Improving the public's willingness to purchase near-expired food to reduce food waste: The case of milk products in China

Cheng, Shujun; Shi, Xuanhao; Ren, Yanjun; Zhao, Minjuan Article — Published Version Improving the public's willingness to purchase nearexpired food to reduce food waste: The case of milk products in China Agricultural Economics (Zemědělská ekonomika) Provided in Cooperation with: Leibniz Institute of Agricultural Development in Transition Economies (IAMO), Halle (Saale) Suggested Citation: Cheng, Shujun; Shi, Xuanhao; Ren, Yanjun; Zhao, Minjuan (2025) : Improving the public's willingness to purchase near-expired food to reduce food waste: The case of milk products in China, Agricultural Economics (Zemědělská ekonomika), ISSN 1805-9295, Czech Academy of Agricultural Sciences, Prague, Vol. 71, Iss. 2, pp. 86-98, https://doi.org/10.17221/166/2024-AGRICECON , https://agricecon.agriculturejournals.cz/artkey/age-202502-0002_improving-the-public-swillingness-to-purchase-near-expired-food-to-reduce-food-waste-the-case-of-milk-product.php This Version is available at: https://hdl.handle.net/10419/312436 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ 86 Original Paper Agricultural Economics – Czech, 71, 2025 (2): 86–98 https://doi.org/10.17221/166/2024-AGRICECON Supported bythe National Natural Science Foundation ofChina (Grant Nos.72173097, 72373117), Key Special Funds ofthe Ministry ofAgriculture and Ministry ofFinance (Grant No.CARS-07-F-1), and the Northwest A&F University (Grant No.JGYJSCXXM202302). © Theauthors. This work islicensed under aCreative Commons Attribution-NonCommercial 4.0 International (CC BY-NC4.0). With the rapid development of the economy and urbanization, global food waste has been further aggravated (Thyberg and Tonjes 2016), and near-expired food (NEF) waste has become an important component. According tothe Food and Agriculture Organization ofthe United Nations (FAO), about one-third of the food (about 1.3 billion tons) is lost or wasted during production and consumption each year globally Improving the public's willingness topurchase nearexpired food toreduce food waste: The case ofmilk products inChina Shujun Cheng1, Xuanhao Shi1, Yanjun Ren1,2, Minjuan Zhao1,3* 1College ofEconomics and Management, Northwest A&F University, Yangling, P.R. China 2Department ofAgricultural Markets, Leibniz Institute ofAgricultural Development inTransition Economies (IAMO), Halle (Saale), Germany 3College ofEconomics, Xi'an University ofFinance and Economics, Xi'an, P.R. China *Corresponding author: [email protected] Shujun Cheng and Xuanhao Shi contributed equally tothis work Citation: Cheng S., Shi X., Ren Y., Zhao M. (2025): Improving the public's willingness topurchase near-expired food toreduce food waste: The case ofmilk products inChina. Agric. Econ. – Czech, 71: 86–98. Abstract: The near-expired food (NEF) isasignificant opportunity toreduce food waste, while consumers often associate NEF with safety issues, which results inalarge amount ofsafe and healthy food being wasted globally. This research focuses onfood date labelling (FDL) and explores how consumers' label cognition impacts their willingness topurchase NEF. Using arandom sampling method online, weobtain 2113valid samples from China and conduct aninformation intervention 'quasi-natural experiment' toobtain participants' FDL cognition and willingness topurchase the near-expired milk (NEM) before and after the intervention and evaluate the impact ofthe intervention through thedifferences-in-differences model. The results show that consumers' initial purchase willingness for NEM islow, and their FDL cognition has apositive effect, especially inEastern China and higher education consumers. Information intervention increases consumers' willingness topurchase NEM bychanging their label cognition, and the intervention has amore pronounced impact among older, male, and higher education consumers. Considering thepressure onresources andthe environment caused byfood waste has become animpediment tosustainable development, thefindings expand theapplication ofthe Knowledge, Attitude, and Practice (K-A-P) theory inthe NEF field and clearly reveal the important role ofeliminating consumer prejudice ofFDL inreducing food waste toachieve the United Nations Sustainable Development Goal 12.3 'Halve food waste.' Keywords: difference-in-differences; experimental economics; food date labelling; food industry; food waste 87 Agricultural Economics – Czech, 71, 2025 (2): 86–98 Original Paper https://doi.org/10.17221/166/2024-AGRICECON (Gustavssonetal. 2011), threatening global food security (Kuiper and Cui 2021). Atthe retail and consumer stages, alarge amount ofedible food isdiscarded (Melbye etal. 2017). About 28% offoods inAustria go unsold because they pass the best-before oruse-by date (Lebersorger and Schneider 2014). Inthe United States, more than 90% ofAmericans throw away food prematurely because they regard date labels asfood safety indicators ( 2012), which brings about 20% offood waste to the consumer stage (FDA 2019). At present, only afew developed economies have paid attention tothis issue, and the current status ofintransition economies isstill unknown. The consumption ofsuboptimal food isavital opportunity toimprove the utilization efficiency ofresources and reduce food waste (Bai etal. 2022), especially the NEF (Zhang etal. 2023). NEF refers tofood about toreach the restriction date onfood packaging (still within the date). However, consumers are reluctant topurchase the NEF (Aschemann-Witzel etal. 2018). Although the NEF can reduce food enterprises' turnover and management costs and enable consumers tobenefit from lower prices, 'Bad quality' isconsumers' inherent impression ofthe NEF (Xu 2013). Nearing expiration date makes consumers feel that food safety will not beuptotheir expectations (Aschemann-Witzel etal. 2018), and existing research points out that 69% to 84% of consumers believe that the quality ofperishable products decreases over time (Tsiros and Heilman 2005). Inaddition, price is an external cue, and product features (such asthe expiration date approaching) are aninternal cue, influencing consumers' willingness to purchase the NEF. Although reducing the price can increase the market attractiveness and isthe most direct way toimprove consumers' purchase willingness of NEF, they often infer the quality level based onthe price and reinforce the cognition ofthe declining quality (Aschemann-Witzel etal. 2017; Theotokis etal. 2012), while this price reduction may just aim toreduce pressure onthe stock. Food date labelling (FDL) isthe criterion for defining the near-expired state, while the definition ofNEF has yet tobeunified globally. InChina, NEF has yet to be unified nationwide, and only a few local governments have made preliminary regulations (see Table 1). FDL in China consists of manufacturing date and quality guaranteed duration, collectively referred toas'quality guaranteed date label,' which isanindicator offood quality. NEF isthe medium ofFDL, and consumers' understanding ofFDL mainly determines their willingness toconsume NEF. However, evidence from advanced economies suggests that many consumers donot know the difference between the various expressions ofFDL onthe market (Milne 2012), and they often misunderstand FDL asanindicator offood safety (Newsome et al. 2014; Verghese etal. 2015), directly leading tothe wrong cognition ofthe NEF and resulting a large amount of safe, healthy, and high-quality food being wasted (Hall-Phillips and Shah 2017). There isasevere information asymmetry inthe market's understanding and use ofFDL, and the FDL has animportant role inreducing food waste (Kavanaugh and Quinlan 2020). Food manufacturers use FDLtoindicate the quality and freshness offood products toconsumers, while consumers mistakenly believe that FDL isintended toconvey food safety information. Based onsuch information asymmetry, itseems understandable that consumers resist NEF for food safety to avoid the risk of foodborne diseases. The theory ofKnowledge, Attitude, and Practice (K-A-P) ismainly used toexplain the effects ofindividual knowledge and original beliefs on changes of individual behaviours, Table 1. The definition ofNEF inChina No. Duration ofquality guaranteed Criterion Example 1More than one year 45 cans, candy, cookies 2More than six months but less than one year 20 instant noodles, aseptically packaged milk and juice 3More than 90 days and less than six months 15 some vacuum-packed and refrigerated cooked food 4More than 30 days and less than 90 days 10 some sterilised packaged meat and fresh eggs 5More than 16 days and less than 30 days 5yogurt, some dessert 6Less than 15 days 1–4 unsterilised cooked food, boxed soy products The criterion column means 'X days before the quality guaranteed date' Source: According tothe different quality guaranteed duration ofdifferent foods, the NEF isclassified into six situations, which isthe most widely recognised standards of'The food date label critical' issued bythe Beijing Bureau ofIndustry and Commerce 88 Original Paper Agricultural Economics – Czech, 71, 2025 (2): 86–98 https://doi.org/10.17221/166/2024-AGRICECON and isthe theoretical basis for formulating behaviour intervention strategies (Pärna et al. 2005), which has been widely used in areas such as waste reduction (Liu etal. 2023) and adaptation toclimate change (Cai etal. 2024). This theory divides individual behaviour transformation into three stages: acquiring knowledge ofinformation, forming beliefs and changing behaviour. Specifically, 'Knowledge' refers tothe acquisition and understanding ofsome common-sense knowledge, 'Attitude' refers tothe correct consciousness and positive attitude toward something, and 'Practice' refers tothe transformation ofindividual behaviour after the acquisition and understanding ofknowledge and producing corresponding beliefs. Consumers will only change their existing consumption behaviour if they aware that itharms their health, and changing itcan reduce the risk (Contento and Murphy 1990). Based onthe above theoretical analysis, random intervention experiments in experimental economics can achieve the realization ofacertain stimulus toconsumers toachieve the change oftheir 'knowledge.' Byinforming the meaning ofthe FDL, weconduct the FDL information intervention onconsumers tochange their cognition ofNEF food tocomplete the cognitive change process ofthe first stage ('knowledge'). After the information intervention, this stimulus gives consumers anew cognition and understanding ofthe FDL and NEF ('Attitude') and is expected to influence their subsequent food consumption decisions ('Practice'). Afurther explanation iswhen consumers realize that NEF will not pose food safety orhealth hazards, the corresponding willingness toconsume NEF may increase. Many studies have focused onthe marketing methods ofsuboptimal food, mainly onthe impact ofprice discounts on consumers' purchase willingness and the role of information intervention in improving consumers' perceived quality. Onthe one hand, merchants attract consumers' attention todifferent types ofsuboptimal food bysetting price discounts tosatisfy their utility atalower price (Aschemann-Witzel etal. 2018). Onthe other hand, providing positive information about food while lowering prices can ensure consumers' perceived quality ofsuboptimal food and sell successfully. Aschemann-Witzel et al. (2019) use the cost-saving information label 'Lower price, save money' and the waste reduction label 'Do not waste food' tointerfere with consumers' purchase willingness for suboptimal food and find that both intervention information improves the perceived quality and purchase willingness for suboptimal food. Giesen and Hooge (2019) use 'Accepting defective food: Joining the fight against food waste' and 'It isanapple (carrot) inaspecial shape: Do not waste them' information interventions to change consumers' purchase willingness for foods with appearance defects, and they find that information works well. Although information intervention asaneffective tool has been widely used infood waste mitigation, the existing information intervention ismainly descriptive language (European Commission 2018 United States Environmental Protection Agency 2019; FAO 2020) and indicates product attributes (Aschemann-Witzel et al. 2019; Giesen and Hooge 2019; Schneider and Ghosh 2020) toinform consumers toreduce food waste. However, there isarelative lack ofstatistical analysis onsuboptimal food through non-price or non-declarative statement intervention methods. For NEF, the relationship between consumers' cognition ofFDL and their willingness toconsume NEF has not been empirically tested, and studies ofinformational interventions based onthe meaning ofFDL are rare. Inparticular, quasi-natural experiments that can simulate real consumption scenarios can beused toobtain more realistic data, while avoiding potential endogeneity problems. Notably, the existing studies are mostly from developed economies, and there isalack ofrelevant studies from transition economies. Our study aims toempirically test the relationship between consumers' cognition ofFDL and their NEF purchase willingness, and then through conducting FDL information intervention to reduce their misunderstandings about FDL to explore the influence ofinformation ontheir purchase willingness ofNEF. Finally, weattempt todiscern the heterogeneity among different groups and contribute new ideas to reduce food waste from transition economy. MATERIAL AND METHODS Sampling and data. Considering that consumers may be more sensitive to the safety of perishables, our research focuses on near-expired perishables, which is representative in the NEF research field. In this study, we take milk (perishable, dairy products), which has a large impact on the environment, arepresentative ofperishable food purchased byconsumers daily and also acommon focus infood waste research (Blondin et al. 2017; Connors and Schuelke 2022), asour research object.Weobtain the per capita milk consumption data from the last five years from the China Statistical Yearbook (The National Bureau of Statistics of China, 2016–2020) and selects six provincial capitals according to the rank 89 Agricultural Economics – Czech, 71, 2025 (2): 86–98 Original Paper https://doi.org/10.17221/166/2024-AGRICECON ofthe per capita milk consumption, per capita gross domestic product (GDP), and regional distribution characteristics. We conduct anonline questionnaire survey from the professional and well-known Chinese online survey platform, 'Questionnaire Star' (https://www.wjx.cn, a platform similar to Synata) in China in 2023 with random sampling. The 'Questionnaire Star' has over 6.2million registered users, including over three million active users, covering 31provincial administrative divisions inChina (excluding Hong Kong, Macao, and Taiwan). The survey company consistently provides qualified survey distribution and collection services through screening participants with anintegrity score system (Lin and Guan 2021). The process isvoluntary and participants can stop answering and exit anytime. Before the start ofour survey, wemake itclear toall participants that the questionnaire isentirely anonymous, the collected data will only beused for scientific research, and there isnoright orwrong inall questions. Participants must beatleast 18years old, and each participant isallowed tosubmit only once. Since the sample groups ofthe 'Questionnaire star' have their unique corresponding codes, the number of answers for each participant isrestricted strictly. Participants are randomly assigned toeither the intervention group orthe control group. When the number ofquestionnaires collected reaches the expected number, the questionnaire collection automatically stops. To ensure the quality of the online questionnaire, weset up multiple attention test questions and repeated questions in different places of the questionnaire toverify the participants' seriousness inanswering the questionnaire. Inaddition, weexclude samples with response times that are too long ortoo short toensure the intervention's effectiveness. After validity screening, weobtain 2209survey data from six cities, and the samples are evenly distributed among regions. Toavoid the influence caused bythe consumers' dietary preferences, wedelete the samples that donot consume milk and obtain 2113valid samples. The descriptive statistical characteristics ofthe samples are shown inTable2, which details the general individual and family characteristics ofour participants, and differences inplace ofresidence. Wealso illustrate the variations ofthe core variables in the study here. Only 17.56% of Chinese consumers understand the meaning ofthe FDL correctly initially, collectively referred toas'quality guaranteed date' in China, which means consumers' cognition oflabels isgenerally low, and the information transmission function ofFDL isseriously damaged. Inaddition, compared tomean4, consumers' initial purchase willingness for NEM isnegative. However, after the information intervention, the corresponding values change. Information intervention experimental design. Before the information intervention experiment, werequire the participants toidentify whether the quality guaranteed date label isaquality orasafety indicator toobtain their initial label cognition, initial purchase willingness for the NEM, and the corresponding demographic characteristics. Then we conduct an information intervention onthe meaning ofthe quality guaranteed date label for the intervention group, which displays the text 'Quality guaranteed date label – Aquality indicator indicating optimal food taste and flavour, which will decline beyond this date but isedible' (in Chinese). For the control group, the screen displays the following content: 'This page isintentionally left blank' (in Chinese). Finally, we obtain their cognition of the quality guaranteed date label and their willingness to consume the NEM again to construct the panel data before and after the information intervention experiment toidentify the effectiveness ofthe intervention. The whole process takes place onparticipants' phones orcomputer screens, and they cannot goback orrevise any questions they answered. Model setting. Firstly, weuse ordinary least square (OLS) regression toexplore the effects ofthe label cognition onconsumers' purchase willingness ofNEM, the specific formula inEquation (1): Yi = β0 + β1Xi + εi (1) where: Yi – the consumers' self-reported willingness (Likert seven-level scale, Y= 1, 2, 3, …, 7) topurchase NEM before receiving the information intervention; the index = 1, … I – the participant; Xi – contains 0–1variable ofthe label cognition that characterize whether the consumer answers the meaning ofquality guaranteed date label correctly and aset ofcontrol variables inTable2; εi – the constant term; β1 – the regression coefficient ofthe variable. Second, we use difference-in-differences (DID) regression toevaluate the net effect ofthe information intervention on consumers' willingness to purchase the NEM. Among the methods toevaluate the effect ofpolicy implementation, the difference-in-differences model isthe most prevalent and the oldest quasi-experimental research method widely used inrecent years (Feng etal. 2021; Wang and Ge 2022), which can not only test the effectiveness ofthe policy orintervention, 90 Original Paper Agricultural Economics – Czech, 71, 2025 (2): 86–98 https://doi.org/10.17221/166/2024-AGRICECON but also further identify the net effect ofthe impact. The basic idea ofthis method istoregard the implementation ofanew policy orintervention asa'natural experiment' or 'quasi-experiment' that is exogenous tothe economic system (Chen and Wu 2015). Onthe one hand, the implementation ofour information intervention may make the participants' label cognition and purchase willingness different before and after the intervention. Onthe other hand, the above two indicators may differ between the intervention and control groups at the same time. The DID model regression estimation can effectively control the impact ofother Table 2. Descriptive statistics Control variables Descriptions Definitions Mean SD Min. Max. Obs. Age the age ofthe participants year 30.24 7.44 18 69 2113 Gender gender ofparticipants male = 1; female = 0 0.42 0.49 0 1 2113 Education education level ofparticipants bachelor degree orabove = 1; below bachelor degree = 0 0.78 0.41 0 1 2113 Risk attitude risk attitude ofparticipants risk aversion = –1; risk neutral = 0; risk preference = 1 –0.45 0.85 –1 12113 Health state the health state ofparticipants' family members all healthy = 1; someone unhealthy = 0 0.76 0.43 0 1 2113 Disposable income per capita household disposable income USD/month 640.22 505.61 46.89 2816 2113 Family size the number ofthe participant's family members people 2.97 1.33 1 8 2113 Sample locations Eastern region Beijing city – – – – 365 Guangzhou city – – – – 361 Central region Hefei city – – – – 350 Changchun city – – – – 341 Western region Xining city – – – – 344 Chengdu city – – – – 352 Urban area 95.60% Core variables Descriptions Definitions Intervention group (N= 1 127) Control group (N= 986) mean SD mean SD Before intervention Label cognition whether the quality guaranteed date label isanswered correctly correct = 1; incorrect = 0 0.20 0.40 0.15 0.36 Purchase willingness I will buy NEM seven-Likert scale 3.95 1.71 3.44 1.72 After intervention Label cognition whether the quality guaranteed date label isanswered correctly correct = 1; incorrect = 0 0.62 0.49 0.14 0.34 Purchase willingness I will buy NEM seven-Likert scale 4.10 1.62 3.39 1.63 Seven-Likert scale: Strongly disagree = 1, disagree = 2, somewhat disagree = 3, general = 4, somewhat agree = 5, agree = 6, strongly agree = 7; wedivide six cities into three regions based onthe regional division method ofLietal. (2020); SD – standard deviation; Obs. – number of obeservations; NEM – near-expired milk Source: Authors' work 91 Agricultural Economics – Czech, 71, 2025 (2): 86–98 Original Paper https://doi.org/10.17221/166/2024-AGRICECON synchronous interventions and the prior difference between the intervention and control groups toidentify the net effect ofthe intervention (Wang and Ge 2022). Therefore, our information intervention isa'quasi-natural experiment,' and the intervention's impact can beevaluated using the DID method. See the following Equation (2) for model construction: Yit = β0 + β1treati + β2postt + β3treati × × postt + β4Xit + εit (2) where: Yit – the participant's willingness toconsume NEM before and after the information intervention; the index t = 0 and 1 – atime series ofdata; treati – adummy variable representing the group effect ofthe intervention group, and this dummy variable isextracted from the survey; postt – atime dummy variable, which captures the time effect ofthe intervention period; treati × postt – the real effect ofthe intervention group during the intervention period; Xit – participants' age, gender, education, risk attitude, per capita disposable income, family health status, and family size; β3 – our key interest, coefficient which represents the net effect ofthe intervention. RESULTS The correlation between label cognition and NEM purchase willingness OLS benchmark regression. Label cognition positively affects NEM's purchase willingness. Wemainly focus onthe impact ofinitial label cognition onconsumers' initial willingness to purchase NEM, and the results of OLS benchmark regression are shown inthe first column ofTable3. For one unit increase inlabel cognition, the corresponding NEM purchase willingness increases by 0.258, which confirms our concerns. The higher the label cognition, the more explicitly consumers know that the NEM issimply food that lasts intaste orflavour while still edible, bringing higher purchase willingness. Although wemainly concern about label cognition, westill have some interesting findings regarding control variables. Among the control variables, the age of participants significantly negatively affects their willingness topurchase NEM, which means that young consumers may have a higher acceptance of NEM. The relationship between consumers' purchase willingness and per capita household disposable income is reversed, possibly related to wealthier consumers' higher demand for fresher foods (Neff et al. 2015). Another possible reason isthat NEM isoften accompanied by sales discounts, which may be more attractive for consumers with lower disposable income tosave ondaily expenses and bring ahigher willingnesstopurchase. Robustness test. Toensure the robustness ofthe results, we test the robustness by changing dependent variables and expanding the sample. i) Changing dependent variable. Weuse the answer obtained by'I will buy NEM toreduce food waste' asasubstitute variable Y2 for the dependent variable for further regression analysis, and the regression results are shown in the second column ofTable3. The effect oflabel cognition onNEM purchase willingness still passes the signifiTable 3. The regression result ofthe correlation between label cognition and NEM purchase willingness Variables (1) (2) (3) Y1Y2Y1 Label cognition 0.258*** 0.167* 0.252*** (0.097) (0.096) (0.096) Age –0.021*** –0.011** –0.019*** (0.006) (0.005) (0.005) Gender –0.038 0.204*** –0.012 (0.077) (0.075) (0.075) Education 0.111 0.007 0.064 (0.097) (0.094) (0.094) Risk attitude 0.002 0.094** –0.016 (0.044) (0.043) (0.043) Disposable income –0.175*** –0.057 –0.172*** (0.049) (0.047) (0.048) Family size –0.054* 0.006 –0.051* (0.030) (0.029) (0.029) Health state –0.433*** –0.246*** –0.420*** (0.089) (0.087) (0.088) City controls YES YES YES Constant 6.099*** 4.682*** 5.987*** (0.353) (0.349) (0.348) R-squared 0.037 0.016 0.033 Observations 2113 2113 2209 *, **, *** P < 0.1, P < 0.05, and P < 0.01, respectively; robust t-statistics inparentheses; toavoid the influence ofmulticollinearity onthe results, weconducted the test, and the result ofmean VIF = 1.31 proves that there isnoobvious multicollinearity inour model; disposable income istreated with logarithm (lnx) inregressions; Y – the participant's willingness topurchase NEM; NEM – near-expired milk Source: Authors' work 92 Original Paper Agricultural Economics – Czech, 71, 2025 (2): 86–98 https://doi.org/10.17221/166/2024-AGRICECON cance test, which verifies the robustness. ii) Expanding the sample. Since wepre-delete the samples that donot consume milk inthe benchmark regression toexclude the influence ofthe food type weselect and the dietary differences onthe results, weexpand tothe whole sample 2209 and conduct the robustness test again. The regression results inthe third column ofTable3 verify the robustness again. Heterogeneity analysis. Considering that there are already some specialised near-expired food stores ineastern China, wewonder if there are significant area differences in consumer cognition and their willingness topurchase. The results inthe first three columns inTable4show that the correlation between label cognition and purchase willingness passes the significance test inthe eastern area, which may berelated tothe fact that consumers inthe east have more opportunities toget intouch with new things and relatively high purchase willingness for NEM. In addition, it isalso curious how the educational level ofconsumers affects this relationship. Wecompare the effect ofthis factor bygrouping regressions for different educational levels (see columns4 and5 inTable4 indetail), and the correlation between label cognition and purchase willingness passes the significance test atthe higher educationlevel. The effect ofinformation intervention Descriptive analysis: Differences between before and after the information intervention. Since label cognition significantly impacts consumers' willingness topurchase NEM, wefurther demonstrate the effect ofinformation onconsumers' label cognition and their purchase willingness through random information intervention experiments. The results show that 61.67% ofparticipants inthe intervention group correctly answers the meaning ofthe quality guaranteed date label after information intervention (with an increase Table 4. Heterogeneity analysis results (OLS model) Variables (1) (2) (3) (4) (5) Eastern China Central China Western China Lower education Higher education Label cognition 0.355** 0.129 0.265 0.196 0.276** (0.157) (0.168) (0.185) (0.216) (0.110) Age –0.022** –0.005 –0.032*** –0.017* –0.022*** (0.010) (0.009) (0.009) (0.009) (0.007) Gender 0.085 –0.050 –0.161 –0.013 –0.043 (0.130) (0.131) (0.140) (0.175) (0.085) Education –0.075 0.248 0.178 – – (0.169) (0.162) (0.173) Risk attitude –0.142* 0.018 0.118 0.156 –0.033 (0.074) (0.075) (0.081) (0.100) (0.049) Disposable income –0.153* –0.191** –0.166* –0.218** –0.162*** (0.081) (0.088) (0.085) (0.111) (0.055) Family size –0.026 –0.076 –0.057 –0.104* –0.040 (0.047) (0.054) (0.056) (0.061) (0.035) Health state –0.582*** –0.347** –0.329** –0.803*** –0.341*** (0.149) (0.158) (0.160) (0.213) (0.097) City controls –––YES YES Constant 5.862*** 5.347*** 5.922*** 6.935*** 5.981*** (0.547) (0.630) (0.573) (0.726) (0.418) R-squared 0.046 0.019 0.049 0.075 0.030 Observations 726 691 696 466 1647 *, **, *** P < 0.1, P < 0.05, and P < 0.01, respectively; robust t-statistics inparentheses; disposable income istreated with logarithm (lnx) inregressions; OLS – ordinary least square Source: Authors' work 93 Agricultural Economics – Czech, 71, 2025 (2): 86–98 Original Paper https://doi.org/10.17221/166/2024-AGRICECON of42.15%), while the cognitive changes inthe control group show adecrease, which means information intervention can effectively improve consumers' label cognition (Turvey etal. 2021). However, itisnot enough for us toexplore changes incognition, and wewould like toknow the further effects onpurchase willingness. Further, weuse the kernel density graphs toshow the changes inparticipants' NEM purchase willingness before and after information intervention inboth the intervention and control groups (see Figure1). Before the information intervention, 38.24% ofparticipants inthe interventiongroup has a negative attitude toward NEM. In contrast, after the intervention, the proportion drops to33.27%, andthe average willingness topurchase NEM increases from 3.95 to4.10, while inthe control group, this initial negative attitude isexacerbated. The net effect of information intervention. Although we can see the changes in label cognition and NEM purchase willingness before and after the information intervention from descriptive statistics, wefurther usethe econometric analysis DID model 1 2 3 4 5 6 7 The intervention group Purchase willingness to NEM Purchase willingness to NEM before after 0.05 0.10 0.15 0.20 0.25 Density before after 1 2 3 4 5 6 7 The control group 0.05 0.10 0.15 0.20 0.25 0.30 Figure 1. The kernel density graphs ofparticipants' purchase willingness toNEM before and after the information intervention NEM – near-expired milk Source: Authors' work Table 5. The net effect ofthe information intervention and robust regression results Variables (1) (2) (3) Y1PSM-DID Y2 DID 0.209** 0.220* 0.191* (0.102) (0.121) (0.103) Constant 5.320*** 5.368*** 4.594*** (0.248) (0.283) (0.247) Control variables YES YES YES Individual fixed effect YES YES YES Time fixed effect YES YES YES ATT – 8.30 – R-squared 0.062 0.062 0.031 Observations 4226 3378 4226 *, **, *** P < 0.1, P < 0.05, and P < 0.01, respectively; robust t-statistics inparentheses; disposable income istreated with logarithm (lnx) inregressions; Y – the participant's willingness topurchase NEM; NEM – near-expired milk; PSM – propensity score matching; DID – difference-indifferences; ATT – average treatment effect on the treated Source: Authors' work –30 –20 –10 0 10 20 Standardised % bias across covariates disposable income health state gender risk attitude education family size unmatched matched age Figure 2. Variations instandardised bias Source: Authors' work