Study title

United Kingdom, Sweden, Netherlands, Switzerland

Organization

MethodologySample:As stated in the initial project proposal, the aim of the project was initially to follow Höpner (2003) in the sampling approach, i.e. to constitute a repeated cross-sectional sample of the 40 largest listed companies by market capitalisation every year. However, in order to be able to use sophisticated longitudinal econometric techniques and, because of uncertainties around the availability of archival data for each one of the firms in the sample, the sample size was increased substantially.We first constituted a list of the largest 150 companies by market capitalisation at year end based on Thomsone One Banker. We then compared this list of companies to the lists of listed companies available in various stock exchange yearbooks for each country (see list of stock exchange yearbooks in SOI_sources.xlsx). We added companies that fell within the initial list by size where they had been omitted. This allowed us to avoid the survivor bias that electronic sources alone would have introduced as they often only contain active companies, or at least the information for defunct companies is incomplete.A small number of companies, which were in the sample in all but 1 or 2 years, but dropped out of the top150, were carried over from one year to another to increase the number of firms that constitute a ‘balanced sample’.The ultimate sample size varies greatly from one country to the other and from one year to the other. This is due to the fact that the archival resources used would not contain information for all companies included in the first steps. Most notably, while the list of UK companies contained a large number of companies incorporated overseas, the stock exchange yearbooks generally did not contain information for these companies. It was therefore decided that foreign subsidiaries had to be dropped from the sample (see ‘nationality criteria’ below for further details). This lead to somewhat smaller sample size for the UK in the earlier years, because the UK sample contained a large number of foreign subsidiaries.The final sample constitutes an unbalanced panel of 1038 unique firms; with 9704 company/year observations over the period 1990-2010 in total.Nationality criteria:The approach adopted in this project was one of ‘nationality’ not listing alone. That is to say we only included companies that were both listed and incorporated in the country of interest, eliminating thus foreign subsidiaries. This choice was made partly due to the research question at hand, partly also, because data availability for foreign subsidiaries was very poor. Two criteria determine hence the nationality of a company: The stock exchange listing (a company had to be listed in the country in question) and country of incorporation. We thus eliminated any companies listed in our countries of interest, but incorporated elsewhere.A small number of exceptions to this rule exist: companies that were taken over or merged with a foreign company, but were from the country of interest at the time of the data point in question were kept in the sample. This was for instance the case of CGE A.B. in Sweden which was acquired by Alcatel-Lucent. Similarly, companies that moved their country of incorporation without being taken over, were kept as well if they remained listed on the stock exchange (e.g. Synthes Inc. in Switzerland).Missing values:The dataset contains a large number of missing values, which are coded as NA. This is due to the comparative and historical nature of the sample and the reliance on archival data. Indeed, for some countries information on certain variables was not reported in the stock exchange yearbooks, which implies that for certain countries and years certain variables are completely ‘empty’. These variables were still maintained in the dataset, as analysis of a cross-section or of the remaining countries may still be possible based on the available data.Another reason for the large number of missing values has to do with the explicit aim of the project to develop a measure of corporate governance practices that is sensitive to ‘functional equivalent’ corporate governance mechanisms across countries. As an example, the four countries vary very considerably concerning the main mechanism of voting-right distortions. Thus, while Switzerland and Sweden extensively use dual class shares, this instrument is formally prohibited in the Netherlands. However, that does not mean that Dutch companies do not ‘distort’ voting rights: Dutch companies use a form of ‘priority shares’ instead, which have de facto the same effect, but are legally a different type of securities. Similarly, multiple voting shares, while not legally prohibited, are used very sporadically in the Netherlands and are hence not reported in the archival documents that we used for the data collection. We coded such cases in the following way: while legally prohibited instruments are ‘structural zeros’ and where hence coded as such (value of 0), legally allowed, but very uncommon instruments that are not reported were coded as NA, although they can be assumed to be 0 in actual fact. We opted for coding ‘structural zeros’ as such, because this project attempted to investigate the extent to which a firm adopts shareholder-orientated practices. The presence or absence of a given instrument is hence relevant information independently of whether the absence was legally imposed (structural 0), or not. Finally, there are some variables, which are only applicable to one country (structure regime dummies for Dutch companies, separate EPS variables for various share categories in Switzerland).Date and Currency:Data from electronic sources are for the year end. Similarly, data from the archival sources are generally for year end, with some exceptions where the figures may be for the end of the financial year.Unless otherwise stated, we have converted local currency figures into US dollars at current prices. The reporting currency of the archival resources are in the last column of the data file. We used annual mid-point averages from oanda.com to convert local currencies into US Dollars where required. The sheet ‘CurrencyFX’ contains the relevant exchange rates.