Yesterday I noticed my analysis on who pays tariffs in US imports from China is totally incompatible with the import price movement in total.
Today I added correction to my analysis. I guess my mistake came from unit revision by Census bureau at the start of the year, which generated a lot of zero price indices.
This is my second major correction. My first one was about [China retaliation values] (https://mitsuoxv.rbind.io/2019/06/21/correction-china-hits-back-as-much-as-the-size-of-its-claim/), and I learned that Harmonized Tariff Schedule is not harmonized at some level.
This time I have learned unit in US trade data may be different over years. In added correction, I just removed zero values and zero price indices. This is tentative correction. I may have to find out how to keep unit consistency.