5.1 Data
5.1.1 Sample identification for the cross-lister sample
The sample identification procedure, which I describe here, follows quite closely the guidelines laid down by Tolmunen (2001) and Tolmunen and Torstila (2005). I define Europe in this thesis as the developed Europe, i.e. as the original EU-15 countries and Norway and Switzerland. The reason for this definition is that I want to exclude European companies, whose country of origin is considered as having an underdeveloped equity market, which might disrupt my analysis. Also cross-listings by companies from Europe‟s less developed countries are extremely rare. Extant literature (see Pagano et al., 2002; Tolmunen and Torstila, 2005) define cross-listing as a listing that takes place before or simultaneously to a home market listing. I follow this same approach in identifying my sample. Tolmunen and Torstila also introduce a view, according to which companies who have cross-listed their shares prior to year 1980 are truly global in nature and cross-listing is likely to have very little effect on their acquisition behavior. I build on this idea and require that the cross-listing in my sample has to have taken place between 1.1.1980 and 31.12.2008.
I start the identification of the cross-lister sample from the web pages of the major U.S. stock exchanges, i.e. New York Stock Exchange (NYSE), the former American Stock Exchange nowadays called NYSE Alternext U.S. (AMEX) and National Association of Securities Dealers Automated Quotations (NASDAQ). I search for companies domiciled in Europe, collecting the name of the company, time of the listing as well as the type of listing. This yields me a total of 160 European companies currently listed in the U.S..
As delistings from the U.S. have become increasingly common due to a number of reasons, the most important being the relaxation of SEC deregistering requirements, I also need to add companies that have been, but are not anymore listed in the U.S. to overcome survivorship bias. I do this by using the Thomson Financial SDC New Issues (SDC) database. I search the database for U.S. equity listings by European companies between 1.1.1980 and 31.12.2008. This yields me a total of 265 potential cross-listers, of which 105 turn out to be duplicates of the companies, which I already identified through the exchanges.
Combining my samples from the two sources and eliminating doubles, I have 320 observations. Prior literature on cross-delistings (see, e.g., Doidge et al., 2008) has laid down a practice to use a number of different databases to make sure to be able to include all cross- delistings. To follow this, I further check for cross-listings from the Citigroup‟s ADR service. Employing the same date range as previously, this yields me 435 observations, of which 86 are new observations that were not found from the other two sources.
All and all, I identify a sample of 406 potential cross-listers. However, it still needs to be verified, whether these are actual cross-listings, or direct orphan listings to the U.S.. I use various sources to confirm this: 1) If the issue synopsis in SDC mentions a simultaneous offering in Europe, the issue is deemed as a cross-listing. 2) Also if, according to Thomson Financial Datastream (Datastream), there is a stock market quote in Europe for the potential cross-lister prior to the listing in the U.S., the company is categorized as a cross-lister. 3) Finally, I use the web page of the company in question, or alternatively LexisNexis key word search, to verify the existence of a prior listing in Europe. By employing this method, I am able to confirm most of my sample companies as cross-listers due to the fact that direct listings in the U.S. by other than North American firms are extremely rare.
For the observations that I have identified solely from SDC, I also have to verify the date of cross-delisting. I again use web pages of the companies as well as LexisNexis for this task. While I add the delisting date to these observations, I also simultaneously cross-check the listing dates found in SDC to verify the correctness of my data. It also turns out that in some cases the date of listing differs between the exchanges, SDC and / or Citigroup‟s ADR service. If the difference is minor, i.e. less than five days, I use first the date from the exchanges, then SDC, and finally that of Citigroups ADR service, if data from the two other sources is not available. However, if the difference is more than five days, I further study
Table 3: Sample sources and filtering of the cross-lister sample
This table shows the derivation of my sample of European companies that have been cross-listed in the U.S. between 1.1.1980 and 31.12.2008. Europe is defined as the original EU-15 countries and Norway and Switzerland. N is number of firms in each stage. I show the initial sample sources and document the process of removing some of the observations to arrive at my final cross-lister sample. Data on potential cross-listings is first gathered from three partially overlapping sources, i.e. directly from the exchanges, Thomson Financial SDC New Issues database (SDC), and Citigroup‟s ADR service. Some observations have to be eliminated due to 1) being a duplicate among the three samples, 2) being a direct listing to U.S. instead of a cross-listing, 3) being a financial company (four digit SIC code beginning with 6) due to differences in their financial reporting, 4) ambiguous listing date, and 5) no data available in the Thomson Financial Worldscope database.
Source / rationale for removing N
Directly from the exchanges 160
SDC 265 Citigroup‟s ADR 435 Duplicates -454 Direct listers -49 Total cross-listers 357 Financial companies -52
Ambiguous listing date -2
Data unavailable -22
Total sample 281
the issue from the web site of the company in question as well as by using LexisNexis. I have to exclude two companies due to being unable to confirm the listing date. Once I employ my final restrictions, i.e. the removal of financial companies due to differences in their financial reporting, and the removal of listings of companies which had no data available in the Thomson Financial Worldscope (Worldscope) database, I arrive with a final cross-lister sample of 281 companies (see Table 3 for summary of the sample identification process).
In this thesis, listing is defined as a quotation on a major U.S. stock exchange, meaning that Level I ADRs as well as SEC Rule 144A private placements are not qualified in the sample, unless they eventually lead to a full quotation. I obtain information on private placements and Level 1 ADR programs from the Bank of New York web page. This allows me to define the original IPO date as the OTC listing date, as identified also in SDC and by the exchanges, for a number of companies. The same definition is used by Tolmunen and Torstila (2005).