top of page

Lake County Cabernet Sauvignon Price for 2016

So, as I have mentioned before, my worst forecast ever was a massive failure to predict the 16% jump in price for Lake County Cabernet Sauvignon between 2013 and 2014. I'm still working on stronger and stronger models for Lake County - and, for that matter, everywhere else - and have a new one to try out. Now, I know, this information isn't so useful for you now. After all, we're about to harvest and anyone on the frontlines has a pretty good idea of where prices are at - certainly better than a purely quantitative model.

Still, it could be useful for some indexed contracts and, for me, it is important to take my models for a test drive in the real world. With no further ado, here is my forecast for 2016 prices, with technical details and instructions on reading this further down (sorry, but you may have to zoom in to read it easily):

The "CPI Price" is the prices, assuming 2.2% inflation between 2015 and 2016, whereas the other number is the price in constant dollars. The CPI price is the actual forecast, but the other number allows me to measure how much any error is cause by different inflation assumptions and how much is weakness in the model's ability to anticipate real price changes.

The numbers for Low 60% and High 60% indicate the boundaries for the region within I predict a 60% of prices falling. That is, there is a 60% chance prices will be between $2.125.04 and $2,350.21. That implies a 20% chance they will be below $2.125.04 and a 20% they will be above $2,350.21. Of course, the other confidence intervals work in an identical manner.

The Multiple R correlation for this forecast is .9708, yielding an R-squared of .9425, a very high level of correlation. The p-value for the intercept is worryingly high, at .5964. The other two variables are one that measures the trend in price, while the other variable is a measure of where we are in the grape market cycle. Both are negligibly different from zero, which is great.

The intercept p-value worries me, though. We'll see how the model performs. I do have a Model B that should perform much more accurately, but less precisely than this one. It would, for instance, yield these numbers:

Low 95% High 95% Low 80% High 80%

$1,827.11 $2,648.14 $1,973.62 $2,501.63

Both models predict nearly identical prices, but clearly, the main model is more desirable, due to the narrower confidence intervals. Model A is a re-work of the model that failed me a couple years ago. I believe it to be a strong model. But, if it fails again, I'll have to rely on Model B.

Anyways, here's to a great harvest! Cheers!

Follow Us
  • Twitter Basic Square
bottom of page