Mexico data at 8 digit level
Closed this issue · 5 comments
Similar to the Romalis tariff data notebook, get 8 digit tariff level data and aggregate up to our classification.
WITS doesn't have 8-digit, at least not from the World Bank (have to check TRAINS).
Some preliminary plots can be found at the very end of the notebook (scroll to the bottom)
To do:
- What is the outlier?
- Why does it matter more in the weighted average apparently?
- Compare this with data from TRAINS
The outlier seems to be industry 27, base metals. However, looking at single product categories, the highest values are 20900, 150100, 20712, 20726, 20735, 20732, 20733, all of which seem to be related to the food industry.
As for the TRAINS comparison, it appears that our original data was from TRAINS. It is much noisier but goes back to 1989. Need to compare AHS tariffs in WTO IDB with those in TRAINS. What was the original idea behind using TRAINS? The WTO data seems better.
EDIT: Scratch that, I'm colour blind and the outlier is indeed industry 15, food, beverages and tobacco.
Summing up:
For the US, it seems clear that the best way to go is to directly use USITC data, which for 1989-2001 is available as AVE from John Romalis, and 2002-2014 is available from USITC itself, without the AVE adjustment. It might be okay to merge the two, which would give us a more or less consistent series 1989-2014
For Canada and Mexico, it seems that WTO IDB dominates TRAINS in terms of data quality, but TRAINS gives a few more observations (1989, 1993, 1995 for Canada, 1991 and 1995 for Mexico). Maybe we can just add those to the WTO database to extend the sample period. The WTO database also has Mexican tariffs dropping to zero for one year in 2001 and 2007, but the levels in 00/02 and 06/08 are so consistent that it appears linear interpolation could be our friend here.
Here's what I'd propose, based on the picture below:
For Canada, use TRAINS PRF for 89-95, then WTO AHS for 96-14
For Mexico, use TRAINS AHS from 89-94, then TRAINS PRF for 95-09
For the US, use TRAINS PRF from 89-96, then WTO AHS for 97-14
This gives us more or less consistent tariff series for all countries from 1989 to 2009, with the only problem being that we don't actually have data for the crucial years 93/94/95 in some cases, and linear interpolation might be dangerous if NAFTA actually let tariffs fall off a cliff (which I think it didn't, but still a linear trend probably wouldn't capture what was going on)
