The Malt Whisky Yield Curve

This is a piece I’ve been thinking about for around a year and have now finally got round to writing up now that the Cardhu Scandal has made it arguably topical again. Basically it’s an idea for anyone who wants an easy way into thinking about capital theory. I’ve thought for a while that the booze industry ought to be used much more as an example for people thinking about time and production, because it allows you to abstract from considerations of technology and the production process; there are any number of ways to produce a chair, some more time-consuming that others, but there’s only one way to produce a cask of ten-year-old whisky[1]; start with a cask full of nine year old whisky and wait. The fact that time is intrinsically part of the production process for wine and brown spirits is why you see “capitalised interest” on the balance sheets of drinks companies; part of the economic cost of whisky production, and therefore part of the eventual sale price and the value of the goods, is the interest foregone during the process of maturation. It’s this interest element which I’m going to concentrate on.

I’ve picked whisky as my example rather than wine, because it’s a product which is more homogeneous over time than wine. Unlike vineyards, whisky producers don’t have vintage years, because whisky is a more industrial and less agricultural product; the growing conditions of the grain make less of a difference to the finished product than they do with grapes. Therefore, there aren’t so many considerations of whether a particular year was “good” or “bad” when it went into the barrel to deal with; the capitalised interest element is the most important reason why older whiskies cost more than younger ones. I couldn’t abstract entirely from these kinds of issues, as you’ll see below, but it was possible to do quite a lot even under the aggressive simplifying assumption that all whisky is homogeneous.

I managed to dig up two datasets; one from the Scotch Malt Whisky Society and one from the WhiskyWeb high end vintage malts catalogue. The SMWS dataset is larger, and it’s more concentrated toward the younger end of the yield curve. It’s also, I think, more homogeneous in the types of malts on offer; the SMWS, as far as I can tell, is selling for the most part whiskies from smaller distilleries rather than the megabrand single malts, and it seems unlikely to me that there will be a lot of variation in the rarity of the malts on offer. The SMWS is also a mutual society rather than a profit-making concern, so I have less to worry about in the area of strategic pricing behaviour. For this reason, I’ll mainly be concentrating on this dataset.

My first step was to go through the current SMWS catalogue on the web, noting the age and price of all the whisky they’re selling through their website (30 data points in all). The next thing to do is to adjust for tax, since UK excise duty is charged at £19.65 for every 1% of alcohol strength per 100 litres, rather than ad valorem [2] Plotting and eyeballing the data below, it’s clear that the relationship is there.

In order to get anywhere, though, we need an estimate of the non-time-related costs of whisky production (mainly excise duty, but also grain, water, fuel and overhead). There’s a number of approaches one could take to this, but I decided on the simplest one I could think of that wasn’t absolutely obviously wrong. I took logs of the prices (because we’re dealing with compound growth rates here) and regressed log price on age, as below:

The regression equation suggested that the intercept of the simple log growth equation was 2.9877, and the antilog of 2.9877 is 19.84, suggesting a notional price of about twenty quid for raw single malt spirit straight out of the distillery. That strikes me as not so bizarrely out of line with the price of a bottle of vodka (an unaged grain spirit) as to be unusable, so I’m sticking with it. As shown lower down, this is quite an important number, though; I’d be grateful for any opinions on what a correct number might be.

Once you’ve got the price of a bottle of whisky of zero age (Pedants may at this point say that it can’t be called whisky if it’s less than three years old), it’s easy to compute the implied rate of return of a bottle of X year-old whisky selling for Y. Here’s the scatterplot, with a fairly boneheaded scatterplot smoother thrown in using Excel’s “Moving Average Trendline” function.

I wouldn’t read too much into the downward slope of the yield curve on this chart; it’s just kicked around a bit by the fact that there’s only a few data points at the short end. I would take away from this scatterplot that the realised rate of return on malt whisky in the cask over the last forty years has been surprisingly constant at four percent, give or take fifty basis points.

One might have expected a positive slope in this yield curve, if one thinks that whisky in the cask gets better at an ever-increasing rate (possibly a whisky snob’s a priori belief), or if one were to think that, as the Cardhu case proves, the supply of properly aged whisky is incredibly inelastic, older whiskies would have a rarity value. But I don’t think you can support this with the data. I played around with my spreadsheet until I got some signs of visible upward slope (basically, at a raw spirit cost of around £24 or above), and you get this chart:

I don’t think that the implied rates of return on the younger malts are credible; they’re well below the rate of return on money in the bank, so why would people be selling? (Also, my guess is that the implied forward rate of appreciation of an 8-year old bottle as it turns into a 10-year old bottle would be implausible). On this criterion, my log-linear estimate of £19.84 looks quite good; the lowest data point is 3.4%, so almost everyone’s earning more from whisky in the cask than the money rate of interest. My guess is that, within the SMWS dataset, the older spirits on offer are from less prestigious distilleries, so any time-varying rarity premium is being offset by a negative distillery-specific premium. Since I couldn’t conveniently get a list of the distilleries from the SMWS website (it could be done, but it seemed more trouble than it was worth [3]), I didn’t check up on this.

Modelling the effects of rarity premia is probably best done with the WhiskyWeb dataset. This is a short but expensive catalogue of older, more prestigious malts. There are fewer datapoints (25), and they range from 16 years to 66 years[4] old, so I felt uncomfortable putting them on a scatterplot. I’ve tabulated them below, sorted by distillery and then by age. Note that I’ve not tax-adjusted the prices because I’m not at all clear on the excise duty position of WhiskyWeb; I crudely altered the implied raw spirit price to sort of compensate. This probably accounts for some of the very low (c2.5%) implied rates of return on some distilleries.

What one gets out of this is first, that the estimate of four percent still looks pretty OK, even going out to some really rather old whiskies. Second, there are clearly distillery-varying premia between distilleries (Macallans seem to do much better, for example), and the very very old whiskies do appear to have an age premium; the assumed upward slope is there if you go out far enough. I’m mildly surprised about the distillery-specific premia in the growth rates (I’d have assumed that the premia would be there simply in the young whiskies, and that growth thereafter would be simply determined by the time value of money). Maybe the best malts age better, or maybe there is more demand for them so they get rarer as they get older.

So there you go. The internal rate of return on Scotch malt whisky in the cask since the war has been four per cent for whisky to be sold today. Mine’s a Chivas.

(For classroom discussion; How does this “malt whisky yield curve” relate to the more normal bond market curve? What rate of return might you expect today on spirit being distilled today?)

[1] Go on, spell it “whiskey” in the comments. I dare you. I just dare you.
[2] I assumed that the prices quoted were for a standard 75cl bottle at cask strength (50%), giving excise duty per bottle of £7.37.
[3] Perhaps a curious decision given the time and trouble gone into this piece overall …
[4] WhiskyWeb gives the “vintage” (year of distillation) rather than the age for most of its whiskies; I hand calculated the age and this might be a source of errors, given that some of the whiskies where they do give an age, it doesn’t seem to match up too well with the stated year of distillation.