Welcome to the second installment of this ongoing series looking at the technical side of the cigar world. In the first installment, I reported on research that was aimed at differentiating Cuban tobacco from non-Cuban tobacco using a fairly straightforward test that many analytical laboratories could execute. In this installment, I will step out of the lab and into the realm of marketing and economics to examine the relationship between quality, ratings, and price.
We’re going to take a look at a 2003 paper by Freccia, Jacobsen, and Kilby, professors of business administration and economics. The title of the article is Exploring the Relationship Between Price and Quality for the Case of Hand-rolled Cigars and it was published in the journal Quarterly Review of Economics and Finance. Their objective was to determine “whether either quality ratings, or hedonic equations including multiple sensory and objective characteristics, can explain price variance among hand-rolled cigars.” In other words:
Does a higher price generally buy a “better” cigar?
- Do ratings and price go hand in hand?
Do certain characteristics rate more highly with reviewers?
- Is there such a thing as the Cuban “twang?”
These are some of the big questions we’re going to explore in this installment. Pull up a chair, pour a tall one, and light that churchill because this will be a long read.
A. General Economics Background:
Freccia, Jacobsen, and Kilby take the basic economic tenet of a direct relationship between a product’s price and its quality and test it in the domain of hand-rolled cigars. The basic assumption is that as quality goes up, so should price, all else remaining equal. The authors use a technique called hedonic analysis (HA). This is a method of studying differentiated products. That is, a type of product that can be broken up into classes such as good, better, best. This technique has been used for over a century with great success on cars, appliances, and other durable (long lasting) goods. More recently, HA has been applied to all sorts of other goods and services. In all cases the analysis is based on inherent characteristics or features that are bound to the product. For example, an HA analysis of automobiles would consider things like type of engine (V6 or V8), type of transmission, fuel economy, leather or cloth seating surfaces. What would not be considered are advertising expenditures or market factors such as supply and demand.
B. Cigar Market Background:
This report covered the tail end of the cigar boom and the beginning of the cigar bust: 1992-1999. The authors reported that between 1992 and 1997 the import of handmade cigars rose from 99 million to over 500 million. Contrary to this pattern, Cuban cigar exports fell from 80 million in 1990 to 50 million in 1995 and did not improve much until 1997 when exports totaled 100 million. Around 1998, the cigar boom came to an end and total sales leveled off and actually started dropping. This was the beginning of the bust.
One important point to keep in mind is that this study covers a time frame when the number and proportion of shoddy “boom time” cigars and brands were at an all time high especially as compared to today. I took a look at my file copy of the October 1997 issue and saw such now defunct brands as Cacique, Don Asa, Don Leo, Habana Gold, Match Play, Tabacos San Jose, Canonero, Maya, Camorra, Jose Girbes, El Sublimado, Don Melo, and Excelsior.
C. Where did the data come from:
The authors used reviews from eight different issues of Cigar Aficianado (CA) published between Winter of 1992 and June of 1999. They only considered the coronas and double coronas sizes due to the fact that these two sizes provided the most number of cigars maximizing the size of the dataset and thus the reliability of the statistical analyses. A total of 689 cigars were included in this study.
Now why, might you ask, did they choose to use the ratings from CA magazine? There are several reasons. First, CA provides some detailed information about their methodology. Second, the reviewers (four senior editors) and the review methodology were stable and consistent over the time period of interest. Third, they provided market (retail) pricing for each cigar. Fourth, the tastings were carried out blind thus decoupling the effects of “band appeal” from smoking performance.
Even so, the authors noted potential risks with using the CA data. First, no justifications were provided as to the cigar tasting qualifications of the four senior editors. What was it about these guys that made their evaluations more worthy than, say, yours or mine? Second, no variance information was provided with the ratings. This means while they could assume the overall rating score was an average of four values, this could have meant that a cigar that scored 90 had ratings of 90, 90, 90, and 90 OR 100, 95, 85, and 80. Obviously the first set of ratings would carry a higher degree of confidence regarding the validity of the overall rating. The second set is such a mess that the average rating would be essentially useless. The authors also excluded the top and bottom scores so as to reduce the overall range and the standard deviation. Personally, I would have excluded only the lowest score as sometimes there was a difference of several points between this cigar and the second lowest scoring whereas the top scores were much more tightly clustered.
D. First hypothesis to be tested: Quality is directly related to price
The authors assumed that there was a positive correlation between quality and price. As quality (the CA blind ratings) goes up, price does as well. When you think about this, it makes a lot of sense. After all, shouldn’t it cost more to make things to a higher standard of quality? In the realm of cigars, “inherent” aspects of quality, or aspects related to the tobacco and its processing come to mind. Better care of the plants, well controlled curing and fermenting operations, longer aging in bales, skilled rollers, etc. These are all things that have costs associated with them.
The statistical analytical approach they took was to run a simple linear regression. This means they crunched the data to see to what degree a cigar’s price was related to its rating score. The key figure that this statistical method returns is called the R-squared (RS). An RS of 1 means that two quantities are perfectly correlated. For example, one’s height in inches and one’s height in feet. The first will always be 12 times the second. If you know one, you can always calculate or predict the other with perfect accuracy. On the other hand, an RS of 0 means that two quantities have no relationship to each other. For example, your shoe size and the amount of snowfall. The first would never be expected to predict the second. A value between 0 and 1 would indicate a relationship that has some slop in it. In the case of this paper, if the RS were high then we could conclude that higher price will buy a higher rated cigar.
So, what did they find out? The analysis showed that price has little relationship to quality (rating score). With RS’s ranging from 0.06 to 0.32 (where 0.75-0.95 would be considered good), one could conclude that price does not predict quality or, as stated by the authors, “high price is little guarantee of a high rating, and low price is little guarantee of a low rating.”
Since their first hypothesis was disproved, that is, price has no relation to quality, they then wondered if perhaps some specific characteristic or quality was more likely found in highly rated cigars. This brings us to their second hypothesis.
E. Second hypothesis to be tested: Certain characteristics are directly related to price and ratings
As an extension of the first analysis, the authors extracted several sets of variables from the data that CA reported in an attempt to find something specific that might explain reviewer ratings or cigar prices. They wanted to know if either ratings or price were correlated with the following things:
1. Vintage (were any years “better” or “worse”)
2. Country of Production (were any country’s cigars “better” or “worse”)
3. Subjective Variables (qualities like strength, smoothness, or good construction)
4. Flavors ( flavors such as cocoa, coffee, leather, or spicy)
With regard to effects on PRICE, the authors found that by far, the most powerful factor affecting price was the country of origin, Cuba. Simply being “made in Cuba” added $7.50 to a cigar’s price tag. Three subjective or flavor characteristics had statistically significant but far smaller effects. Good build, creaminess, and spiciness all increased the price of a cigar by $0.54-$0.65. The overall RS of this analysis was 0.58, which is not bad but not great. In essence, about 60% of the effect on price was explained by these variables. About 40% remained a result of things unknown.
Although the authors made quite a bit of this result, I am not convinced. In fact, I think they overlooked one very obvious fact that casts serious doubt on the significance of this finding. The non-Cuban cigars were purchased at retail mostly in the United States. The Cuban cigars, however, were purchased in the United Kingdom. While the cigar taxes can vary quite widely in the U.S., cigar prices in the UK are perhaps some of the highest in the world. On top of import duties, there is VAT (value added tax), and the additional markup levied by Hunters & Frankau, the sole official Habanos importer/distributor to the United Kingdom. This fact alone is able to explain almost the entire price bump by the “Cuba” variable.
One thing the authors might have done to address this weakness would have been to run a parallel regression using Habanos prices from Spain. In Spain, the single largest market for Cuban cigars, the prices are set by the state alone and are closer to those of comparable premium non-Cuban cigars. Prices at the Estancos (official Spanish cigar outlets) are close to the prices in Cuba and much lower than UK prices. This situation, in general, suggests that caution is called for when attempting to carry out cross-market studies using methodologies that are, by definition, blind to key market factors. In this case, ignoring the tax and value added costs scuttled the analysis.
With regard to effects on RATING, the results were a bit more interesting. Eighteen out of the 19 subjective and flavor characteristics were found to have statistically significant effects on the reviewer ratings. Curiously, only “tobacco flavor” was not statistically significant. The characteristics that increased ratings most strongly were leather, cocoa, spicy, and nutty flavors and good construction and strength. The characteristics that decreased ratings most strongly were papery, vegetal, and grassy flavors and inconsistency and mildness. These results probably capture the overall preferences and perceptions of cigar smokers at the most general level. I know it sort of describes my likes and dislikes in a generic sense so no surprises here.
The really interesting result lies in the results by country. As in the preceding analysis, the variable “Cuba” was the strongest factor. Unlike that analysis, the results were not invalidated due to market factors ignored by the hedonistic model. If a cigar was Cuban, that alone added over four points to its score. This effect was nearly three times stronger than any other subjective or flavor factor. The significance of this finding becomes apparent when we realize that these taste tests were done blind. The authors stated that “the ability of the judges to identify the Cuba characteristic in a blind taste test suggests the presence of a unique Cuban flavor (or potentially another identifying characteristic like color or shape).” I believe this could be the fabled “Cuban twang” that so many smokers profess to detect when smoking cigars from the island.
But could it be something else? Since these were all either coronas or double coronas, I do not believe shape or size was a distinguishing or prejudicial factor. Further, the wrappers of these cigars all varied from claro to maduro and Habanos fit well within this range. I think two slightly more plausible counterarguments might be the scent (barnyard) or the presence of a “triple cap”. It’s also worth noting that a cigar’s being Cuban added to the score indicates not only that this Cuban characteristic is detectable but also that it is something perceived as good. I’ve long held that this is not an absolute but rather a culturally imprinted preference.
Although the data were correlational and not from an experiment designed to determine cause and effect relationships, this finding is compelling. I believe the door has been opened on the issue of whether or not there is an underlying Cuban quality. If it does exist, it might be something that persists and is detectable no matter the superficial character of the cigar. The “Cuban twang” might not be just a myth after all.
Summing up:
1. Price and ratings are not related to each other.
2. Certain characteristics raise ratings while others lower ratings. Some unidentified characteristic of Cuban cigars was detectable by the raters and resulted in Cuban cigars being rated more highly.
F. Snob appeal, vanity, and utility
In the final section of their paper, the authors discussed what might account for the unknown 40% of the explanation. Why was this hedonistic model only partly successful in explaining the effect of cigar characteristics on price and rating? They proposed that this missing factor might be an aspect of utility. Utility is exactly what it sounds like: the “usefulness” that a product provides. Functional utility is one type of utility that encompasses those qualities the product provides through look, feel, taste, and size. These are the things inherent in the product and be discerned by holding, inspecting, or using the product. Nonfunctional utility are things applied to the product irrespective of its inherent characteristics. Price, for example is a nonfunctional utility.
The authors then talked about two effects related to nonfunctional utility. One, the “bandwagon” effect is an increase in NF utility as more people use the product. In other words, what’s popular is more desirable. The “snob” effect is almost the inverse of this. Here, the fewer people who are able to use something makes it more desirable. That is, exclusivity is desirable. A related effect that links to the “snob” effect is the “vanity” effect, a consequence of conspicuous consumption. Cigars are a product that are particularly vulnerable to these effects and Cuban cigars, especially so.
Postscript: Controversy Regarding the CigarAficianado Ratings
The CA cigar ratings have been controversial almost from the very beginning. The most fundamental concern deals with a suspected relationship between advertising and ratings. In other words, advertisers who pay for ad space get preferential treatment in the ratings process. However, if we are to believe CA’s long held claim that the reviews are done blind, then logically this cannot be a factor. The only way it could be a factor is if identifying information were provided outside of the blind process violating their own stated policy and thus the trust of their readers and the cigar market.
So then how might one explain the Cuban to non-Cuban ratings shift over the last 10 years? In the mid to late 1990’s it was not uncommon to see Cuban cigars dominate the top spots in each rating. In recent years, however, Habanos scores have spread out and non-Cubans have infiltrated the top of the ratings in greater proportion. Could this have anything to do with advertising fees as leverage? I’m inclined to say no. One reason is that there have never been any ad revenues from producers of Cuban cigars. Another explanation that is entirely plausible is that there are simply more good non-Cuban cigars being made these days. Yet another reason for the shift might have to do with the fact that the reviewers have learned to abstract essential qualities of a cigar no matter the national origin and as a result judge Cuban and non-Cuban cigars on the same unified quality scale. It is these last two explanations that, together, make the most sense to me.
Wilkey
We’re going to take a look at a 2003 paper by Freccia, Jacobsen, and Kilby, professors of business administration and economics. The title of the article is Exploring the Relationship Between Price and Quality for the Case of Hand-rolled Cigars and it was published in the journal Quarterly Review of Economics and Finance. Their objective was to determine “whether either quality ratings, or hedonic equations including multiple sensory and objective characteristics, can explain price variance among hand-rolled cigars.” In other words:
Does a higher price generally buy a “better” cigar?
- Do ratings and price go hand in hand?
Do certain characteristics rate more highly with reviewers?
- Is there such a thing as the Cuban “twang?”
These are some of the big questions we’re going to explore in this installment. Pull up a chair, pour a tall one, and light that churchill because this will be a long read.
A. General Economics Background:
Freccia, Jacobsen, and Kilby take the basic economic tenet of a direct relationship between a product’s price and its quality and test it in the domain of hand-rolled cigars. The basic assumption is that as quality goes up, so should price, all else remaining equal. The authors use a technique called hedonic analysis (HA). This is a method of studying differentiated products. That is, a type of product that can be broken up into classes such as good, better, best. This technique has been used for over a century with great success on cars, appliances, and other durable (long lasting) goods. More recently, HA has been applied to all sorts of other goods and services. In all cases the analysis is based on inherent characteristics or features that are bound to the product. For example, an HA analysis of automobiles would consider things like type of engine (V6 or V8), type of transmission, fuel economy, leather or cloth seating surfaces. What would not be considered are advertising expenditures or market factors such as supply and demand.
B. Cigar Market Background:
This report covered the tail end of the cigar boom and the beginning of the cigar bust: 1992-1999. The authors reported that between 1992 and 1997 the import of handmade cigars rose from 99 million to over 500 million. Contrary to this pattern, Cuban cigar exports fell from 80 million in 1990 to 50 million in 1995 and did not improve much until 1997 when exports totaled 100 million. Around 1998, the cigar boom came to an end and total sales leveled off and actually started dropping. This was the beginning of the bust.
One important point to keep in mind is that this study covers a time frame when the number and proportion of shoddy “boom time” cigars and brands were at an all time high especially as compared to today. I took a look at my file copy of the October 1997 issue and saw such now defunct brands as Cacique, Don Asa, Don Leo, Habana Gold, Match Play, Tabacos San Jose, Canonero, Maya, Camorra, Jose Girbes, El Sublimado, Don Melo, and Excelsior.
C. Where did the data come from:
The authors used reviews from eight different issues of Cigar Aficianado (CA) published between Winter of 1992 and June of 1999. They only considered the coronas and double coronas sizes due to the fact that these two sizes provided the most number of cigars maximizing the size of the dataset and thus the reliability of the statistical analyses. A total of 689 cigars were included in this study.
Now why, might you ask, did they choose to use the ratings from CA magazine? There are several reasons. First, CA provides some detailed information about their methodology. Second, the reviewers (four senior editors) and the review methodology were stable and consistent over the time period of interest. Third, they provided market (retail) pricing for each cigar. Fourth, the tastings were carried out blind thus decoupling the effects of “band appeal” from smoking performance.
Even so, the authors noted potential risks with using the CA data. First, no justifications were provided as to the cigar tasting qualifications of the four senior editors. What was it about these guys that made their evaluations more worthy than, say, yours or mine? Second, no variance information was provided with the ratings. This means while they could assume the overall rating score was an average of four values, this could have meant that a cigar that scored 90 had ratings of 90, 90, 90, and 90 OR 100, 95, 85, and 80. Obviously the first set of ratings would carry a higher degree of confidence regarding the validity of the overall rating. The second set is such a mess that the average rating would be essentially useless. The authors also excluded the top and bottom scores so as to reduce the overall range and the standard deviation. Personally, I would have excluded only the lowest score as sometimes there was a difference of several points between this cigar and the second lowest scoring whereas the top scores were much more tightly clustered.
D. First hypothesis to be tested: Quality is directly related to price
The authors assumed that there was a positive correlation between quality and price. As quality (the CA blind ratings) goes up, price does as well. When you think about this, it makes a lot of sense. After all, shouldn’t it cost more to make things to a higher standard of quality? In the realm of cigars, “inherent” aspects of quality, or aspects related to the tobacco and its processing come to mind. Better care of the plants, well controlled curing and fermenting operations, longer aging in bales, skilled rollers, etc. These are all things that have costs associated with them.
The statistical analytical approach they took was to run a simple linear regression. This means they crunched the data to see to what degree a cigar’s price was related to its rating score. The key figure that this statistical method returns is called the R-squared (RS). An RS of 1 means that two quantities are perfectly correlated. For example, one’s height in inches and one’s height in feet. The first will always be 12 times the second. If you know one, you can always calculate or predict the other with perfect accuracy. On the other hand, an RS of 0 means that two quantities have no relationship to each other. For example, your shoe size and the amount of snowfall. The first would never be expected to predict the second. A value between 0 and 1 would indicate a relationship that has some slop in it. In the case of this paper, if the RS were high then we could conclude that higher price will buy a higher rated cigar.
So, what did they find out? The analysis showed that price has little relationship to quality (rating score). With RS’s ranging from 0.06 to 0.32 (where 0.75-0.95 would be considered good), one could conclude that price does not predict quality or, as stated by the authors, “high price is little guarantee of a high rating, and low price is little guarantee of a low rating.”
Since their first hypothesis was disproved, that is, price has no relation to quality, they then wondered if perhaps some specific characteristic or quality was more likely found in highly rated cigars. This brings us to their second hypothesis.
E. Second hypothesis to be tested: Certain characteristics are directly related to price and ratings
As an extension of the first analysis, the authors extracted several sets of variables from the data that CA reported in an attempt to find something specific that might explain reviewer ratings or cigar prices. They wanted to know if either ratings or price were correlated with the following things:
1. Vintage (were any years “better” or “worse”)
2. Country of Production (were any country’s cigars “better” or “worse”)
3. Subjective Variables (qualities like strength, smoothness, or good construction)
4. Flavors ( flavors such as cocoa, coffee, leather, or spicy)
With regard to effects on PRICE, the authors found that by far, the most powerful factor affecting price was the country of origin, Cuba. Simply being “made in Cuba” added $7.50 to a cigar’s price tag. Three subjective or flavor characteristics had statistically significant but far smaller effects. Good build, creaminess, and spiciness all increased the price of a cigar by $0.54-$0.65. The overall RS of this analysis was 0.58, which is not bad but not great. In essence, about 60% of the effect on price was explained by these variables. About 40% remained a result of things unknown.
Although the authors made quite a bit of this result, I am not convinced. In fact, I think they overlooked one very obvious fact that casts serious doubt on the significance of this finding. The non-Cuban cigars were purchased at retail mostly in the United States. The Cuban cigars, however, were purchased in the United Kingdom. While the cigar taxes can vary quite widely in the U.S., cigar prices in the UK are perhaps some of the highest in the world. On top of import duties, there is VAT (value added tax), and the additional markup levied by Hunters & Frankau, the sole official Habanos importer/distributor to the United Kingdom. This fact alone is able to explain almost the entire price bump by the “Cuba” variable.
One thing the authors might have done to address this weakness would have been to run a parallel regression using Habanos prices from Spain. In Spain, the single largest market for Cuban cigars, the prices are set by the state alone and are closer to those of comparable premium non-Cuban cigars. Prices at the Estancos (official Spanish cigar outlets) are close to the prices in Cuba and much lower than UK prices. This situation, in general, suggests that caution is called for when attempting to carry out cross-market studies using methodologies that are, by definition, blind to key market factors. In this case, ignoring the tax and value added costs scuttled the analysis.
With regard to effects on RATING, the results were a bit more interesting. Eighteen out of the 19 subjective and flavor characteristics were found to have statistically significant effects on the reviewer ratings. Curiously, only “tobacco flavor” was not statistically significant. The characteristics that increased ratings most strongly were leather, cocoa, spicy, and nutty flavors and good construction and strength. The characteristics that decreased ratings most strongly were papery, vegetal, and grassy flavors and inconsistency and mildness. These results probably capture the overall preferences and perceptions of cigar smokers at the most general level. I know it sort of describes my likes and dislikes in a generic sense so no surprises here.
The really interesting result lies in the results by country. As in the preceding analysis, the variable “Cuba” was the strongest factor. Unlike that analysis, the results were not invalidated due to market factors ignored by the hedonistic model. If a cigar was Cuban, that alone added over four points to its score. This effect was nearly three times stronger than any other subjective or flavor factor. The significance of this finding becomes apparent when we realize that these taste tests were done blind. The authors stated that “the ability of the judges to identify the Cuba characteristic in a blind taste test suggests the presence of a unique Cuban flavor (or potentially another identifying characteristic like color or shape).” I believe this could be the fabled “Cuban twang” that so many smokers profess to detect when smoking cigars from the island.
But could it be something else? Since these were all either coronas or double coronas, I do not believe shape or size was a distinguishing or prejudicial factor. Further, the wrappers of these cigars all varied from claro to maduro and Habanos fit well within this range. I think two slightly more plausible counterarguments might be the scent (barnyard) or the presence of a “triple cap”. It’s also worth noting that a cigar’s being Cuban added to the score indicates not only that this Cuban characteristic is detectable but also that it is something perceived as good. I’ve long held that this is not an absolute but rather a culturally imprinted preference.
Although the data were correlational and not from an experiment designed to determine cause and effect relationships, this finding is compelling. I believe the door has been opened on the issue of whether or not there is an underlying Cuban quality. If it does exist, it might be something that persists and is detectable no matter the superficial character of the cigar. The “Cuban twang” might not be just a myth after all.
Summing up:
1. Price and ratings are not related to each other.
2. Certain characteristics raise ratings while others lower ratings. Some unidentified characteristic of Cuban cigars was detectable by the raters and resulted in Cuban cigars being rated more highly.
F. Snob appeal, vanity, and utility
In the final section of their paper, the authors discussed what might account for the unknown 40% of the explanation. Why was this hedonistic model only partly successful in explaining the effect of cigar characteristics on price and rating? They proposed that this missing factor might be an aspect of utility. Utility is exactly what it sounds like: the “usefulness” that a product provides. Functional utility is one type of utility that encompasses those qualities the product provides through look, feel, taste, and size. These are the things inherent in the product and be discerned by holding, inspecting, or using the product. Nonfunctional utility are things applied to the product irrespective of its inherent characteristics. Price, for example is a nonfunctional utility.
The authors then talked about two effects related to nonfunctional utility. One, the “bandwagon” effect is an increase in NF utility as more people use the product. In other words, what’s popular is more desirable. The “snob” effect is almost the inverse of this. Here, the fewer people who are able to use something makes it more desirable. That is, exclusivity is desirable. A related effect that links to the “snob” effect is the “vanity” effect, a consequence of conspicuous consumption. Cigars are a product that are particularly vulnerable to these effects and Cuban cigars, especially so.
Postscript: Controversy Regarding the CigarAficianado Ratings
The CA cigar ratings have been controversial almost from the very beginning. The most fundamental concern deals with a suspected relationship between advertising and ratings. In other words, advertisers who pay for ad space get preferential treatment in the ratings process. However, if we are to believe CA’s long held claim that the reviews are done blind, then logically this cannot be a factor. The only way it could be a factor is if identifying information were provided outside of the blind process violating their own stated policy and thus the trust of their readers and the cigar market.
So then how might one explain the Cuban to non-Cuban ratings shift over the last 10 years? In the mid to late 1990’s it was not uncommon to see Cuban cigars dominate the top spots in each rating. In recent years, however, Habanos scores have spread out and non-Cubans have infiltrated the top of the ratings in greater proportion. Could this have anything to do with advertising fees as leverage? I’m inclined to say no. One reason is that there have never been any ad revenues from producers of Cuban cigars. Another explanation that is entirely plausible is that there are simply more good non-Cuban cigars being made these days. Yet another reason for the shift might have to do with the fact that the reviewers have learned to abstract essential qualities of a cigar no matter the national origin and as a result judge Cuban and non-Cuban cigars on the same unified quality scale. It is these last two explanations that, together, make the most sense to me.
Wilkey
