How Expert Opinions Influence Consumer Wine Purchase Behavior: “Not All Publicity is Good Publicity”

Product quality and consumer awareness of that quality has a huge influence on consumption patterns.  Companies and advertisers have come up with several methods for increasing product awareness to consumers, the methods by which to inform consumers of quality are very important in products that fall under the “experience goods” category.  In a nutshell, experience goods are those products that one typically has to purchase or try before making a decision about quality.  These products include goods such as restaurants, movies, books, and wine.

In the literature, there have been many studies examining the extent to which product quality information affects consumer behavior.  One method, which has been less frequently studied, is the influence of quality information provided by “experts” to consumers.  One major problem with examining this method for conveying product quality information is that products that are high in quality are likely to receive high ratings and are in fact high in quality (not just saying they are).  Therefore, it becomes difficult to determine how much of the consumer behavior is influenced by expert opinion, since the researchers would need to be able to control for product quality (which is extremely difficult to do).  Another complication with this method is that it is difficult to determine if consumer behavior is influenced completely by expert opinions, or if the review simply alters the consumer to the product, and it doesn’t matter what the overall quality score ended up being.

The goal of the paper briefly presented today (I’m not going to go into all the gory details this time unless requested) was to examine the impact of expert opinion on retail wine purchases using an experimental approach at stores of a national grocery chain in northern California.

Brief Summary of Methods

150 wines that were typically stocked in a national grocery chain were randomly selected.  Wines were only selected if they had a wine score, which somewhat limited the pool of wines to be randomly selected from (14% of all the wines in the store). 

Wine scores were presented on the price tag of wines in the “treatment store” for four weeks during April of 2006.  Half the wines received scores between 82 and 86, and 90% of the wines received scores between 78 and 89.  For all wines sold by the treatment store, weekly sales data were collected.  In addition to the treatment store, the weekly sales data from 38 other “control stores” from Northern California were collected.  Weekly sales data were aggregated to monthly sales, and included the total number of bottles sold per month, average pre-discount price, average post-discount price, and whether a bottle of wine was discounted at all during the month.  Sales data were then merged with wine score data.

To examine the extent to which treatment wines increased wine sales, the researchers used several mathematical models comparing treatment wines with control wines that did not have the wine score labeling treatment.  (I will spare you the mathematical details, but if you have specific questions, I can try to find out the answer for you.)


  • The average number of bottles of “treated” wines sold increased by 1.5 bottles from March to April in the “treated” store.
  • The average number of bottles sold decreased by 1.1 bottles from March to April in the control stores.

o   “Treatment” (providing wine scores to price tags) increased consumer demand of treated wines by 2.6 bottles, or 27%.

  • In the treatment store, average sales of treated wines is driven by an increased demand for high scoring wines, which increased from 9.4 to 11.5 bottles on average.

o   In the control stores, high scoring wine sales decreased from 10.8 to 9.6 bottles on average.

  • In the treatment store, Sales of low scoring wine sales decreased from 9.9 to 8.2 bottles on average.

o   In the control stores, low scoring wine sales decreased from 13.8 to 12.7 bottles on average.

§  Low scoring wines showed a treatment effect of a decrease of 0.7 bottles, whereas high scoring wines showed a treatment effect of an increase of 3.2 bottles.

  • Wines with higher scores have larger increases in demand due to the treatment.
  • The average score that the treatment effect was zero was 80.

o   Wines scoring higher than 81 experienced a positive increase in consumer demand.

o   Wines scoring lower than 81 experienced a decrease in consumer demand.

  •  Demand for wines in the treatment store was less price-sensitive than demand at the control stores.
  • Consumers responded positively to high-quality information signals and negatively to low-quality information signals.

o   This suggests “not all publicity is good publicity”.

o   Consumers exposed to high quality indicators are less influenced by price.

  • At the treatment store, consumer demand for “untreated wines” remained steady.

o   It appears as though average purchases of “unlabeled wines” remained constant, while purchases of “labeled wines” increased.


According to the authors of this study, they were able to successfully analyze the relationship between product quality and expert opinion in a retail grocery chain setting.  By attaching the wine’s score to the price tag of the wine, the results showed that sales increased on average by 25% and that high-scoring wines displayed a higher increase in sales compared to their low-scoring counterparts.  It is important to note that these increases were only found in treated wines, meaning that if there was another high-scoring wine present but it wasn’t advertised as such, sales would not increase for that particular wine.

Some limitations of the study are that the treatment period may not have been long enough to observe the full effect of expert opinion on consumer purchase behavior of wine.  Also, results may not be generalizable to the public as a whole, as only one treatment store was included in this study and the demographics of that particular store may not represent the larger consumer public as a whole.  Specific demographics of the treatment store clientele were not described in the paper, except for the fact that they were wealthier than many in the general public.

Overall, the results of this study suggest that perhaps expert opinion has a strong influence on consumer purchase behavior; in that higher scoring wines are more often purchased than lower scoring or non-reported wines.  Wine shops may wish to take advantage of this information and post wine scores when applicable.   Of course, much more research needs to be done in order to get a stronger sense of the mechanism behind this relationship, in addition to how important it is compared to other consumer purchase behavior influencers, but it is in and of itself interesting and a good addition to this field.

I’d love to hear what you all think about this topic!  Please feel free to leave your comments below!  (Reminder: any unapproved html tags will be promptly deleted.  Links to personal blogs are accepted).

Source: Hilger, J., Rafert, G., Villas-Boas, S. 2011. Expert Opinion and the Demand for Experience Goods: An Experimental Approach in the Retail Wine Market. The Review of Economics and Statistics 93(4): 1289-1296.
I am not a health professional, nor do I pretend to be. Please consult your doctor before altering your alcohol consumption habits. Do not consume alcohol if you are under the age of 21. Do not drink and drive. Enjoy responsibly!

6 comments for “How Expert Opinions Influence Consumer Wine Purchase Behavior: “Not All Publicity is Good Publicity”

  1. Christian Miller
    September 15, 2012 at 9:31 am

    Fascinating study, thanks for bringing it to our attention. One question – was price controlled for, either relative or absolute changes in price?

    • Becca
      September 15, 2012 at 9:32 am

      Thanks for reading and commenting, Christian!

      While the authors didn’t intentially set out to control for price, they did keep record of the prices of wines in each store. For the treatment store and the control stores that had wine scores on their wines already (not purposely placed their by investigators), the average prices of wines did not appear to be different. The average prices of wines at untreated stores with no scores appeared to be a little higher, however, these statistics are all merely descriptive.

      They did find that price was less important if wine scores were higher, though other than that, there wasn’t much of a focus on it. I would be interested to see what would happen if price were more controlled than it was!

  2. Tom
    September 15, 2012 at 9:32 am

    How many treatment stores were included? Also if 80 was the no-change point and 78 was the low score, it strikes me that people were looking at the scores like letter grades, instead of knowing that most reviewers don’t give numerical scores that go too low anyway.

    • Becca
      September 15, 2012 at 9:32 am

      Hi Tom,

      Thanks for your comments! One of the major limitations of this study was that there was only one treatment store. I’m a little confused as to why they only had one treatment store, since they had no trouble finding 38 control stores, but having been in research in the past, I think it may have been a result of financial limitations or simple lack of permission from the other stores.

      Anyway, I’d mentioned in one of the last paragraphs that demographics of that one particular treatment store may not represent the larger consumer public as a whole, and that more research needs to be done that includes many more stores. I think this was a pilot study that they managed to get published, the results of which are interesting, but a result to the single treatment store, must be taken with a grain of salt.

      Thanks for reading!

      • Tom
        September 15, 2012 at 9:33 am

        I don’t see that financial limitations are a real concern here — they put shelf tags on in supermarkets and got the sales from scanner data. It may have been a permission issue, but that also seems unlikely. If they were able to use previous years’ scanner data to select control scores, it seems ridiculous that they couldn’t have done better than this.

        And no one does a market research study without controlling for prices. What were they thinking? I’m amazed this was published as is.

        • Becca
          September 15, 2012 at 9:34 am

          I suppose I was just trying to give them the benefit of the doubt. Who really knows why they didn’t do more. Maybe the store owners were all just really not into it and wouldn’t allow them to do it? No clue!

          Maybe the editors were asleep at the wheel the day they read this article submission…

Comments are closed.