Decision tree

View or edit on GitHub

This page is synchronized from trase/models/brazil/cocoa/Decision-tree.md. Last modified on 2025-12-14 23:19 CET by Trase Admin. Please view or edit the original file there; changes should be reflected here after a midnight build (CET time), or manually triggering it with a GitHub action (link).

Thoughts after quick exchange between Vivian and Javier, following some few data digging by Javier. I am including some new suggestions with the goal of discussing them here.

1) The ICMS data (a tax associated to movement of products between states) could provide us with the total amount of cocoa fleeing Para to Bahia and other potential destinations like SP. Vivian will make an e-petition to get this data. Ideally we would get it for key municipalities, in case the cocoa destination would vary. Altamira is the key municipality since it collects the cocoa from the larges cocoa producers in the Transmazon highway.

2) In any case it is important to understand that domestic direct demand for Para cocoa is very very small (Cacauway mostly, and it is a tiny company). This means that the beans go almost 100% to Bahia (and maybe SP, see point 1). HOWEVER, this does not mean they are all exported. In Bahia they have the vast majority of the industrial capacity. This industrial capacity does not segregate origins, and the majority of the exports are low quality products from 1803-1804. Thus for codes 1802,1803,1804 etc... we can safely assume that the proportion of Pará cocoa in exports associated to Ilheus (and for that matter in domestic consumption) is just dependent on the proportion of production Pará vs Bahia, for those companies with activities in both. So if Pará is producing 60% of the cocoa arriving in Bahia, and Bahia the other 40%, then if a company exports 1000 kg of 1802 then 60% of that volume will come from PA. This is just in the case that the company has CNPJs in both Bahia and Pará (which is the case for the major producers anyway). If it has CNP just in Bahia then there is no issue, origin in Bahia. This simple assumption will be very useful to determine the majority of the volumes (1804: Cocoa; butter, fat and oil; 1805: Cocoa; powder, not containing added sugar or other sweetening matter)

3) For 1801 I suggest that we saturate the Pará supply so it makes it into exports, given its better quality. In other words cocoa bean exports from Ilheus would have their origin in Pará if the exporter has operations in Pará. When the Pará supply is met then it would be filled in with Bahia produce. The code 1801 is pretty small. For this to work this branch needs to come in before 1802-1805.

4) Beyond the above LP constrains, to allocate cocoa production to LH we need to "play" with multiple CNPJ in the 2 states for some major exporters (for 1802-1805, industrial exports). For example, if Cargill would have 2 CNPJs in Para (A with municipal production 100 and B with municipal production 300) and 1 in Bahia (C with municipal production 500), and the ratio PA/BA would be 1.4, then if the exporter is exporting 1000 the sourcing would be:

From C: 1000/(1.4+1) = 416.6

From Pará: 583.3

From A: 583.3 x 100/400 = 145.8 From B: 583.3 x 300/400= 437.5

Once we have generated this kind of simple level 1 approach, we can improve these assumptions and potentially add new available data by engaging with experts on the ground.