NEWS Flash: Labelled alcohol level frequently understated

NEWS Flash: Labelled alcohol level frequently understated
http://www.sfgate.com/cgi-bin/article.cgi?f=/c/a/2011/04/24/FD311J4I7H.DTL&type=wine&ao=5

Perhaps this is as well known as the tendency to understate one's income to the IRS, but it was a surprise to me. The Chronicle did a study of 19 wines using a very accurate (and, admittedly, expensive) method to compare the labeled alcohol percentage to the liquid in the bottle. Of the 19 wines, 8 were understated by 0.5% or more, and only one was overstated by that much. I'm no statistician, but that doesn't seem like measurement error to me! For those whose binomial distribution calculation skills are a little rusty, the odds of that happening by chance (assuming there's a 50% chance of "meaningful" understatement or overstatement) are about 2%.

For those who didn't know, a measurement of 12.5% on a label means, legally, that the wine can be between 11 and 14%! That's a lot of leeway, but I could understand when you're dealing with inaccurate instruments and barrel variation. Not every winery has a tank big enough to fit an entire vintage, and they certainly don't all have $16,000 Alcolyzer instruments. It's also likely that a winery or negociant could have multiple cuvées with different percentages. I'm the last one to impose an onerous restriction on small wineries, but this is systematic understatement, not understandable variation. To put the amounts in perspective a 1 percentage point difference is about 7 calories worth of alcohol, so also within the nutritional labeling guidelines, if they applied to wine. With an eye toward libel law, I am not saying that any of the wineries deliberately understated the alcohol percentage (and in the Chronicle's study, the amount of understatement was within the law), but it is almost impossible for random variation to be the cause.

Certainly it does not help that the federal government charges a fee (not to mention the increased headache) to change the alcohol level on a label. To make it easier to be compliant, there should be no charge for an alcohol change, even mid-vintage. If the winery is willing to print the labels, don't punish them for accuracy. The freakonomist in me says that maybe you should charge the wineries that *don't* update their labels yearly, to give them an incentive for accuracy.

As for enforcement, I'm somewhat cheered that the label alcohol is reasonably accurate. Without a regular testing program, there doesn't seem to be much reason for honesty on the part of the winery. I would like to see the TTB sample wines at retail every few years. It could be an opportunity to test for alcohol accuracy, pesticide residue, etc., and would not place a burden on wineries to provide a sample for testing.

For the consumer, the take-home message is, expect that the alcohol level is understated by, on average, .4%. It's not a big amount, but it is real.

[David M. Bueker]

Certainly it does not help that the federal government charges a fee (not to mention the increased headache) to change the alcohol level on a label.

Actually, according to winemakers who have posted on this topic on other sites, the government label approvals do not include alcohol level in the approval.

What states do (and yes, states get their say as well) I am not sure.

...Not every winery has a tank big enough to fit an entire vintage, and they certainly don't all have $16,000 Alcolyzer instruments...

Instrumentation is really not necessary. There are several independent labs that provide detailed wine analysis, and they're not that expensive. In fact, from posts I read on another website, some home wine makers even use the labs. One popular lab is at Virginia Tech. That being said, I don't use a lab, but do put the abv on my home wine labels, based on the potential alcohol scale of my hydrometer, which is about 55% of the Brix (% sugar). This is pretty close and works for whites, rosés and reds that don't see a barrel.

Actually, according to winemakers who have posted on this topic on other sites, the government label approvals do not include alcohol level in the approval.

What states do (and yes, states get their say as well) I am not sure.

Dear David,
Interesting. I was relying on the article, which says, that Adam Lee, owner of Siduri, "rarely modifies the alcohol numbers on his labels, in part because of the cost of filing for a new label approval from federal and state agencies." Whatever the reason, infrequent modification would not cause the systematic understatement shown here.

Dear Howie,
I was assuming that most wineries bottled and labeled at the same time, and therefore would need several weeks (?) lead time to get labels approved and printed. Therefore they might not have time to send off for a measurement after making the final blend. Mainly I'm trying to see any possible reason why there might be systemic understatement of alcohol levels. It seems like the most likely explanation is a desire to lower the printed level that the consumer sees.

Walt

In addition, with a wine stated as 12.5% and a range up to 14%, the winery will pay a lower tax rate than if the wine is over 14%. Here's a link: http://www.drvino.com/2010/08/24/wine-alcohol-tax-wine-spirits-magazine/

Wine and Spirits did a similar type of snapshot last fall. They used City Winery and Crushpad to test 84 wines using Ebulliometer to test alcohol. Half were US (all west coast) and the other half were rest of the world. Of the US wines tested 27 had more alcohol than stated on the label. Of the world wines tested 26 had more alcohol than stated on the label. All total there were 53 wines that had higher alcohol than stated, 13 where it was as stated, and 18 where the alcohol was lower than the lable. A few interesting points; of the 20 Australian wines tested 7 had less alcohol than on the label and the only country that had the same over as under was Argentina but it was only 3 wines tested. France and Italy were both 3 over to 1 under.

Not sure how much one can conclude from small sample sizes such as this as they relate to the larger context of alcohol content reported on the label. Also not sure how acurate one test method is compared to another or the process variability. When I look at the label I just assume +/- 0.3%.

That being said, I don't use a lab, but do put the abv on my home wine labels, based on the potential alcohol scale of my hydrometer, which is about 55% of the Brix (% sugar). This is pretty close and works for whites, rosés and reds that don't see a barrel.

Howie, I have read that the Brix to alcohol conversion can vary from 0.55 to 0.61 depending upon numerous factors. Most I know use a factor somewhere in the .57 to .59 range. I agree its close enough but the difference between 0.55 and 0.59 is almost 1% at 23 brix. Because I don't test and since some people make such an issue out of the alc% I am thinking of not labeling my bottles anymore and only noting the level in my notebooks.

I believe it is still legal and I still see it done to not reveal any alcohol level on a wine if it is under %15 and labeled "table wine".

This is a non-story.

It is not as if US law required wineries to print an accurate alcohol percentage.

To note that a wine that their test showed had 15.46% but was labelled as 15.5% is just silly -- they appear to have rounded to 2 decimal points (since they don't show any wines with more) and the winery has rounded to one.

The sole European wine is labelled as per EU regulations to the nearest 1/2 percent at 12.5% and their test showed it had 12.71%.

As others have noted, labels are usually printed before wines are bottled, so the abv will be measured while wine is being made and there may be small changes in the actual abv by the time the wine bottled.

And, again as previously noted, in the USA wines aren't required to show the abv if the use the words 'table wine' on their labels.

I believe it is still legal and I still see it done to not reveal any alcohol level on a wine if it is under %15 and labeled "table wine".

Dear Jim,
Close, the magic number is 14%. However, I don't see that done much any more. In fact, the only time I see it done is with imported wines where the existing front label will not pass muster with TTB (due to using a comma in place of the decimal point or similar issues) and the importer slaps a label on the back with "White table wine" in place of a real label.

This is a non-story.

It is not as if US law required wineries to print an accurate alcohol percentage.

It is true that there is a lot of legal leeway in the U.S. As I said above, I'm actually encouraged that all of the wines were more or less within the legal limit. And that's why, when I was giving my analysis, I ignored the wines that were within .5% either way. Still, I don't think the average consumer knows to add a fudge factor due to the fact that, as Brian confirmed above, systematic understatement is the norm.

From a stylistic standpoint, I would really prefer more accuracy. I might take a flyer on a unknown California pinot noir at 14.0%, but almost certainly not at 15.0%. (Not making a statement here, there are pinot noirs that I enjoy at 15%, but that level is really hit or miss for my tastes).

Walt

Wine and Spirits did a similar type of snapshot last fall. They used City Winery and Crushpad to test 84 wines using Ebulliometer to test alcohol. Half were US (all west coast) and the other half were rest of the world. Of the US wines tested 27 had more alcohol than stated on the label. Of the world wines tested 26 had more alcohol than stated on the label. All total there were 53 wines that had higher alcohol than stated, 13 where it was as stated, and 18 where the alcohol was lower than the lable. A few interesting points; of the 20 Australian wines tested 7 had less alcohol than on the label and the only country that had the same over as under was Argentina but it was only 3 wines tested. France and Italy were both 3 over to 1 under.

Not sure how much one can conclude from small sample sizes such as this as they relate to the larger context of alcohol content reported on the label. Also not sure how acurate one test method is compared to another or the process variability. When I look at the label I just assume +/- 0.3%.

Dear Brian,
Thanks for the follow-up. The sample size is small but the data are overwhelming. There's just no way that you can have random variation that runs 3 to 1. Putting +/- implies says that you're as likely to be over as under, so I think my rule of thumb will be to add +.5%.

Walt

Wine and Spirits did a similar type of snapshot last fall. They used City Winery and Crushpad to test 84 wines using Ebulliometer to test alcohol. Half were US (all west coast) and the other half were rest of the world. Of the US wines tested 27 had more alcohol than stated on the label. Of the world wines tested 26 had more alcohol than stated on the label. All total there were 53 wines that had higher alcohol than stated, 13 where it was as stated, and 18 where the alcohol was lower than the lable. A few interesting points; of the 20 Australian wines tested 7 had less alcohol than on the label and the only country that had the same over as under was Argentina but it was only 3 wines tested. France and Italy were both 3 over to 1 under.

Not sure how much one can conclude from small sample sizes such as this as they relate to the larger context of alcohol content reported on the label. Also not sure how acurate one test method is compared to another or the process variability. When I look at the label I just assume +/- 0.3%.

Dear Brian,
Thanks for the follow-up. The sample size is small but the data are overwhelming. There's just no way that you can have random variation that runs 3 to 1. Putting +/- implies says that you're as likely to be over as under, so I think my rule of thumb will be to add +.5%.

Walt

Not sure I agree. I finally looked at the SFChron article and numbers reported and if you note it reports a discrepency of 0.4% between what they tested the Pepperbridge and what the winery tested it at using and Ebulliometer. Also note that they report 0.2% discrepency on the Hall CS using the same technology but with different testings. The original story reports an over/under of 14/3 but if you consider the potential for a 0.2% discrepency that drops to 12/7 and a 0.4% discrepency drops it down to 10/9. I don't have the Wine and Spirits numbers with me right now but the majority of the wines were within 0.3% of what was on the label. So if there is a variance of ~0.3% from what was reported the conculsions change from a non random pattern of underreporting to random variation. The obvious question I have is why is the SFChron testing showing higher alc than the other testing? Both retests metioned resulted in lower alcohol measurement. To me it appears that there is not a standard test method that one can be sure is providing the definitive answer on how much alcohol is actually in the wine. Lacking this, other conclusions are difficult to make.

From talking to CA producers, my sense is that it's common to understate by the permissable 1.0. The government only gets involved if the taxes aren't being properly paid, ie if the wine says 13.9 and it's actually 14.1 they are very interested. (I'd be curious to know how many producers understate by more than 1.0...)

I would be curious whether the accuracy of the label was related to the level of alcohol. I would expect to see more understatement at the higher levels of alcohol, there's no motive to understate if you're naturally low.

EU rules say 0.5 either way, and AFAIK my producers label using that convention.

In my view a much more important number than alcohol is potential alcohol at harvest. It's common for expensive CA wine to show eg 14.8% as bottled wine, but if the wine was picked at 28 or 30 brix (or higher) the texture and flavors of the wine come from that number*. The alcohol is then lowered by de-alcoholisation or the addition of water. In other words the character of the wine (soft tannins, jammy flavors) is determined in large part by picking at a certain level of sugar, the alcohol is easily managed thereafter.

*Adam Lee referred to a batch of Pinot fruit that was picked at 33 brix as 'not overripe' in the recent Pinot Noir conference, apparently because the pH was low.

[David M. Bueker]

FYI, in another discussion of this same topic elsewhere, Adam Lee (who is quoted in the original article) discussed the cost in label changes as being from state registration processes. Some require re-registration even for a vintage change. Some charge fees while others do not.

From talking to CA producers, my sense is that it's common to understate by the permissable 1.0. The government only gets involved if the taxes aren't being properly paid, ie if the wine says 13.9 and it's actually 14.1 they are very interested. (I'd be curious to know how many producers understate by more than 1.0...)

I would be curious whether the accuracy of the label was related to the level of alcohol. I would expect to see more understatement at the higher levels of alcohol, there's no motive to understate if you're naturally low.

So I did this quickly last night from the W&S numbers reported. Of the samples tested 8 were understated by 1.0 or more. The label reported alcohol for those eight were 12.5, 13.5, 13.5, 13.5, 14.0, 14.5, 14.5, 14.9. There were 4 from the US and 4 from outside the US. Notice that 7 of the 8 are standard label alcohol readings. As the 14.9 is a US wine, I could argue that all 8 are label values that probably never change from year to year.

As for all the reported understatmens 33 were for wines listed at 14% or above and 21 were for those listed under. Of those 21 below 14%, 9 were below 13%. For the overstatments, 7 were over 14% and 10 were under 14%. (Note that I either miscounted for my previous post or this one as the over/under numbers are off by one).

Probably the most telling thing I noticed is that out of the 61 wines that were noted to test different than the label, 41 of those had a labeled alcohol level that ended in either .0 or .5. suggesting that these are standard label values not adjusted yearly.

Brian, could you please post a link to the raw data? I'm having trouble finding it. Thanks.

I could not find it online either. Been using the print mag. When I get to a real computer will add link to the proper vol & % issue.

i remember john konsgaard once telling me that he used to redo the labels every year with the exact alcohol. one year's chard was popular at 13.8%; and the next year was 13.3. he got complaints that that vintage was too 'light'. from that point on they were all 13.8. this was when he was with Newton i believe; and a long time ago. not many chards that low anymore unless they are dealcoholized. i'm impressed from the numbers that noone is screwing around with the 14% tax threshold.

Brian, could you please post a link to the raw data? I'm having trouble finding it. Thanks.

I found the story before on the wine and spirits page but can not even find that now. I did find it on Dr. Vino but also without the table.
http://www.drvino.com/2010/08/24/wine-alcohol-tax-wine-spirits-magazine/

Some might find the following link interesting-
http://www.wineandspiritsmagazine.com/p ... x_man.html


Salute