a random night forecasting an edible oil trader's stock price
hi again
A couple of days before 2026 started, I was getting really, really bored at home. Not looking at stocks was biting away at me and the ignored watchlist glared accusingly every day.
I didn't start anything from that watchlist though. I went on a mad pursuit of a forecastable stock at 10pm. I ended up with gokul agro resources, an edible oil trading company. I didn't see any annual reports, no concalls, and got down to the financial statements with a small brief that screener gives about the business model to everyone.
this is why I think it's an ideal candidate for a relatively simple forecast.
- a homogeneous product with a close to perfectly competitive market; not a lot of product differentiation or branding.
- majority of the market is domestic, which makes it convenient because i don't need to split up markets and forecast independently.
- low margins that makes forecasts dependent on sales; primarily, sales volume because sales can be deflated easily with a price proxy
What followed was a set of excel sheets over the course of three hours that ended up forecasting world edible oil prices by using the FAO oil index as a proxy for edible oil prices. I noticed sales are cyclical with upward shocks. A rolling average would not work for obvious reasons. I deflated sales with the FAO index and decomposed the Index and the Deflated Sales into structural and cyclical parts. the structural part of both assumes an average straight line growth rate while the cyclical part connects peaks and troughs. The cyclical part is where I use my discretion and try to recreate the cycle.
I'm still a broke college student so you can see the excel here. The price is off because there are multiple flaws with the model that I plan on updating later:
- The beta is wildly high. I need to take more price data into account or use another way to measure beta.
- Capex and NWC should be linked to concall assumptions and not be assumed constant. Capex increases to sales are incorporated through structural growth in the deflated sales, although I think it can be modelled differently perhaps
- I use a rolling average to forecast margins and that smoothens out margin pretty quickly. There's a pattern I found. Margins usually increase with a less than expected increase in sales and decrease with an upward shock in sales. Modelling that would probably change forecasts.
The main problem is the beta though. It doesn't help that it's a very low-debt company so WACC is mostly cost of equity. suggestions to help model this better are welcome.
help a fellow forecaster out. thanks!