 1.99 3.33 0.20 0.75 2.1 2.35 Noneconomic Demographic Groups 6.99 4.64 5.75 10.73 5.43 9.84 6.03 7.04 12.7 6.4 As it results, there is a consistency in the thematic emphases across the five programmes but, at the same time, there are important evolutions in those emphases. Among the categories related with the economy only two, Technology and Keynesian Demand Management, appear highly mentioned; but only the last one could be considered as a traditional left-wing issue (in fact, the majority of the mentions to 24

Keynesian Demand Management make reference to Support of Employment). Adding the percentages of the more leftist categories in the domain of the economy (Regulation of Capitalism, Economic Planning, Controlled Economy, and Nationalization) the data are the following: 3.72 in 1996, 4.59 in 1993, 8.73 in 1989, 5.17 in 1986, and 7.97. Therefore, it seems that a tendency in reducing the emphases in such categories exists. Also the emphasis in support to Labor Groups could be decreasing. Other traditional left-wing issues as Education or Social Services show an importan weight. The clearest pattern of increasing emphasis is shown in Environment. Hence, the relative emphases seem to reflect the combination in IU s electoral platforms of traditional leftist concerns and new issues. Finally, the notable weight of Decentralization could be understood as the result of Spanish intense debates around the territorial organization of the State. In the next table (Table 8) the ten categories that have the bigger proportion of mentions are summarized. Table 8. More important emphases 1996 1993 1989 1986 1982 Military: Negative 1.55 2.09 2.37 2.77 4.52 Democracy 4.33 4.64 3.22 7.57 3.98 Decentralization: Positive 3.86 3.51 3.68 2.77 3.62 Government Efficiency 9.94 8.06 5.21 9.84 7.60 Keinesian Demand Management 1.19 1.89 1.99 2.39 4.16 Technology 4.22 5.61 4.98 3.91 2.89 Environment 10.24 11.06 6.44 6.43 3.80 Art, Sport, Leisure and Media 2.53 3.87 1.76 4.92 4.34 Social Services Expansion: Positive 5.29 5.80 7.59 2.77 5.79 Education: Pro-expansion 3.03 7.93 8.43 6.06 5.43 Labor Groups: Positive 2.26 4.41 5.82 4.67 12.8 Farmers and Agriculture 1.66 1.09 2.53 1.01 3.26 Noneconomic Demographic Groups 6.99 4.64 5.75 10.73 5.43 Underlined figures are the ten higher percentages for each electoral programme. The analysis of party manifestoes enables to calculate a measure of the location of those playforms in the left-right scale and, therefore, it is a technique to locate parties in the left-right dimension. Given that I have only analyzed the electoral platforms of one Spanish party I can not locate IU in the left-right continuum through factor 25

analysis 9. Hence, I will use exclusively the method used by Klingeman, Hofferbert and Budge 10 to locate parties in the left-right space throuhg party manifestoes. This technique consists, as it is well known, in a substraction between the percentages corresponding to left-wing categories and right-wing categories. According to this calculus, the positions of IU in the years considered were the following. Table 9. Party Position in the Left-Right scale: Party Manifestoes Year of Party Manifesto Calculus made using the % over the total number of sentences Calculus made using the % over the total nuber of codifiable sentences 1982 40.23 47.35 1986 26.52 31.5 1989 40.88 50 1993 29.78 38.65 1996 21.46 30.21 Given that the calculus consists in substracting from the percentages of the selected left-wing categories the percentages of the selected right-wing categories, a positive number indicate a leftist value. In this way, a bigger positive number indicates a more leftist party position (in this calculus the scale would be from 100, all the mentions in the left-wing categories selected, to -100, all the mentions in the righ-wing categories selected). Therefore, the evolution of IU would indicate a progressive moderation in its left-wing stances, only with a radicalization in 1989 always according to the calculation of the position with Klingemann, Hofferbert and Budge s method. Comparing the data on the trajectory of IU according to the analysis of party manifestoes (Graph 8), with the data shown in previous pages on the ideological evolution of IU in the left-right scale (1-10) according to expert and mass surveys (Graph 3), there is a coincidence in showing a movement of slight moderation. 9 P. Mair, Locating Irish Political Parties on a Left-Right Dimension: An Empirical Enquiry, Political Studies 34 (1986) 456-465. S. Bartolini and P. Mair, Identity, Competition and Electoral Availability, Cambridge & New York, Cambridge University Press, 1990. 10 H.D. Klingemann, R.I. Hofferbert and I. Budge, Parties, Policies, and Democracy, Boulder, Westview Press, 1994. 26

Left-Right Scale (-100 to 100) Izquierda Unida: Left-Right position of party manifestoes 100 80 60 40 20 0-20 -40 % over total sentences % over codifiable sentences -60-80 -100 1982 1986 1989 1993 1996 Years of party manifestoes 27

4. Concluding Remarks This research note has shown that different measures of locating political parties in the left-right dimension frequently bring about similar results (at least in the Spanish case). In spite of this, a certain caution should be made, since the results match differently for different parties: while results were more consistent in the case of IU it was less so in the case of PP and of smaller parties. The analysis of party manifestoes is more easily biased by certain ideosincracies of the platform that may not correspond to real trends in party positions. For instance, in the case of IU, the 1989 platform appears to be more left-wing than all the rest, while the other measures are not consistent with this piece of information, and neither would be the qualitative knowledge of this party positions during that time. Furthermore, in the case of IU, the party positions appear too centred. It should be considered that the location of IU in the left-right scale according to the analysis of electoral manifestoes, has been made applying a distribution of categories between left and right that was extracted from the analysis of party platforms between 1945 and the late 1980s. When analyzing in detail a party s manifestoes, it seems clear that many coding categories that should be incorporated as a left category, are not so. On the contrary, many of those currently considered to form the left pole of the dimension are relatively unimportant in the 1980s and 1990s platforms. In conclusion, a reconsideration of which are the issues around which the left-right competition is established is probably necessary. In general terms, the comparability of all of these measures in the Spanish case is greater when trying to determine the longitudinal trend of party positions, given that the differences between them are minor with respect to the evolution of locations in time. 28