Here is the paper by Auvray&Rabinovich.
You can also look at Joel’s presentation of sept-19. Additionnal paper on the (false) financialization of non financial firms.
Lecture on quantification and Desrosieres’ paper
Oct-7
- Install R and Rstudio. You can use this tutorial and this textbook.
- Import the reduced database.
- Install the Tidyverse package. My slides.
- My R script.
- Download the complete database.
How to work on the long run
- Write a journal, a log (where the data comes from, what you do to it, issues you face and options you take, ideas, paste your plots)
- save and comment your scripts
- start writing your report from the first day
- make a copy of the whole folder on another computer or on the cloud, each time you work on it.
Oct 14
- Tristan Auvray presentation
- My panel models slides
- Our workshop pad
- My script
Oct 21
- My slides on GMM
- Try GMM on the toy database
- Import the complete database
- Calendar :
- Oct 28 no class
- Nov 4, class on sequence analysis in the morning (10:50-12:30) and from 14:00 to 16:30, for individual advise on the replication. Register here for an appointment, max 5 students per 30 minutes session.
- Nov 11 no class
- Nov 18 last class, at UTC at Boulevard Sebastopol. 14:00 – 17:00. Individual meeting again, register here
- Dec 5 Report due date
Advice for implementation:
- the variables are already trimmed, lagged, normalized… in the complete database ready4.
- do not use the size dummies (du_size_big), they are for another table, and do not correspond to table 6 categorisation. You have to split yourself. You should try not to split differently for different years, otherwise it will produce missing information in the panel.
- do not use “summary” of your output, it takes a lot of time, use stargazer instead
- if the computation takes to much time, try with less lags in the instruments, with less companies, with less covariables…
Additionnal advice (Nov 18)
- You might find a negative coefficient for investment_l. THere is no need to “fix” it. Comment on why it’s strange.
- If computing is too long, you can use “collapse=T” as an option of pgmm, to reduce the number of instruments and computation time.
Directives for the report:
- The report is individual. It’s length is approximately 6 pages. It should contain
- an abstract of Auvray and Rabinovich’s paper
- a discussion of a variable of the model, with a sociology of quantification perspective. Here is a very very complete example.
- the replication of table 6, p.24, at least column 5 (model with offshoring variables for large companies). Compare your results to those of the original paper and comment. You might want to use the stargazer package to produce a nice table.
- Send me your report at harari at univ-orleans.fr with [econometrics] in the object of the email by December 5.