Monday, February 3, 2020

Are you ready for the cyclical downturn in the US wine market?

There has been much discussion recently about the effects of import tariffs in depressing sales of wine in the USA, as well as discussions about a current glut of domestic wine grapes (eg. The perilous state of the US wine industry?). However, to put these data in context we need an “expectation” for what US wine sales might look like in the short-term future, independent of such influences.

By this, I mean that discussions of the current situation in any wine market need to be based on more than just reports of the current situation compared to, say, last year. We need some insight into what has happened in the past and how it might relate to the near future. What we have at the moment is apparently contradictory reports (and opinions) about the current status of the US wine market. We need some sort of formal “expectation” of where we are at, in order to put things into context.

Such a thing can be produced using mathematical forecasts, in the same manner as a weather bureau produces forecasts for tomorrow’s weather. We take our knowledge of the current situation, and formally add to it our knowledge of what has happened in the past under similar circumstances. This often works remarkably well, until something comes along to disturb the current situation. For example, James Lawrence recently re-visited his wine-industry forecasts from last year, and discovered a number of disrupting factors that nullified those forecasts (The folly of wine prophecy).

The US wine market

As far as US wine sales are concerned, the bw166 group has recently released some long-term data for the U.S. Beverage Alcohol Market, based on information from The Alcohol and Tobacco Tax and Trade Bureau. Their summary of the US Beer Wine & Spirits January 2020 provides a graph of the number of wine cases entering commerce from 1960 to 2019, as shown here.

bw166 graph of recent time trends in the size of the US wine market

The data are interpreted by the author(s) as showing that “wine experienced strong growth for 25 years but that growth has slowed.” It seems to me that we can turn this into a quantitative forecast for the next few years.

We need to ignore the dotted polynomial provided by bw166 in their graph, as it is of no use for forecasting. When I look at the graph, I see two super-imposed patterns: a long-term linear trend (market growth) and a repeated cycle (ups and downs in the size of the market). So, to make the mathematical forecast, we first need to separate these two patterns. The next graph shows the linear trend as a straight line.

Linear time trend in the size of the US wine market

If we now subtract this straight line from the data, we are left with a clear view of the cyclical pattern, as shown in the next graph.

Cyclical time trend in the size of the US wine market

It is this cyclical pattern of ups and downs in the US wine market that is of most interest for our forecast. It shows (at the left) a dip in the market until the mid-1960s, followed by an uptrend until the mid-1980s, then a sharp downturn until the mid 1990s, and followed by another uptrend until the mid-2010s. The regularity of these cycles provides a clear forecast for the near future.

The current growth of the US wine market has covered the past c. 20 years, as it also did in the previous cycle (mid-60s to mid-80s). That previous growth was followed by a c. 10-year decline (mid-80s to mid-90s). My forecast as to what comes next should now be obvious — a repeat of the previous cycle.

That is, what looks like a current “market slow-down” in the original bw166 data (graph 1) looks very much like the beginning of a decline, to me. This is actually what we would (sadly) expect to happen about now, based on the previous cycle (graph 3). The decline is obscured in the original graph (graph 1) because it is masked by the long-term linear growth pattern (ie. linear growth + cyclical decline = apparent slow-down) — it is only when the two patterns are separated that they become clear.


This forecast helps explain the supposed contradiction between the bw166 data analysis, which reports an annual 1.1% increase in the size of the US market, and the recent data analysis from the International Wines and Spirits Record (IWSR), which reports a 0.9% decrease (US wine consumption falls for first time in 25 years). Neither of these reports is intrinsically “better” than the other — the former simply reflects the linear trend (graph 2) and the latter reflects the cyclical pattern (graph 3).

Anyway, there is my (unfortunate) forecast. As with all mathematical forecasts, its limitations are that it assumes that: (i) the current data are accurate and relevant; (ii) there are sufficient past relevant data to recognize the pertinent long-term patterns; and (iii) nothing new comes along to disrupt the current situation. The recent wrangle about tariffs is a relevant example of a potentially disruptive influence, as would be increasing effects of climate change.

However, if the forecast is accurate, at least in the short-term, then US wineries may soon need to place more reliance on their export markets (California export boosters aim for 37% more wine abroad by 2030). The alternative, I guess, is to reduce their vineyard area (eg. Wine slump threatens California vineyard cull).

PS. You may like to try the same analysis on the bw166 graph for the size of the US spirits market, which shows remarkably similar cycles.


  1. One bright light: the thousands of tasting rooms from the Rockies to the Atlantic welcome literally hundreds of thousands of new customers every week. There they learn to drink varietals, not brands and are a source of wine education/appreciation that will help all of us sell more wine to those a bit intimidated by some of our jargon and occasional pretense. A rising tide.....

  2. Good analysis. The degree is the downturn is the most uncertain aspect. Does it make sense to use the history of down turns, including and excluding the big 10 year decline for guidance?

    1. Forecasts don't really offer guidance. They simply say: "what happened last time will happen again next time". At best, that will often be approximately true, but is unlikely to ever be true in detail. The weather bureau says "it will rain a lot tomorrow", but they can't say exactly how much rain there will be.

    2. It's rained a zillion times, but there has only been one sustained market downturn in your analysis, so it seems to be quite a leap to base a prediction on a single occurrence. I have no clue, but do you have any idea what happened in the mid-80s to mid-90s? Doesn't appear to be related to economic activity.

      Also, you are using nominal cases, not %, so if your graph holds then the downturn would only be around half as severe as the previous one.

      I have no real doubt that a slowdown is coming, but maybe it's driven by factors not found in historic data.

  3. Plantings numbers need to be mandated to allow growers the opportunity to see 'the forecast' of wine grapes coming into the market in three years. Over planting is the reason for the downturn as more cases of wine are produced and not selling.

    1. As I emphasized in the post, it is a forecast not a prediction. It simply says "if things continue the same way, then this is what happens next". There may be individual explanations for each pattern, nevertheless.