Consumers’ ability to distinguish between milk types: results of blind taste testing

Abstract

Objective:

To assess consumers’ ability to correctly identify different types of milk in a blind taste test and correlates of plans to purchase lower-fat milk.

Methods:

Adults from eight supermarkets in low-income neighborhoods tasted three types of unlabeled lower-fat or fat-free milk samples and guessed the type of each sample.

Results:

Of the 1074 participants, only 7.6% were able to identify all three unlabeled samples correctly.

Conclusions and Implications:

Most adults in this study reported consuming higher-fat milk and could not correctly identify milk type by taste alone. Blind taste tests may encourage consumers to drink lower-fat milk.

INTRODUCTION

Dairy products, including milk, contain essential nutrients that are vital to maintaining overall health¹. Fat-free and low-fat products provide the same nutrients as full-fat products, but with less fat, and are recommended in the US Dietary Guidelines as healthier options for adults¹. Despite health benefits, including increased calcium intake and reduced calorie and saturated fat intake, milk sales have declined significantly in recent years. Between 2000 and 2012, annual dairy milk sales (fat-free, 1%, 2%, and whole) in the United States declined by more than ten million pounds² or five percent. Currently American adults and children consume below the recommended three servings, or cups, of fat-free or low-fat (1%) fluid milk or dairy equivalent in milk products per day. ^{1, 3}

Milk consumption patterns vary across demographic groups by race, income, gender, and age^4-8. Non-Hispanic whites consume more low-fat milk (1% fat or lower) than Hispanics and non-Hispanic blacks.^4,5 This may be due to differences in access to low-fat milk. One recent study found that in majority black communities, there are 31% to 67% lower odds of local stores carrying milk than in majority white communities⁹. Grocery stores in low-income communities have three to five times the odds of carrying whole milk (3.25% fat) than grocery stores in higher-income communities. ^9,10. Milk consumption also varies by gender and age^6-8. On average, males age 20 years and older consume more milk than females⁶. However, over time, from age 20 to age 70, milk consumption among males has decreased while it increased among females⁶.

Milk-drinking beliefs and established purchasing habits affect the type of milk that people choose to drink ^11,12. The fat content of milk is a critical driver of purchasing, and common descriptions of low-fat milk include perceptions that it has a “watered down taste,”^11,12,13 a belief that may be a driving cause of continued purchasing of higher fat milk. Blind milk taste-testing has demonstrated that despite shoppers’ confidence in their ability to correctly identify milk fat content, many are actually not able to accurately distinguish between types.^14,15 Further, taste testing has proven a successful strategy in opening consumers’ minds to lower-fat milk^14-16. Shoppers who participated in blind taste tests have often been unable to identify the different types of milk sampled, and following a blind taste test, many consumers report that they like fat-free or low-fat milk, and would consider purchasing it as a replacement for fuller fat milk ^14,15,17-18.

The overarching goal of this study is to determine whether shoppers in low-income neighborhoods would consider purchasing lower-fat milk after participating in a blind taste test. Our study builds on prior research but uses a larger sample, enabling a more detailed analysis of the characteristics of those who are willing to switch from a whole milk to a low-fat milk purchase after taking part in the blind taste-test. Specifically, this study aims to (1) examine consumers’ ability to correctly identify milk samples of varying fat content; (2) determine consumers’ willingness to consider purchasing lower-fat milk following a blind taste test; and (3) describe demographic characteristics associated with consumers’ willingness to switch to lower-fat milk.

METHODS

Study Design

Blind milk taste tests were completed in stores participating in a cluster randomized controlled trial evaluating the effects of healthy in-store marketing strategies on sales of healthier food items in 33 supermarkets in low-income neighborhoods in the Philadelphia, PA metropolitan area. This study was conducted at the first eight intervention stores in the study. To be eligible for the blind taste-test, participants had to be 18 years of age or older, be the primary household shopper, read and speak English, do most of their shopping at the participating store, and regularly consume white cow’s milk.

Procedures

Taste tests were conducted by trained research staff once a month at each store, for 12 months. Procedures were similar to those reported in Weiss et al.,¹⁵ conducted as part of a pilot study¹⁹ that formed the basis for the present larger trial. Taste tests took place in stores on varied weekdays between the hours of 10 am and 4 pm. Staff continued the milk taste tests with consumers until 25 participants were enrolled or a until 4-hour time period elapsed.

Shoppers were recruited to participate in the supermarket and, if they agreed, they completed a short survey on which they reported their age, gender, race, ethnicity, and the type of unflavored cow’s milk (fat-free/skim, 1%, 2% or whole) they typically consumed. They were then provided three unlabeled samples containing fat-free, 1%, and 2% milk which was purchased from the store where it was sampled. Whole milk was not in the taste test both to simplify the task, and to see if participants might think lower-fat milk tasted like whole milk without having a direct comparison. Samples were provided in 3-ounce paper cups pre-labeled with different color stickers to allow staff to pour the designated type of milk in the cup and to record participant responses. Cups were set on the table for the taste tests. Milk cartons were kept cold, in coolers, on ice, and staff ensured that the milk samples were not confused by color coding cartons with stickers that were not visible to shoppers during taste testing. Participants were instructed to taste the milk samples, in an order of their own choosing, and attempt to identify type of milk tasted by placing each sample on a game board-type table display which included four large quadrants labeled with 4 options; non-fat/skim, 1%, 2%, and whole milk . Participants took one sample at a time and placed it on the quadrant that they believed matched that milk selection. Responses did not need to be mutually exclusive (i.e., respondents could re-select the same quadrant) and were each recorded by staff. After the taste test, participants were informed about the accuracy of their responses and of the correct answers, if they gave incorrect responses. Those who reported usually drinking 1%, 2%, or whole milk were asked if they would consider switching the type of milk they typically drink, and if yes, which type of milk they would switch to. All study procedures were approved by the University’s Institutional Review Board.

Data Analysis

All analyses were conducted using Stata MP 15 for Windows (StataCorp; College Station, TX). First, descriptive data were summarized to characterize the participants and to assess whether they correctly identified the type of milk samples they tasted. Chi-square analyses were used to test whether the pattern of correct/incorrect identification of milk types differed for each level of fat content (fat-free, 1%, 2%). Next, participants who reported they would consider changing the milk type they drink were compared to those who reported they would not; an independent t-test was used to assess the mean age differences and chi-square tests were used to test for differences in gender, race, ethnicity, and milk type typically consumed. Participant characteristics that were significant in bivariate analyses were included in a logistic regression model to assess predictors of participants’ consideration to purchase a different type of milk.

RESULTS

A total of 1,074 participants (64.2% female) completed milk taste tests from December 2016 through November 2017. Participants were an average of 52.4 ±18.4 years old; 65.4% non-Hispanic white, 24.9% non-Hispanic black, 5.0% Hispanic, and 4.9% other racial/ethnic identity.

Overall, participants reported typically consuming 2% milk (37.1%) or whole milk (34.2%) most often, followed by 1% milk (15.9%), and fat-free milk (12.8%). This pattern was consistent across non-Hispanic black, non-Hispanic white and Hispanic race categories. Non-Hispanic black participants reported consuming whole milk more often than non-Hispanic white and Hispanic participants (42.4% vs. 29.3%, p < 0.001). Non-Hispanic white participants were more likely to report consuming 1% milk than were other groups (17.7% vs 10.7%, p < 0.001). Table 1 shows how many participants identified each sample as fat-free, 1%, 2% or whole milk; and the proportion who correctly identified each type of milk. Two-percent milk was most frequently misidentified 44.5% of the time as whole milk and participants were most likely to identify fat-free milk samples correctly (34.7%), followed by 2% (31.2% correct) and 1% milk samples (25.5% correct). While for any one type of milk, correct selections ranged from 34.7% to 25.5%, only 7.6% of participants were able to identify all three types of milk correctly. The average number of correct guesses did not differ by gender, race/ethnicity or type of milk usually purchased.

Table 1:

Participants' guesses for each sample tested and percentage correct^* (n=1074)

	Guess (%)
	Fat-free Milk (n=605)	1% Milk (n=794)	2% Milk (n=964)	Whole Milk (n=859)
Milk Sample
Fat-free Milk  (n=1074)	34.2	33.2	20.7	11.9
1% Milk  (n=1074)	13.0	25.5	37.9	23.6
2% Milk  (n=1074)	9.2	15.1	31.2	44.5

^*Pearson X² statistic = 553.59, p < 0.001

Shading indicates percent of guesses that correctly identified the milk sample

When asked whether they would consider purchasing a different type of milk following the taste test, 40.6% of participants said yes (Table 2). Those willing to consider a different milk-fat were significantly younger, more likely to be female, non-Hispanic black, and current whole milk consumers than those unwilling to consider a change. There was no significant association between the ability of a participant to correctly identify milk samples, and whether or not they said that they would consider switching to a different type of milk.

Table 2:

Bivariate Analysis: Participant Characteristics by Consideration for Switching Milk Type (n=936)

	Consider Switching Milk Type
	No		Yes		Test Statistic*	P Value
	Mean	±SD	Mean	SD
Age (n=933)	54.3	18.5	46.5	17.2	6.6	<0.0001
Gender (n=936)	n	%	n	%	21.1938	<0.001
Male	234	69.2%	104	30.8%
Female	322	53.8%	276	46.2%
Race (n=923)					21.0907	<0.001
Black or African American	124	48.6%	131	51.4%	17.9435	<0.001
White	371	64.6%	203	35.4%	15.382	<0.001
Hispanic or Latino	25	52.1%	23	47.9%	1.2198	0.269
Other	31	67.4%	15	32.6%	1.1914	0.275
Milk Type Purchased (n=936) **					19.3094	<0.001
1%	122	73.5%	44	26.5%	16.6168	<0.001
2%	240	59.1%	166	40.9%	0.0247	0.875
Whole	194	53.3%	170	46.7%	9.2055	0.002
Guesses Correct (n=936)					3.5438	0.315
Zero	207	59.0%	144	41.0%
One	238	59.6%	161	40.4%
Two	65	54.6%	54	45.4%
Three	46	68.7%	21	31.3%

^*Age was compared with an independent t-test; categorical variables were compared with X² Statistic

^**Fat-free milk drinkers (n= 138) were not asked whether they would consider switching milk type

Age, gender, race, and milk type purchased were included in a logistic regression model of predictors of participants’ consideration to purchase a different type of milk (Table 3). Variables were assessed for multicollinearity, and all observed variance inflation factors were below ∣0.70∣. In this model, participants’ age and being female were significant predictors of considering purchasing a different type of milk. Participants who typically purchased 2% milk or whole milk had higher odds for considering switching to a different type of milk. In the multivariate analysis, race was no longer a significant predictor of considering purchasing a different type of milk.

Table 3:

Logistic Regression; Predictors of Willingness to Consider Switching to a Different Type of Milk (n=936)

Independent Variable	Odds Ratio	Standard Error			95% CI
			Z	P >∣z∣	Lower	Upper
Age	0.98	0.00	−4.81	<0.001	0.97	0.99
Gender	2.00	0.31	4.52	<0.001	1.48	2.70
Race
White	--	--	--	--	--	--
Black	1.32	0.22	1.65	0.098	0.95	1.83
Hispanic	1.10	0.35	0.3	0.762	0.59	2.05
Other	0.63	0.22	−1.34	0.182	0.32	1.24
Milk Type Purchased
1%	--	--	--	--	--	--
2%	1.80	0.38	2.78	0.005	1.19	2.72
Whole	2.44	0.53	4.12	<0.001	1.60	3.73
Constant	0.30	0.12	−2.98	0.003	0.14	0.66

^**Fat-free milk drinkers (n= 138) were not asked whether they would consider switching milk type

Log likelihood = −579.60179

Number of obs = 921

LR chi2(7) = 83.34

Prob > chi2 = 0.0000

Pseudo R2 = 0.0671

DISCUSSION

Consistent with prior research, our study found that the majority of participating shoppers typically consume higher fat milk, with a slightly stronger preference for 2% milk (37.1%) than whole milk (34.2%). In comparison, only 28.7% reported consuming any type of low-fat milk. Our prior research also found that customers typically purchase higher fat milk; more than 78% of customers reported most frequently purchasing whole (40%) or 2% (38%) milk in our 2015 study.¹⁶ Such findings are of particular concern in light of long-standing recommendations to consume low-fat (1%) and fat free milk.¹ The discordance between dietary recommendations and dietary behaviors, underscores the need for new or innovative ways to shift healthier product perceptions, norms and ultimately consumer behaviors.

Our results also show that while a critical attribute for purchase decisions,¹¹ most participants cannot accurately identify the type of milk they are drinking by taste alone, and while many are able to generally categorize a milk as a lower fat or higher fat variety, a substantial proportion of consumers believe that lower fat varieties of milk taste like higher fat content (2% or whole) milks; 32.6% of skim milk tastings were believed by participants to be 2% (20.7%) or whole milk (11.9%) while 37.9% of 1% fat milk tastings were identified as 2% and 23.6% as whole milk. Indeed, like in prior studies, fewer than 10% of customers are able to correctly identify all three samples of milk, and on average only one-third of consumers could correctly identify any one type of milk. Even fat-free milk, which has been reported as tasting distinctly “watery,” was correctly identified by only 34.2% of participants.

Following a milk taste-test experience, 40.6% of participants indicated they would consider purchasing a different type of milk, and younger females and those who typically purchased whole milk were most likely to consider a switch. Thus, the overarching goal of our study – to determine whether shoppers in low-income neighborhoods would consider purchasing lower-fat milk after participating in a blind test test – was achieved with a strong favorable main finding. Our results are aligned with prior research by Bakke et al. that found that milk taste-tests hold promise for testing consumer beliefs, particularly if they prefer higher fat milks.⁷ However, as the study also reveals, consumers who typically consume full fat milk also prefer higher fat content milks, suggesting that consumer acceptance may be bound to more modest shifts in fat content for higher fat consumers.

The results of this study provide strong evidence for the idea that milk taste testing can be an effective tool to help convince individuals to consider selecting lower-fat milk options, especially in areas where many are still drinking higher fat milk, and at higher risk of obesity and heart disease. This approach is easily scalable and a relatively inexpensive activity, with well-documented procedures. It can be incorporated into nutrition outreach to low-income populations in collaboration with SNAP-Ed, WIC, and local Dairy Council units, thus reaching many of the tens of millions of people who receive food assistance. Through these approaches, the taste-testings will reach the intended population in the supermarket, near the point of purchase.

At the same time, the context of availability of low-fat milk is an important consideration, particularly in lower income areas. Prior research has shown that stores in lower income communities, and smaller stores, such as corner stores and bodegas, are less likely to carry fat-free milk as compared to grocery stores and stores located in more affluent communities.¹⁰ As such, future intervention efforts which may build on our research should seek to ensure adequate access to low-fat milk.

The study reported here has some limitations. The convenience sample, while substantially larger than samples examined in earlier studies and recruited at stores in low-income neighborhoods, is limited to those customers shopping at the study stores, and while race/ethnicity and gender were captured, other characteristics like income, participation in WIC or SNAP and family size and composition were not. While the chain supermarkets in this study were willing to make lower fat milk more prominent, this may not be the case in some smaller groceries or corner stores. Further, this study only examined the ability of the participants to correctly identify the type of milk tasted, and was not a ranking task that examined how, or if, participants were able to generally discriminate between higher or lower fat levels; and the findings should not be interpreted as such. The study is based on a small tasting sample of milk (3 ounces) which was consumed without any option of pairing with any other products, such as cereal. It is possible that participants could have been better able to correctly identify the milk-fat content in a context more similar to how milk consumption occurs at home (i.e. with cereal, ice cold, in coffee, while baking, etc.) Participants were also allowed to taste any of the three milk types in any order they preferred, and were not provided with the samples in a random order. Instead, customers were free to taste one sample, taste a second sample of a different type of milk and return to re-taste the first, changing their decision until they were comfortable with their determinations. Further, our study is limited in that it did not track actual purchases but instead relied on consumer reports of the types of milk most often consumed, as well as willingness to purchase a different type of milk, rather than actual changes in purchases. “Willingness to switch” is not the equivalent of intention to switch, but is only one step toward considering a change in the type of milk purchased. Further research is needed to determine whether consumers actually change their purchasing and consumption patterns after taking part in a blind taste-testing.

We recognize that milk consumption can be a controversial topic, embedded in a context of concern for environmental effects related to production and processing, as well as antibiotic resistance. ²⁰ Our study did not seek to address such concerns.

IMPLICATIONS FOR RESEARCH AND PRACTICE

Our study offers new information about the value of taste testing in creating dietary behavior change. Findings may also have implications for other healthy foods whereby dispelling myths related to taste, through a blind rating which includes taste-testing, may enable consumers to reconsider perceptions and product purchasing. When it comes to shifting consumption patterns to lower-fat milk purchases our data suggest that marketing efforts which targeting younger populations including mothers with children, and those predominately consuming whole milk may result in the greatest change in purchasing lower-fat milks. This approach is easily scalable and a relatively inexpensive activity, with well-documented procedures. It can be incorporated into nutrition outreach in collaboration with SNAP-Ed, WIC, and local Dairy Council units.

It is important to note that this research is nested in the context of current debates about the benefits of low-fat milk as compared to full-fat milk and dairy foods in general.²⁰ Recent findings suggest that the issue is complex and more data are needed to understand how dairy consumption in general, and low-fat and full fat dairy consumption in particular, intersect with other dietary consumption habits and cultural patterns.²¹