# How can the "Share of merchandise imports at current PPPs" be negative?

I'm looking at the Penn World Table 9.0 data, and find that the "Share of merchandise imports at current PPPs" (csh_m) can be negative. For example,

   country  year  csh_m
<chr>   <dbl>  <dbl>
1 Aruba   1970. -0.485
2 Aruba   1971. -0.459
3 Aruba   1972. -0.436
4 Aruba   1973. -0.441


My initial thought is that the share of import (csh_m) here actually means "net import as share of GDP." However, this is not correct, because share of import (csh_m) and share of export (csh_x) don't have the same magnitude.

   country  year  csh_m csh_x
<chr>   <dbl>  <dbl> <dbl>
1 Aruba   1970. -0.485 0.427
2 Aruba   1971. -0.459 0.435
3 Aruba   1972. -0.436 0.441
4 Aruba   1973. -0.441 0.477

1. Where can I find documentation explaining the construction of csh_m and csh_x from the Penn World Table?

2. If a want to construct a variable of trade openness, should I add the absolute values of csh_m and csh_x?

• Have you looked at p3 of rug.nl/ggdc/docs/national_accounts_in_pwt80.pdf ? If it's only very small countries, could it be that statistical discrepancies are larger than imports?
– Dan
Mar 15, 2018 at 17:39
• All the csh_m (Share of merchandise imports at current PPPs) numbers in that table for all years and all countries are negative (apart from some years for Bermuda) while all csh_x (Share of merchandise exports at current PPPs) are positive (again except for a few years for Bermuda). These shares look like $Y=C+I+G+X-M+R$ all divided by $Y$. Places with big absolute values seem to include Hong Kong, Singapore, BVI, Malta, Turks & Caicos, where trade is indeed open. So you could add the absolute numbers together Aug 5, 2019 at 22:28
• See rug.nl/ggdc/productivity/pwt and the legend tab of the Excel spreadsheet for more explanation about the data Aug 5, 2019 at 22:30