Footnote 45, Chapter 4, Page 98:

The compilation of Minority Veto data is described on pages 97-102. A more detailed version of that section is provided here. The first data indicating the presence of minority groups’ veto power (MV) in a large set of countries were made for analysis here. With the exception of two sets of cases which are described in Appendix E, this data was made for all of the country cases, including those four which Lijphart has identified most consistently as having experienced long-term success with consociation. These data values were calculated through reference to demographic statistics and the contents of all constitutions and constitutional amendments that could be found which governed each country’s entire populations. The unfeasibility of finding out whether groups veto policies in practice meant that the MV variable is confined to indicating whether, if groups were represented in legislative bodies in proportion to their sizes, they could exert a veto resulting in a changed policy decision. This criterion contains a highly significant and unfortunately necessary assumption. However, if the MV and PR variables are used in conjunction with each other in the statistical analysis, inability of groups to gain such representation at least will be partly controlled for. With the exception of “schedules” used to provide details and attached to some constitutions, every word of each constitutional document used between 1975 and 1995 and found for use in this analysis was read in search of provisions that could result in MV. An inventory was made that lists all documents found for each state, the few that were not found, and the periods of time they were used. Individual pieces of data were made for each year of a country’s experience, which are particularly important because some countries only used constitutions for a few years of this period. Data could only be made for years when such documents were in force. Notes made for each document were compared to the percentages of each country’s population made up of individual MARs and all of its MARs collectively. If MARs and countries were listed in Gurr’s 1993 publication, the 1990 population figures used in it were provided for this purpose. If they were included in his 2000 publication but not the 1993 one, the 1998 figures provided in the 2000 publication were referred to. Austria is the only one of Lijphart’s core country examples of consociation for which MV data could not be made and it was not included in Gurr’s study. Since two of those core country examples, Belgium and the Netherlands, were not identified as containing MARs for Gurr’s dataset, the population figures used to make MV data for them were taken from the 2004 edition of the CIA World Factbook. If it is found that one MAR in a country could potentially exert a veto resulting in a changed policy decision, if it was proportionally represented in legislative bodies, each year when that was true is allocated a “1” for this variable. If this feat could be accomplished by all of a country’s MARs working in concert, that case is given a “.5.” If neither of these scenarios was possible in a year of a country’s experience, it was allocated a “0.” Lijphart’s core country examples other than Austria, which are Belgium, the Netherlands, and Switzerland, were all found to provide potential MV power for each year between 1975 and 1995, corresponding to a value of “1” for this variable. This analysis contains the first data designed to represent the ability of minority groups to veto their governments’ policy decisions, in a large set of countries. However, the absence of earlier comparative, and particularly quantitative, work in this area severely limited the existing sources that could be used to make this data. Therefore, there are a number of drawbacks related to the applicability of the data which need to be recognized. The detection of rules potentially allowing this veto power had to be restricted to analysis of constitutions because information concerning its actual use was unavailable. Representation of actual use also was not feasible, given the resources available for construction of this analysis. In a couple cases when I knew that information contained in constitutions did not reflect the reality in a specific country, I did a little more research through internet searches to confirm my impression and made the data reflect reality. For example, one of the MARs for South Africa, from as early as 1993, is “Europeans.” If one MAR could possess potential veto power, its country was usually given the highest “1” score. However, its constitutional documents before 1993 indicated that all legislative decision-makers had to be “white” so they were the only people capable of exerting this veto and I knew that they dominated all policy there. For this reason, although Europeans had lots of veto power, South Africa is coded “0” until the very end of the 1975-1995 period because this “minority at risk” was at risk regarding some matters but completely dominant in terms of political power. Unfortunately, the data is also adversely affected by the fact that data showing the exact proportions of legislative seats held by members of different cultural groups is also unavailable. Keefer's World Bank Database of Political Institutions contains information showing whether any of the three largest government parties during each year of country’s experience were “nationalist” or “religious” but this information only covers members whose parties are in government. These individuals also cannot be relied upon even to be representatives of "nationalist" or "religious" types of political parties. Rather than representing the existence of veto power over policy initiatives by MARs, the data for this variable actually corresponds to the existence of MV in countries with operational, single, written constitutions that reflect real conditions and relatively proportional representation of MARs in decision-making bodies. Another drawback of this data is that it inadequately represents the domination of some majority groups by minority ones. “Minorities at Risk” that are actually numerical majorities are very rare within Gurr’s dataset but they do exist in a few countries. The method for constructing this data treats all MARs the same so, if one of a country’s MARs had a numerical majority of the population and was judged to have potential veto power, the country was given a positive “1” score. However, examination of Gurr’s cases indicates that the countries with a majority MAR either had potential veto power for other MARs as well or the majority MAR was labeled an MAR because it was dominated by a smaller group. The first category resulted in no introduction of inaccuracy. For countries in the second category, constitutional provisions indicating potential veto power for the majority MAR did constitute promises of empowerment for groups that traditionally were politically dominated. However, this second category of countries, consisting of South Africa, Burundi, and Iraq, often did not operate in practice in a manner reflecting this power portrayed in their constitutional documents. That is a problem related to the drawbacks of using constitutions to make this data, rather than treating all MARs the same way. Yet another regrettable aspect of this data is that, since countries are given a positive “.5” score if all of their MARs can apparently exert a veto if they act together, countries whose populations contain large proportions of MARs are actually more likely to be coded as if they have more MV. Introduction of an additional independent variable corresponding to size of population comprised of all MARs could control for this effect but such a variable would have a systemic, directly proportional, and unfalsifiable relationship with the dependent variable, stability. More people at risk in a society obviously makes it more likely to be unstable. It should also be pointed out that one piece of data exists for each year that each country was using a constitution, unless a copy of that document could not be collected. Since most constitutions covered several years of individual countries’ experiences, a country is often allocated the same value for all years between 1975 and 1995. Values of “0” were not allocated to years for which data was unavailable because that would introduce a problematic bias, since countries with constitutions that were difficult to find or unformalized constitutions would be penalized. Notwithstanding all of the small drawbacks associated with this data, it is the first designed to represent minority veto power and it will provide great insight into the operation of this core component of consociation. There are exceptions to this methodology for the creation of data representing MV. The following paragraphs list and discuss country-specific exceptions for independent variable data. With regard to MV data, brief qualitative analysis of the individual countries which make up these categories suggests that the lack of data for them probably does not introduce substantial bias. Cases without MV data include countries widely considered to have experienced successful consociation, such as Austria and Canada, but also others such as Djibouti and Saudi Arabia which it seems reasonable to assume have not provided MV power, as it is defined for this project. Exceptions to Data Coding Methodologies Involving Individual Countries If the methodology used to code a country’s data differed from that used to code the other countries’ data, I included that information in the new Appendix E. However, if a country’s data is mentioned because it is of interest, even though it was coded using the same methodology, I have retained discussions of it in the text. Appendix E is organized as you suggested, listing methodology exceptions by country. Also, where each of these exceptions existed in the text, there is now a footnote directing the reader to Appendix E. Austria: (Methodology involving minority veto data): There is no minority veto data for Austria. The minority veto data were calculated for all of the country cases, except for those falling into two categories. The first contains a very small set of countries for which access to constitutional documents could not be obtained. The second group includes countries which have not had operational, codified, single constitution documents operating between 1975 and 1995, such as Austria, Bhutan, Canada, Israel, New Zealand, Saudi Arabia, and the United Kingdom. This group also contains four countries which adopted their constitutions at the end of 1995 or in 1996. These include Azerbaijan, Bosnia, Georgia, and Ukraine. Although some of these countries have systems of government which use groups of documents that are internationally, collectively recognized to be analogous to constitutions, these groups of documents seem to cover significantly more subjects than normal, codified constitutions. Since they contain more material, it is likely that these countries’ “constitutions” would generally tend to have more mentions of potential MV so their inclusion in my MV dataset would introduce a systematic bias in favor of identifying these countries as containing MV. Some countries like Bhutan simply did not use anything like a constitution during the 1975-1995 period so they also cannot be represented by this MV data. These two categories of exclusions together comprise 15 countries for which no MV data is supplied. Azerbaijan: (Methodology involving minority veto data): There is no minority veto data for Azerbaijan because its constitution was adopted in November, 1995. Bahrain: (Methodology involving minority veto data): There is no minority veto data for Bahrain because I was not able to obtain the country’s constitutional documents, at the time I was making this data. Bhutan: (Methodology involving minority veto data): Data involving minority veto for Bhutan was not made because its constitutional documents are not sufficiently comparable to those of most other countries. Bolivia: (Methodology involving minority veto data): There is no minority veto data for Bolivia because I was not able to obtain the country’s constitutional documents, at the time I was making this data. Bosnia: (Methodology involving minority veto data): There is no minority veto data for Bosnia because its constitution was adopted in December, 1995. Canada: (Methodology involving minority veto data): Data involving minority veto for Canada was not made because its constitutional documents are not sufficiently comparable to those of most other countries. Czechoslovakia: (Methodology involving proportionality data): Keefer’s dataset includes a variable indicating whether, if multiple political parties are legal, a country uses PR or does not use PR. His dataset had missing values for a substantial number of years of countries’ experiences and data could not be made to replace these missing values. The one exception to this is that some missing values for Czechoslovakia were compiled, and included in the dataset, for this project. Djibouti: (Methodology involving minority veto data): There is no minority veto data for Djibouti because I was not able to obtain the country’s constitutional documents, at the time I was making this data. It seems that its first constitution was adopted in 1992. Fiji: (Methodology involving segmental autonomy and, in particular, boundary coincidence): Except for Fiji, all of the other country cases were allocated the same value corresponding to segmental autonomy for each year of their experiences covered in this dataset. Fiji’s data for this variable includes two different values because information was found that indicated the presence of segmental autonomy involving the country’s demographic situation, beginning in 1987. This information involves the extent to which the boundaries of its geographically delineated constituencies and its group boundaries coincide. Fiji was originally coded as not containing such boundary coincidence but Reilly’s discussion of the country’s experiences indicates that, while geographic and group boundaries have not coincided between 1975 and 1995, the electoral system operated to produce the same sort of results to be expected from such coincidence between 1987 and 1995. In a situation with perfect correspondence between geographic and group boundaries, a voter would only have an opportunity to vote for regional representatives from his/her own group. Horowitz and Reilly espouse avoidance of this boundary coincidence to promote incentives for people to vote for candidates from outside of their groups. Between 1987 and 1995, in elections to all decision-making posts in the country, voters in Fiji were only permitted to vote for candidates of their own groups, two of which are MARs. Reilly observes that this situation, with electoral and group coincidence performing the same function that geographic and group coincidence would, “ensured that true interethnic political competition was virtually impossible.” Since this did not constitute geographic and group coincidence, Fiji might arguably be coded with negative values for this variable, for 1987 to 1995. However, positive or negative values must be chosen and, while it may not perfectly fit the positive definition, this situation certainly does not fit the negative one implying the dispersal of individuals’ power across group lines. Prior to 1987, all voters in Fiji voted for candidates to fill a substantial number of “national” seats so negative values, indicating an absence of boundary coincidence, were allocated to the years 1975-1986 of the country’s experience. Lijphart does not seem to examine Fiji at any length so his perspective on the group and power boundaries in Fiji cannot be compared to that of Reilly. Georgia: (Methodology involving minority veto data): There is no minority veto data for Georgia because its constitution was adopted in August, 1995. Ghana: (Methodology involving minority veto data): There is no minority veto data for Ghana because I was not able to obtain the country’s constitutional documents, at the time I was making this data. Israel: (Methodology involving minority veto data): Data involving minority veto for Israel was not made because its constitutional documents are not sufficiently comparable to those of most other countries. Malaysia: (Methodology involving segmental autonomy and, in particular, boundary coincidence): Examination of the Minorities at Risk dataset and the list of countries containing federalism and/or autonomy arrangements initially indicated that Malaysia did contain a coincidence of geographic constituency and group boundaries during the 1975-1995 period. However, the inaccuracy of this preliminary conclusion is emphasized by the agreement between Horowitz and Lijphart that the potentially antagonistic groups in Malaysia are more divided, than concentrated, between the country’s federal units. Horowitz identifies Malaysia as containing “heterogeneous states” and Lijphart confirms that the groups (identified as MARs in Gurr’s dataset) experience “mutual isolation” that is “only social and not territorial,” because the country’s fragmentation pervades the substates. These conditions appear to have existed throughout the 1975-1995 period so each of those years of Malaysia’s history was allocated a negative value for boundary coincidence. New Zealand: (Methodology involving minority veto data): Data involving minority veto for New Zealand was not made because its constitutional documents are not sufficiently comparable to those of most other countries. This issue is described in more detail in the note above relating to Bhutan (concerning the same footnote). Saudi Arabia: (Methodology involving minority veto data): Data involving minority veto for Israel was not made because its constitutional documents are not sufficiently comparable to those of most other countries. Sri Lanka: (Methodology involving segmental autonomy and, in particular, boundary coincidence): The information originally consulted to make data for Sri Lanka regarding this variable indicated that it should receive a negative value, indicating an absence of boundary coincidence. This value resulted from the fact that Sri Lanka is not a federal country and it does not contain autonomous units. However, Horowitz depicts its subnational governance units, which apparently do not have sufficient autonomy to justify its identification as federal, as “homogeneous” with regard to group composition. Lijphart does not discuss conditions in Sri Lanka so his perspective on the nature of these subcountry governance units cannot be consulted, to ascertain whether he agrees with Horowitz regarding the country’s demographic situation. Since Sri Lanka’s subnational governance units appear to have been homogeneous throughout the 1975-1995 period, the country receives positive values for each year contained in this analysis. Ukraine: (Methodology involving minority veto data): There is no minority veto data for Ukraine because its constitution was adopted in June, 1996. United Kingdom: (Methodology involving minority veto data): Data involving minority veto for the United Kingdom was not made because its constitutional documents are not sufficiently comparable to those of most other countries.