Country-pair-year observations

[Rik Velthoven]

Dear Statalist,

I have a question about country pair-year-observations. I am doing a research about cross-border mergers. I do a follow-up on another paper. However I am confused about some statements they make regarding country-pair-year observations. In the paper they aggregate the total volume from country a to country b in a specific year. Volume is in this example the transaction volume, which is recorded by Thompson one, that is paid by a firm from country a to acquirer a firm in country b. I will give a short example how this will look like

Country-A Country-B Year Transaction-Value
Germany France 2000 1 mil
Germany France 2000 1 mil
Germany France 2005 5 mil
Belgium U.S.A. 2006 3 mil
Belgium U.S.A. 2006 8 mil

In this case the total aggregate volume from the country-pair Germany to France is 2 mil in 2000 and 5 mil in 2005. The total aggregate volume for the country-pair Belgium-U.S.A. is 11 mil in year 2006. In stata I would run the following command.
bys (Country-A Country-B Year): egen Aggregate Volume=sum(Transaction-Value).

So far so good. However in the paper I am doing a follow-up on they have a total of 20.083 cross-border mergers in their sample. They state that they aggregate these 20.083 cross-border mergers into 27.000 country-pair-year observations. I double checked the paper multiple times, but don't seem to get my head around it. Maybe it is because my English isn't advanced enough to understand, but I don't understand how they can get more country-pair-year observations than there are cross-border mergers in the first place. I hope someone could clear things up for me and help me understand. I thank you in advance for your time.

Kind Regards,
Rik

[Clyde Schechter]

I don't understand how they can get more country-pair-year observations than there are cross-border mergers in the first place.

I don't understand it either. Either you are misinterpreting what they say, or there is a mistake in the paper. Perhaps 27,000 is a typo for 17,000 or something like that. Or, is it possible that the same merger might be counted in more than one country-pair year observation, say, in both Belgium-USA-2006 and USA-Belgium-2006. Or is it possible that some mergers span more than a single year in time and are counted, say, in both Belgium-USA-2005 and Belgium-USA-2006?

If there is no way that a single merger can be allocated to more than one country-pair-year, then there must be a mistake in the paper, or some other misunderstanding on your part. Providing a complete reference, or better, a link, to the paper might enable somebody to see if there is something in the paper you have missed.

[Rik Velthoven]

Dear Schechter,

Thank you for your reply I have thought about the double counting of the country-pair year like Belgium-USA and USA-Belgium, but this seems given the nature of the research unlikely to me.
The link of the paper is as follows:

https://papers.ssrn.com/sol3/papers....act_id=1545514

In the paper the following is mentioned:
"Due to data restrictions on our other variables (described below),our original sample of 127,950 is reduced to 104,652 mergers with 20,893 cross-border mergers and 83,759 domestic mergers across 52 different countries.These mergers are aggregated into 27,753 country-pair-years, including domestic mergers,that form the sample we use to test the role ofculture on the volume of cross-border merger activity. The domestic mergers serve as a benchmark for merger volume where no national cultural differences exist."

"Where later they mention the following: There are 27,086 cross-border country-pair-year observations and 27,753 cross-border and domestic country-pair-year observations in Panels A and B".

It good be a typo but that would also sound onlogic to me because the paper is updated multiple times and published so I would assume they would sort the typo out. If anyone could solve this riddle it would be great. I thank you in advance.

Kind regards,
Rik

[Clyde Schechter]

So, with the paper available the mystery deepens. According to Table 1, there are 29 different countries involved. That means there are at most 29² = 841 distinct country pairs (including USA-USA type pairs). Not all of those pairs actually occur, but 841 is clearly an upper bound. The years range between 1985 and 2008 inclusive, so the maximum possible number of country-pair-year combinations is 841*(2008-1985+1) = 20,184! The 27,753 is impossible unless I have seriously misunderstood what is written in the paper's methods section.

Just because a paper has been revised several times, and reviewed, and published, does not mean that all typos or other errors have been caught.

My advice is that you contact one of the authors of the paper and inquire about this.

[Rik Velthoven]

Dear Schechter,

In table 1 not all the countries are included in the matrix only the top 30-nations, so in the entire sample there are more nations which are involved with mergers. I will try to reach out to the authors of the paper and see if they will respond. Furthermore I thank you for the time and your input in helping me understand this riddle.