1

I'm having some trouble at replicating this paper (text here: http://www.michiganjb.org/issues/52/text52c.pdf) The paper wants to assess the economic impact of the Olympic games in the long run . The author has a big dataset cointaing data on GDPpc for host and runner-up countries from 10 years before to 10 years after the Games. Data are from page 88 for host countries and from page 89 for runner-up countries. Regressions and variables are explained in page 73. I'm just interested in regression (1) and (2a). The problem is that I can't get the author strategy. In the paper, it looks like the author is using a simple OLS regression, but since he collected data on different countries across 21 years shouldn't he using a panel data regression? If I run an OLS i get similar (but not identical result), while with FE or RE I get totally different results. Which technique would you use in such a case?

Paper: Mehrotra A. To host or not to host? A comparison study on the long-run impact of the Olympic Games. Michigan Journal of Business. 2012;4 (2) :61-92.

Luca
  • 11
  • 2
  • 3
    If this is just about Stata programing to replicate these regressions then this might be the wrong place to ask. We can help you more with conceptual questions regarding the techniques and statistics. – Andy Apr 28 '16 at 09:40
  • I'm sorry if I wasn't clear. The problem is that I can't get the author strategy. In the paper, it looks like the author is using a simple OLS regression, but since he collected data on different countries across 21 years shouldn't he using a panel data regression? If I run an OLS i get similar (but not identical result), while with FE or RE i get totally different results. Which technique would you use in such a case? – Luca Apr 28 '16 at 10:11
  • 1
    Please edit the information in the comment into the question. Do you get the same results using OLS? Also, if the paper has been published, please add a full reference (title, journal, page numbers, name of author etc.) in addition to the link. – Juho Kokkala Apr 28 '16 at 10:14
  • I edited the question. My results are different in both case but with OLS the magnitues of the coefficient are similar , while with FE even the signs are different – Luca Apr 28 '16 at 10:56
  • 3
    Cross-posted here http://www.statalist.org/forums/forum/general-stata-discussion/general/1337767-problem-with-replicating-a-paper It's polite in any forum to tell people about posts elsewhere. – Nick Cox Apr 28 '16 at 11:12
  • When you say that "you run OLS", you mean the you pooled the panel data sample as though it were a cross-sectional sample and then run "pooled OLS" estimation? – Alecos Papadopoulos Apr 28 '16 at 11:38
  • This seems to be two questions merged in one post - I think it would be better to ask separate questions about the replication issue (getting the same numbers using the method the author reported to have used) and about the methodology issue (what kind of regression model should be used here). – Juho Kokkala Apr 28 '16 at 12:13
  • Why is this tagged stata when Stata is not mentioned in the question at all? – Juho Kokkala Apr 28 '16 at 12:13

0 Answers0