Due to ginger holds a special and indispensable place in Chinese cuisine, understanding consumers’ preferences for organic ginger is of significance, especially given the growing interest in organic food products and sustainable agriculture. This study thus examines Chinese consumers’ preference for fresh ginger and the sources of their preferences heterogeneity for organic ginger consumption.
The study is using choice experiment (CE) method and mixed logit (MXL) modeling with 1,312 valid samples. The participants are regular consumers who are 18 years old or above and had bought fresh ginger within the past 12 months.
The results show that consumers prefer organic product certification labeling ginger to conventional ginger, preferred to purchase ginger at wet markets to at supermarkets or online, and preferred either ginger with regional public brand or private brand to unbranded ginger. Results also indicate that age, education level, income, purchasing experience of organic and branded ginger, and cognition of ginger health benefits are the sources of heterogeneity in consumer preferences for organic ginger.
This study contributes to ginger growers, marketers and policy makers. This study tracks how consumers' preferences change under different attribute combinations, capture the complex preference structure of consumers, and help reveal the motivations behind consumers' preferences for organic ginger. These findings will be crucial for developing marketing strategies, promoting organic products, and meeting consumer needs.
1. Introduction
China is the largest consumer of ginger in the world. According to data from the China Bureau of Statistics and China Customs, the gross domestic consumption for ginger in China in 2021 was around 11.363 million tons (https://www.chinairn.com/news/20221211/102528436.shtml, accessed on 11 December 2022). Due to its versatile applications and the appreciation of its unique flavor and potential health benefits, ginger has become an indispensable ingredient in Chinese culinary traditions. Although ginger has a small proportion in Chinese consumers' total vegetable consumption, it is a hardly substitutable seasoning vegetable on the dining table in most Chinese households. Unlike consumers in other countries who prefer processed products, Chinese consumers prefer raw and fresh ginger. In the daily diet of Chinese people, fresh ginger consumption accounts for about 70% of China’s total ginger consumption.
In recent years, rising incomes and living standards have driven consumer demand for organic food (Britum and Yannaka, 2019; Ahmed et al., 2020). China is one of the countries with the fastest growth in the organic food consumption market. According to statistics, the organic food consumption market in China is growing at an annual rate of 25% (http://www.ce.cn/xwzx/gnsz/gdxw/202111/07/t20211107_37066202.shtml, accessed on 7 November 2021). Consumers are concerned that conventional methods of planting ginger may have excessive pesticide residues, so consumers' demand for organic ginger is also increasing (Han et al., 2021). What attributes of fresh ginger do consumers have preferences for? What are the differences in consumers’ preferences for organic ginger compared to traditional ginger? What factors influence the consumption of organic ginger? This study will focus on consumer preferences for organic ginger and answer the above questions.
Few studies have studied on consumers’ preferences for organic fresh ginger, and no further research has been conducted on the heterogeneity of consumer preferences for organic ginger. Therefore this study will analyze Chinese consumers’ preferences for fresh ginger, and further explore the heterogeneity of organic ginger consumption preferences and examine whether factors such as consumers’socio-demographic characteristics, purchasing experience and habits, and awareness of organic products are the sources of consumer preference heterogeneity. We adopt the choice experiment (CE) method to design alternative options and choice sets composed of four attribute variables: price, certification label, purchase place, and brand. We use a mixed logit (MXL) model to empirically analyze the preferences of Chinese consumers for different combinations of ginger. The use of CE and MXL model can track how consumers' preferences change under different attribute combinations, capture the complex preference structure of consumers, and help reveal the motivations behind consumers' preferences for organic ginger.
This study contributes to ginger growers, marketers, and policymakers. This knowledge can help growers improve product market competitiveness, adjust cultivation methods, and adopt organic farming techniques to meet consumer needs and achieve better economic benefits. For policymakers, understanding consumer preferences can inform decisions on regulations, incentives, and support for organic ginger production, ultimately promoting sustainable agricultural practices and meeting the growing demand for organic products.
2. Literature review
Research on consumer preferences for organic food has gained significant interest. Several studies have identified key motivations driving consumers to choose organic food. Health consciousness is a primary factor, as consumers are motivated by concerns over the potential health risks associated with pesticides and chemicals in conventionally grown produce, with organic options perceived as safer (Hempel and Hamm, 2016; Ranjbarshamsi et al., 2016; Ceylan et al., 2018). Ethical considerations also play a role, with consumers feeling a sense of moral responsibility to support environmentally friendly products (McEachern and McClean, 2002; Cho and Krasser, 2011; Prakash et al., 2023). Economic development and environmental awareness in developing countries have further influenced consumer attitudes towards organic food (Yang et al., 2023). Additionally, consumers are attracted to organic products for their perceived natural growth, retention of original flavors, and perceived higher quality (Liang et al., 2021; Bernard and Liu, 2017; Nadricka et al., 2020).
Studies have examined the influence of demographic and socio-economic factors on consumer behavior regarding organic food consumption. Age, education, income, and health awareness are key factors. There are differing perspectives on the impact of age on organic product purchases, with some suggesting a positive correlation with age (Sangkumchaliang and Huang, 2012) while others propose a negative relationship (Byrne et al., 1991; Hamide et al., 2016). Education is another variable, with higher education levels often associated with greater awareness and likelihood of purchasing organic food (Chung, 1996; Liu and Zheng, 2019; Li et al., 2021). Income is also a significant factor, with higher-income households generally showing a greater inclination towards buying organic products (Gracia et al., 2007; Smith et al., 2009; Su-Huey and Andrew, 2009; Dimitri and Dettmann, 2012; Ceylan et al., 2018). Other demographic factors such as marital status, gender, ethnicity, presence of children in the family, and urban or rural residence are also considered in the literature (Wee et al., 2014; Akgüngör et al., 2010; Mehree, 2015). Additionally, researchers are exploring psychological variables such as emotions, morality, and self-identity perception among organic food consumers. Some consumers view purchasing organic products as a lifestyle choice and a reflection of their moral values, emphasizing concepts like ecologism, hedonism, benevolence, and self-discipline in describing characteristics of organic food consumers (Arvola et al., 2008; Robinson and Smith, 2002; Aertsens et al., 2009; Kaur et al., 2023).
The features of food itself can significantly impact consumer behavior Caswell et al. (2002) classify food features into intrinsic and extrinsic features, mention that intrinsic features include safety, nutrition, sensory experience, and functional attributes, and extrinsic features include quality detection indicator attributes and quality clue attributes. Extrinsic attributes such as price, brand, and place of origin are often used by consumers as clues to food quality (Veale et al., 2006). Moreover, extrinsic cures (such as price and country of origin) have a significant impact on consumers' product quality evaluation (Veale and Quester, 2009). Thus, many studies have studied the impact of extrinsic attributes such as price, origin, certification labels, and brand of organic food on consumer behaviors (Xie et al., 2016; Alphonce and Alfnes, 2012; John et al., 2019; Larceneux et al., 2012).
Due to ginger being a minor agricultural product with limited consumption, there is relatively limited research on ginger consumption. Akerele et al. (2017) examined ginger consumption patterns and factors including income, demographic characteristics, health food awareness, and electronic media influencing ginger purchase decisions among consumers in tertiary institutions in Nigeria. They found that awareness about the health benefits of ginger, electronic media as source of information, consumers’ concern for their health and previous gratifying experience were factors substantially influenced ginger purchase decision. Sendanayake et al. (2017) analyzed Sri Lankan consumers' preferences for the pungency, aroma and color and found consumers prefer the color of Chinese ginger, the spiciness of local ginger, and the aroma of Rangoon ginger. Amoah et al. (2022) found that Ghanaian consumers prefer dried ginger over fresh ginger, preferring ginger powder to sliced dried ginger. They also identified that a significant factor influencing consumer preferences is the intended use of ginger (for medicine or for spicing). Khoza et al. (2021) discovered that awareness of the health benefits of ginger, the cost of ginger and frequency of consuming ginger positively influence rural households’ WTP in South Africa. And their study also indicated that the females were more likely willing to pay for ginger than males. In addition to the aforementioned researches, there are also more studies focused on processed ginger products or foods containing ginger ingredients, such as ginger-garlic paste (Hude et al., 2021), ginger drinks (RM et al., 2012; Onuegbu et al., 2023), ginger milk chocolate (Prihantari et al., 2024), ginger beer (Nutakor et al., 2020), etc.
In summary, many existing studies have focused on high daily consumption or common organic foods, such as milk, apples, tomatoes, broccolis, meat, and fish that are commonly consumed and consumed in large quantities in daily life (Kim et al., 2018; Scozzafava et al., 2020; Skreli et al., 2017; Jürkenbeck et al., 2020; Rahman et al., 2021; Xie, 2016)，Some studies have also analyzed consumer preferences for conventional ginger (Akerele et al., 2017; Amoah et al., 2022), and analyzed consumer preferences for processed ginger products or products containing ginger ingredients (Njoya et al., 2020; Faiqoh et al., 2021; Onuegbu et al., 2023; Nutakor et al., 2020). However, few studies have studied on consumers’ preferences for organic fresh ginger, and no further research has been conducted on the heterogeneity of consumer preferences for organic ginger. Given the significant influence of consumer demand on organic agriculture, the growing potential of organic ginger production and consumption, and the lack of similar studies, the necessity and significance of conducting this study are apparent.
3. Materials and methods
3.1 Choice experiment design
A Choice experiment (CE) was designed to examine consumers’ willingness to purchase ginger under a given combination of attributes. CE method originated from Lancaster’s consumer theory and random utility theory.
Lancaster’s consumer theory is an economic theory that explains consumer decision-making by considering both the price of a product and its attributes. According to Lancaster, consumers not only consider the price of a product but also its characteristics when making purchasing decisions. Lancaster (1966) argues that consumers, when purchasing goods, take into account both the attributes and utility of the goods, and choose a combination of goods that maximizes utility within a limited budget. There are many factors that affect ginger consumption, but too many attributes cannot be put into attribute combinations in CE method. Setting too many attributes may conceal key influencing factors and also affect consumers' rational choices at short notice (Gao et al., 2010). Based on previous literature and pre-survey, we selected four attribute variables: price, certification label, purchase place, and brand, and determines the levels of each attribute (Table 1). These four attributes play important roles in consumer purchasing decisions. Price directly affects consumers' willingness and ability to purchase, certification labels convey information about the quality and safety of products, purchase places affect the convenience and availability of products, and brand reflects the visibility and reputation of products. Taking all the above considerations into account, we ultimately decide to include these four attributes in the CE design.
Ginger product attributes in the choice experiment
| Attribute | Level | Description |
|---|---|---|
| Price (Chinese yuan) | 4 | 5 yuan per 500 g 15 yuan per 500 g 25 yuan per 500 g 35 yuan per 500 g |
| Certification label | 3 | CGFC label COC label No certiffcation |
| Purchase place | 4 | Wet markets Supermarket JD.com (Online) Pinduoduo (Online) |
| Brand | 3 | Regional public brands for agricultural products Private brand of agricultural product enterprises No brand |
Note(s): In September 2023, 1 USD ≈ 7.286 RMB
Source(s): Authors’ own creation
Price has four levels: 5 yuan, 15 yuan, 25 yuan, and 35 yuan, which are selected to match the retail prices of 500g gingers. We used the lowest market price (RBM5.00/500g) of fresh ginger as the baseline and generated four price levels by increasing it by 10 yuan for each level. Certification label has three levels: China Green Food Certification (CGFC) label, China Organic Certification (COC) label and No certification. CGFC label and COC label are the most recognized certification labels by Chinese consumers. We use no certification as the baseline to verify whether consumers are willing to pay a higher premium for these two types of certified ginger. Purchase place has four levels, two of which are physical stores: wet markets and supermarkets, and two are online platforms: JD.com and Pinduoduo. By including both physical stores and online platforms as two purchasing channels, we can gain a more comprehensive understanding of consumer purchasing decisions and preferences in different purchasing environments. JD.com and Pinduoduo are major e-commerce platforms in China. JD.com is a self-operated platform that can ensure product and service quality and the prices of agricultural products are relatively high, while Pinduoduo is a third-party e-commerce platform that cannot guarantee the quality and services of agricultural products and the prices are relatively low. Brand has three levels: Regional public brand, Private brand and No brand. We use no brand as a baseline to test consumer preferences for regional authentication or brand recognition.
In our CE, there were a total of 144 (4*3*4*3) choice sets. We used the fractional factorial orthogonal method in SAS9.4 to design reasonable choice sets, and select the optimal choice set according to the D-efficiency principle. Ultimately, 11 independent choice sets were obtained. In each choice set, participants can either choose one of two given types of ginger composed of four attributes, or they can choose the “I would choose neither product” option. Including a “I would choose neither product” option helps make the shopping scenarios more realistic (Gao et al., 2016). One example of the choice sets is shown in Figure 1.
3.2 Econometric models
Random utility theory in economics models individual choice behavior when individuals select from unpredictable alternatives. It posits that individuals opt for the alternative offering the highest utility, reflecting their satisfaction or preference. McFadden (1973) introduce random utility theory in economics to analyze individual behavior of making choices among limited choice sets, transforming the choice problem into a comparison problem. We defined a random utility function for consumers to purchase ginger, and assume that consumers choose their preferred products based on the principle of utility maximization. The random utility function is as following:
where indicates the total utility when consumer i choose alternative j, indicates the deterministic utility based on the observable characters (attributes) of product, indicates the random error of the utility function based on unobservable characters (attributes).
Based on the principle of utility maximization, consumers will choose the alternative that brings them the maximum utility. That is to say, if consumer i chooses alternative j, it means that holds true for consumer i choosing any alternative k (k≠j). Therefore, the probability of alternative j being chosen by consumer i in a choice experiment is:
Assuming that the error term follows an independent and identically distributed (IID) distribution and belongs to the Gumbel distribution/Type 1 asymptotic extremum distribution, the probability of any given alternative being chosen can be represented by a multinomial logit (MNL) model:
In (3), the deterministic utility function generated in MNL model can be expressed in linear form:
where ASC is an alternative specific constant used to explain the impact of unobservable attributes on making choice, and represents the benchmark utility when consumers choose “I would choose neither product” option. indicates attribute variables if consumer i choosing alternative j, and is the corresponding coefficient to be estimated.
However MNL model needs to satisfy that the random error term strictly follows the IID assumption and the parameter value is fixed. This assumption implies treating consumers as a homogeneous group, and therefore the randomness of various consumer’s preferences cannot be expressed, the heterogeneity of choice preferences cannot be further analyzed either. The mixed logit (MXL)model overcomes these limitations of MNL. The MXL model assumes a random variable that follows a certain probability distribution, so it can reflect the randomness characteristics of individual preferences, which can be used for heterogeneity analysis of consumer’s preferences (Hensher and Greene, 2003). The MXL model is an integral formula of MNL models, which represents the probability of consumer i choosing alternative j as follows:
In MXL, as is a random variable, it can be represented as:
where is the coefficient of a random utility variable, which is a fixed value and can be the mean, is a random perturbation term, which can be the deviation of the mean (Hensher and Greene, 2003). If is assumed to follow a normal distribution, and can be considered as the mean and variance of the normal distribution. In this way, the observable utility in the formula (5) for consumers i to choose alternative j can be expressed in a simple linear form as followings:
In order to identify the sources of consumer heterogeneity preferences, we introduce the interaction term between consumer’s demographic and socio-economic characteristics variable and the chosen attribute , which is simply expressed as:
where, is the interaction coefficient of and .
3.3 Data collection
From November 23 to December 6 of 2022, a survey was conducted in six cities: Beijing, Shanghai, Xi’an, Qingdao, Guangzhou, and Harbin. These six cities represent different geographical regions, economic development levels, and cultural characteristics of China. These cities are located in the northern, eastern, southern, and western regions of China, covering different climate environments, population sizes, and economic activities. Beijing in Northern China is the capital city and the political center of China. Shanghai is an Eastern coastal city, the largest economic center and important global financial center. Guangzhou is the capital of Guangdong Province and a major financial and international commercial center in southern China. Xi’an in western China, the capital city of Shanxi province, is an important central city in western China. Harbin is the capital city of Heilongjiang Province and is also the economic and cultural center of northern China. Qingdao is a famous coastal city in northern China and one of the economic centers of Shandong Province with strong economic strength. In our study, we conducted random sampling of urban consumers in the selected cities. In our study, we initially conducted a proportional sampling based on the population of the sample cities. After removing some invalid samples, we ended up with 1,312 valid samples, including 255 in Beijing, 266 in Shanghai, 222 in Guangzhou, 195 in Xi’an, 187 in Qingdao, and 187 in Harbin. At the beginning of the survey, two screening questions were used to ensure that all respondents were 18 years old or above and had purchased ginger within the past 12 months. Before the CE began, the definitions of all the attributes of the CE were clearly presented to participants. A cheap talk script was designed before conducting the CE to remind participants their budget constraints and ask them to make choices as if they were choosing products in real situations (Tonsor and Shupp, 2011; Gao et al., 2019). Once participants completed the CE, they answered a series of questions including daily ginger consumption habits, knowledge of organic products and ginger benefits, and socio-demographic characteristics information.
3.4 Characteristics of the sample
Among the 1,312 participants, the ratio of female was 57.24%, and the ratio of male was 42.76%. Regarding age, 34.60% of the participants were between 18 and 25 years old, and only 5.8% of respondents were aged 50 years old and above, indicating that the average age of the respondents is relatively young. Most participants were educated from 15 to 16 years or held a bachelor degree (59.68%), and education level below high school only accounted for 11.28%. As to income, the proportion of respondents with a personal monthly income between 5,000 and 10,000 yuan was the highest, at 60%. Comparing between respondents from the different cities, a majority of participants in Beijing, Shanghai, Guangzhou, and Xi’an were female. The proportion of different age range and the proportion of education level were roughly in the same order in six cities There are differences in personal monthly income among cities, with Beijing participants having the highest monthly income, with 45.09% of Beijing participants having a monthly income of over 10,000 yuan, followed by Shanghai (37.59%), Guangzhou (32.44%), Harbin (18.18%), Qingdao (14.98%), and Xi’an (13.34%) (Table 2).
Comparison of socio-demographic characteristics between survey participants (N = 1,312) in six cities in % of the sample
| Sample variables | Beijing (N = 255) | Shanghai (N = 266) | Guang Z (222) | Xi’an (195) | Harbin (187) | Qingdao (187) | Total (1,312) |
|---|---|---|---|---|---|---|---|
| Gender | |||||||
| Female | 58.43 | 62.03 | 64.86 | 54.36 | 50.27 | 49.73 | 57.24 |
| Male | 41.57 | 37.97 | 35.14 | 45.64 | 49.73 | 50.27 | 42.76 |
| Age (in years) | |||||||
| 18∼25 | 29.80 | 31.58 | 40.54 | 36.92 | 31.02 | 39.57 | 34.60 |
| 26∼30 | 22.75 | 18.42 | 26.58 | 16.92 | 17.65 | 14.44 | 19.74 |
| 31∼40 | 29.80 | 30.45 | 25.23 | 20.51 | 29.41 | 22.99 | 26.75 |
| 41∼50 | 11.76 | 14.66 | 5.86 | 19.49 | 13.90 | 13.90 | 13.11 |
| 51∼60 | 4.31 | 3.01 | 1.35 | 5.64 | 3.74 | 5.88 | 3.89 |
| >60 | 1.57 | 1.88 | 0.45 | 0.51 | 4.28 | 3.21 | 1.91 |
| Education level (in years) | |||||||
| ≤9 years | 0.39 | 2.26 | 2.25 | 3.59 | 4.28 | 8.56 | 3.28 |
| 10∼12 | 5.49 | 5.64 | 3.60 | 12.82 | 12.30 | 10.70 | 8.00 |
| 13∼14 | 7.84 | 10.90 | 18.02 | 25.64 | 21.39 | 25.67 | 17.30 |
| 15∼16 | 68.24 | 68.80 | 67.12 | 47.18 | 52.94 | 45.99 | 59.68 |
| >16 | 18.04 | 12.41 | 9.01 | 10.77 | 9.09 | 9.09 | 11.74 |
| Personal monthly income(in RMB yuan) | |||||||
| <2,000 | 8.63 | 9.77 | 9.01 | 21.03 | 18.18 | 20.86 | 13.87 |
| 2,001∼5,000 | 12.55 | 15.79 | 18.47 | 31.28 | 24.60 | 29.41 | 21.11 |
| 5,001–10,000 | 33.73 | 36.84 | 40.09 | 34.36 | 39.04 | 34.76 | 36.43 |
| 10,001–15,000 | 24.31 | 19.92 | 18.02 | 8.72 | 10.16 | 10.70 | 16.08 |
| 15,001–20,000 | 13.33 | 10.53 | 9.01 | 3.08 | 3.74 | 2.14 | 7.55 |
| >20,000 | 7.45 | 7.14 | 5.41 | 1.54 | 4.28 | 2.14 | 4.95 |
Source(s): Authors’ own creation
4. Empirical result
4.1 Regression results from empirical models
We used SAS9.4 software for regression. Due to our experiment having 11 choice sets, each with 3 options, and 1,312 valid questionnaires, a total of 43,296 (1,312 × 11 × 3 = 43,296) valid observations were obtained. We adopted a MXL model for analysis. When we set organic authentication as a random parameter variable and assume a normal distribution, and set other variables as fixed variables simultaneously, the mean and standard deviation of the random variables, as well as the estimated values of the fixed variables coefficients, are the most significant, and the model has the best fitting effect. The regression results are exhibited in Table 3.
Estimation results of the mixed logit model (model 1)
| Variables/Attributes | Coefficient | Standard error | |
|---|---|---|---|
| Random variable | |||
| COC label | Mean | 1.9085*** | 0.0603 |
| Standard Deviation | 1.2882*** | 0.1823 | |
| Fixed variables | |||
| Price | −0.0427*** | 0.002485 | |
| CGFC label | 1.7217*** | 0.0370 | |
| Supermarket | −0.1553*** | 0.0540 | |
| JD.com (Online) | −0.1359** | 0.0600 | |
| Pinduoduo (Online) | −0.8425*** | 0.0542 | |
| Regional public brands for agricultural products | 0.5948*** | 0.0671 | |
| Private brand of agricultural product enterprises | 0.4698*** | 0.0405 | |
| ASC | −1.1512*** | 0.0621 | |
| No. of respondents | 1,312 | ||
| No. of observations | 43,296 | ||
| McFadden’s LRI | 0.3552 | ||
| Adjusted Estrella | 0.6188 | ||
| Log Likelihood | −10223 | ||
Note(s): ① ***, ** and* indicate significance at 1, 5 and 10% significance levels, respectively (the same below)
②Price attribute is the actual price, the base of Certification label attribute is “No certification”, the base of Purchase place attribute is “Wet markets”, and the base of Brand attribute is “No brand”
Source(s): Authors’ own creation
The results of MXL model 1 (Table 3) show that the McFadden LRI value is 0.3552, indicating that the overall fit of this model is good (usually a value between 0.2 and 0.4 is considered a very good fit result). The standard deviation coefficient of the random parameter variable COC label is biggish (1.2882) and significant at the 1% level, which indicates heterogeneity in participants' preferences for COC label. The followings are the analysis of the preferences of participants for each attribute:
The coefficient of the price attribute is negative (−0.0427), which is significant at the 0.01 level, indicating that consumers' utility will decrease as prices rise. This result aligns with the fundamental principle of economics that an increase in price leads to a decrease in demand.
The coefficients of CGFC label and COC label are both significant at the 1% level, indicating that certification attributes are the important factor affecting consumer preferences. Both of these two coefficients are positive (1.72 and 1.91, respectively), indicating that compared to “No certification”, both CGFC label and COC label increase consumer’s utility. Moreover, the higher coefficient for the COC label indicates that organic certification can further increase consumer utility.
The coefficients of supermarkets, JD.com, and Pinduoduo are all significantly negative (−0.1553, −0.1359, and −0.8425, respectively), indicating that the traditional wet market is still the first choice for consumers to buy ginger, and supermarkets, JD.com, and Pinduoduo cannot improve consumer’s utility. Further comparison between online and offline channels respectively. As to offline channels, consumers prefer to purchase at the wet markets to supermarkets, which is in line with the vegetable shopping habits of most Chinese people. As to online channels, consumers prefer JD.com to Pinduoduo and may trust the quality of JD’s products more. Specially, the absolute value of Pinduoduo’s coefficient is the highest, indicating that consumers are the least likely to purchase ginger from Pinduoduo.
The coefficients of regional public and private brands are both significantly positive (0.5948 and 0.4698, respectively), indicating that compared to no brand, either ginger with regional public brand or private brand can increase consumer utility, and the utility of regional public brands is greater. This result is consistent with the fact that consumers have more trust in branded agricultural products. Consumer preference for regional public and private brands indicates that consumers are more inclined to purchase products with brands, especially those with regional public brands.
Finally, the coefficient of the term ASC is significantly negative (−1.1512) at the 1% level, indicating that the given choice sets are attractive to most participants.
4.2 Analysis of preference heterogeneity in organic ginger
In order to further analyze the sources of preference heterogeneity in organic ginger, we continue to establish 3 more MXL models (Model 2, Model 3, and Model 4). The regression results are exhibited in Table 4.
Estimation results of the mixed logit model (model 2, model 3 and model 4)
| Variables/Attributes | Model 2 | Model 3 | Model 4 | |
|---|---|---|---|---|
| Random variable | ||||
| COC label | Mean | 1.7231*** | 1.6547*** | 1.2610*** |
| Standard Deviation | 0.0319 | 1.2475*** | 0.5801* | |
| Fixed variables | ||||
| Price | −0.0387*** | −0.0426*** | −0.0387*** | |
| CGFC label | 1.6619 | 1.7215*** | 1.6539*** | |
| Supermarket | −0.0638 | −0.1551** | −0.0566 | |
| JD.com (Online) | −0.1736*** | −0.1362* | −0.1704 *** | |
| Pinduoduo (Online) | −0.7639*** | −0.8427*** | −0.7513*** | |
| Regional public brands for agricultural products | 0.4042*** | 0.5943*** | 0.4081*** | |
| Private brand of agricultural product enterprises | 0.4586*** | 0.4697*** | 0.4661*** | |
| ASC | −1.0950*** | −1.1501 *** | −1.0949*** | |
| Interaction terms | ||||
| Gender × COC label | −0.0424 | |||
| Age × COC label | −0.1314*** | |||
| Education × COC label | 0.0754** | |||
| Income × COC label | 0.0688*** | |||
| Whether often buy vegetables or not? × COC label | −0.0734 | |||
| How often do you buy vegetables a week? × COC label | −0.0260 | |||
| Have you ever bought organic ginger? × COC label | 0.3763*** | |||
| Have you ever bought branded ginger? × COC label | 0.3329*** | |||
| Whether know about food safety certification? × COC label | 1.05 | |||
| Whether know about health benefits of ginger? × COC label | 1.25 | |||
| Whether to increase purchases after knowing the health benefits of ginger ? × COC label | 5.31*** | |||
| No. of respondents | 1,312 | 1,312 | 1,312 | |
| No. of observations | 43,296 | 43,296 | 43,296 | |
| McFadden’s LRI | 0.3566 | 0.3582 | 0.3564 | |
| Adjusted Estrella | 0.6205 | 0.6215 | 0.6194 | |
| Log Likelihood | −10,202 | −10,176 | −10,204 | |
Source(s): Authors’ own creation
The results of Model 2 show that interaction coefficients between age and COC label, education and COC label, income and COC label are significant, indicating that age, education level, and income are the sources of consumer’s preference heterogeneity. In detail, note that with a positive coefficient (1.9085) for the COC label, the interaction coefficients between age and COC label is negative (−0.1314) and significant, implying that older consumers have lower utility in consuming organic ginger. The interaction coefficients between education and COC label, income and COC label are both significantly positive (0.0688, 0.0754), implying that consumers with higher education and higher incomes have higher utility in consuming organic ginger.
The results of Model 3 show that interaction coefficients between “Have you ever bought organic ginger?” and COC label, “Have you ever bought branded ginger?” and COC label, are significantly positive (0.3763, 0.3329) at the significance level of 0.01. These results indicate that consumers who have previously purchased organic ginger or branded ginger are more likely to purchase ginger with organic certification labels.
The results of Model 4 show that interaction coefficients between “Whether to increase purchases after knowing the health benefits of ginger ?”and COC label is significantly positive (5.31) at the significance level of 0.01, indicating that consumers knowing the health benefits of ginger will increase their willingness to purchase organic ginger.
Based on the regression results of the above models 2, 3, and 4, it can be concluded that age, education level, income, purchasing experience of organic and branded ginger, and cognition of ginger health benefits are the sources of heterogeneity in consumer preferences for organic ginger with COC label.
4.3 Robust test
We have established four mixed logit models with and without interaction terms. By comparing the results of these four models, it can be seen that the variable coefficients and significance levels of each model are basically consistent, indicating that the results are relatively robust. Using a conditional logit model to test the empirical findings is a common method due to its simplicity, broad applicability, parameter interpretability, and ability to control for other variable effects. The regression results from conditional logit model are exhibited in Table 5. Compared with the regression results in Table 5, except for the coefficient of supermarket attribute which is not significant, the coefficients of other attributes are completely consistent in direction and significance. The regression results are relatively robust.
Estimation results of the conditional logit model (model 5)
| Variables/Attributes | Coefficient | Standard error |
|---|---|---|
| Price | −0.0386*** | 0.002171 |
| CGFC label | 1.6589*** | 0.0338 |
| COC label | 1.7380*** | 0.0397 |
| Supermarket | −0.0606 | 0.0428 |
| JD.com (Online) | −0.1745** | 0.0560 |
| Pinduoduo (Online) | −0.7609*** | 0.0440 |
| Regional public brands for agricultural products | 0.3977*** | 0.0422 |
| Private brand of agricultural product enterprises | 0.4579*** | 0.0362 |
| ASC | −1.0963*** | 0.0565 |
| No. of respondents | 1,312 | |
| No. of observations | 43,296 | |
| McFadden’s LRI | 0.3545 | |
| Adjusted Estrella | 0.6171 | |
| Log Likelihood | −10234 | |
Source(s): Authors’ own creation
In order to further test the robustness of the regression results, all samples are divided into two sub-samples based on gender, male group and female group. The MXL model is still used for regression according to formula (7), with the settings of random and fixed parameter variables consistent with the previous full sample analysis. The regression results in Table 6 indicate that, except for the insignificance of the coefficients of the supermarket and JD.com attributes in the male sample regression, the coefficients of the mean of random and fixed parameter variables in each sub-sample are consistent in direction and significance with those of the full sample. The coefficient direction of the standard deviation of the random parameter variables is also consistent, confirming that gender does not drive preference heterogeneity. Overall, the regression results are robust.
Estimation results of the mixed logit model based on gender (model 6)
| Female | Male | ||||
|---|---|---|---|---|---|
| Variables/Attributes | Coefficient | Standard error | Coefficient | Standard error | |
| Random variable | |||||
| COC label | M | 2.0356*** | 0.0858 | 1.7672*** | 0.0841 |
| S.D. | 1.3906*** | 0.2437 | 1.1800*** | 0.2688 | |
| Fixed variables | |||||
| Price | −0.0521*** | 0.003437 | −0.0302*** | 0.003659 | |
| CGFC label | 1.7241*** | 0.0498 | 1.7270*** | 0.0554 | |
| Supermarket | −0.2192*** | 0.0766 | −0.0947 | 0.0767 | |
| JD.com (Online) | −0.1668** | 0.0804 | −0.0953 | 0.0906 | |
| Pinduoduo (Online) | −0.9504*** | 0.0767 | −0.7205*** | 0.0766 | |
| Regional public brands for agricultural products | 0.7005*** | 0.0976 | 0.4852*** | 0.0907 | |
| Private brand of agricultural product enterprises | 0.5065*** | 0.0547 | 0.4323*** | 0.0606 | |
| ASC | −1.2563*** | 0.0845 | −1.0166*** | 0.0937 | |
| No. of respondents | 751 | 561 | |||
| No. of observations | 24,783 | 18,513 | |||
| McFadden’s LRI | 0.3586 | 0.3555 | |||
| Adjusted Estrella | 0.6218 | 0.6173 | |||
| Log Likelihood | −5,821 | −4,370 | |||
Source(s): Authors’ own creation
5. Conclusion and discussion
This study examines Chinese consumers’ preference for fresh ginger, and further discusses their preferences heterogeneity for organic ginger consumption. We found that consumers preferred ginger with organic product certification labels to conventional ginger, preferred to purchase ginger at wet markets to at supermarkets or online, and preferred either ginger with regional public brand or private brand to unbranded ginger. We also found heterogeneity in consumer preferences for organic ginger. In detail, younger, higher-educated, and higher-income consumers tend to prefer organic ginger. This result may be due to, younger consumers are often more conscious of environmental and ethical concerns, leading them to choose organic products. Additionally, higher education and income levels may provide the means and knowledge to seek out and afford organic options. Consumers who have purchased branded or organic ginger before tend to prefer ginger with COC label, which may be due to these consumers may be more concerned about the quality and safety of their food and they believe organic certification labels can serve as a marker of quality and authenticity, as well as trust and transparency. Awareness of the health benefits of ginger may increase consumers' willingness to purchase organic ginger due to concerns about potential chemical residues, the desire for higher nutrient content, and a broader consideration of environmental and ethical factors. Overall, age, education level, income, purchasing behavior, health cognition are the sources of consumer’s preference heterogeneity for COC labeled organic ginger consumption. Understanding consumer preferences for organic ginger can encourage agricultural producers to better understand market demands, promote sustainable organic farming methods, and reduce negative impacts on the environment.
Our study provides essential information for both ginger growers and marketers in China. Our study indicates that consumers value COC labeled organic ginger. This can incentivize growers to expand organic cultivation, comply with certification standards, and meet consumer demand. The increased supply of organic ginger can enhance market competitiveness, promote organic agriculture, and contribute to the healthy development of the organic agricultural industry chain. Our study also suggest that consumers have a preference for branded ginger. For marketers, they can create a unique brand image for organic ginger through brand design, packaging, and promotional activities, highlighting its green, healthy, and environmentally friendly characteristics. Marketers can introduce the advantages and characteristics of organic ginger to consumers through advertising, promotional activities, online and offline promotion.
Our study also provides suggestions for policymakers. Firstly, with the increasing demand for organic ginger, the government can establish policies supporting organic ginger production, such as providing financial support, tax exemptions, setting up special funds, etc., to encourage agricultural producers to grow organic ginger. Secondly, enhance certification and regulation of organic ginger production, promote the popularization of organic ginger product certification labels to increase consumer trust in product quality and credibility. Thirdly, offer organic ginger cultivation technology training and support to help agricultural producers enhance production techniques and management skills. Finally, support the establishment and promotion of regional public brands for organic ginger agricultural products to enhance the regional brand awareness and competitiveness of organic ginger.
However, this study has some limitations that can inspire future research. First, we have only studied the impacts of four attributes including price, certification label, purchase place and brand. There must be other attributes such as place of origin, appearance, traceability information and freshness that affect ginger consumption. Hence, future studies should consider these attributes, which may lead to a deeper understanding of consumer purchasing decisions and preferences for ginger. Second, our sample is limited to consumers from six cities of China. Ginger is widely consumed around the world due to its versatile applications and the appreciation of its unique flavor and potential health benefits. Future research should be conducted comprehensive studies that examine ginger consumption patterns and differences across diverse regions and cultures.
The authors are grateful to Jikun Huang for giving us suggestions on the study design.
Funding: This study has been funded by Projects of Shandong Social Science Planning (grant number: 21CDCJ16) and Shandong Provincial Party School Innovation Project, and Shandong Social Science Planning (grant number: 23CKFJ27).
Competing interests: The authors declare that they have no competing interests.
Authors contributions: WL, XH and ZG conceptualized and coordinated the study including the attribute selection and level. WL and XH designed the choice experiment. XH and XC supervised the survey and edited, and cleaned the data. XH wrote and edited the manuscript and ZG reviewed it. WL performed data analyses. All authors read and approved the final manuscript.
Availability of supporting data: The data that support the findings of this study can be obtained from the corresponding author upon request.
