What Reduces Household Food Waste in Japan? Nation-Wide and Region-Specific Contributing Factors in Urban and Rural Areas

Abstract

We analyze the contributing factors (i.e., attribute factors and behavioral factors) that lead to household food waste in Japan by region (i.e., large cities, small cities, and villages) using a nationwide sample and an ordered probit model. As a result of the analysis, it was found that “gender”, “age”, “children in the household”, “occupation”, “safety awareness”, and “time” were related to the occurrence of food waste in terms of consumer attributes, which were common throughout Japan. In terms of consumer behavior, “action”, “checking labels”, and “food management” are related. Among these, only “checking labels” had a positive relationship with the occurrence of food waste. By region, “children in the household”, “time”, “safety awareness”, and “food management” were particularly affected in large cities, while they were less affected in small cities and towns. For those who had “agricultural experience”, the occurrence of food waste was significantly lower in towns and villages.

1. Introduction

The issue of food waste has gained intense international interest in recent years. The UN Summit adopted “Transforming our World: the 2030 Agenda for Sustainable Development” [1] in 2015, setting “reduction of food loss and waste” as a goal, specifically mandating member countries to “halve per capita global food waste at the retail and consumer levels and reduce food losses along production and supply chains, including post-harvest losses” by 2030 (Target 12.3).

The declarations made at the G20 Summit held in Japan in 2019 also included a commitment to reduce food loss and waste. The G20 Agriculture Ministers’ Declaration (May 2019) [2] specified: “Productivity needs to increase and distribution needs to be more efficient, including by reducing food loss and waste, in order to achieve food security and improve nutrition for the growing world population”. Subsequently, “agricultural productivity needs to increase and distribution needs to be more efficient, including by reducing food loss and waste” was included in the G20 Osaka Leaders’ Declaration (June 2019) [3].

Regarding food waste around the world, according to Global Food Losses and Food Waste by the United Nations Food and Agriculture Organization [4], the total per capita food waste, including the production, retail, and consumption stages, is higher in developed countries than in developing countries, and especially high at the consumption stage.

This research focused on Japan, where the Ministry of Agriculture, Forestry and Fisheries estimated 25.3 million tons of food waste in FY2018, including inedible portions, of which 6.0 million tons (mt) was food waste, which is defined as food that could have been eaten but was thrown away (leftovers, unsold items, excess removal, direct disposal, etc.) [5]. This estimation divided food loss into business-related (3.24 mt) and household-related (2.84 mt) at the generation stage. Of the food waste including inedible portions, the share of food loss that is actually edible but thrown away is 18% for business-related and 36% for household-related. These numbers are quite high; hence, for Japan to achieve the international development goals, it is necessary to promote the reduction of food loss, more so for household-related loss.

The issue of food waste raises the question of how to define the term. Although different countries and institutions have different perceptions of this definition, we establish the definition in this study. First, there is the issue of the distinction between “food waste” and “food loss”. According to the FAO definition [6], we define "food loss" as that which occurs during production and processing, and “food waste” as that which occurs during retail and consumption. In this study, we use the term "food waste" to focus on households. There is also the question of whether to make a distinction between edible and inedible parts. In this study, we use the term “food waste” to mean “food that should have been eaten but was thrown away”. This is because in Japan, reduction targets have been set by focusing on the edible portion of food waste, and the data used in this study are also based on this definition.

Among studies on food waste in households, Canali et al. [7], Aschemann-Witzel et al. [8], and Nonomura [9] broadly surveyed research trends, while Roodhuyzen et al. [10] presented a list of contributing factors. In these studies, the contributing factors of household food waste are divided into behavioral factors (such as planning, purchasing behavior, and preservation behavior) and attribute factors (such as gender, age, income, children, and knowledge of best-before date).

Behavioral factors. People who do not plan well are expected to generate more food waste [8]. Similarly, people who shop less frequently are more likely to generate losses [11]. The relationship between checking food labeling and generating food waste is well established. Watson and Meah [12], for instance, conducted interviews and a record survey of actual waste in the United Kingdom to analyze when food is discarded. They pointed out that food labeling, in particular, has a significant effect. Van Boxstael et al. [13] conducted a questionnaire survey on knowledge about the best-before date and food disposal behavior in Belgium. Their survey indicates that consumers do not fully understand the difference between best-before dates and shelf life and that the food disposal based on best-before dates leads to food waste. Waitt and Phillips [14] interviewed people in Australia about food refrigeration and disposal. In deciding whether to dispose of food, the interviewees were influenced more by their subjective feelings toward the food than by an awareness about food waste or the labeling information. There are studies on the relationship between food management and the generation of food waste. Graham-Rowe et al. [15] interviewed consumers in the United Kingdom to show the need for proper food management technology to reduce food waste in households. Terpstra et al. [16] interviewed consumers in the Netherlands about food storage methods, period, and motivations for storage, and found that inappropriate storage management led to food waste, depending on the type of food.

Attribute factors. Families with children tend to generate losses [17]; high-income families, especially, tend to be busier and generate more food waste [18]. Numerous studies have analyzed the relationship between family structure and generation of food waste. Wenlock et al. [19] studied the amount of food wasted in households in the United Kingdom and found that food waste was greatly influenced by family structure, in that adults have a larger absolute quantity of food waste than children and larger households have less waste amount per capita than smaller households. Koivupuro et al. [20] conducted a monitoring study of households in Finland and found that household size and other factors were related to generation of food waste. On the relationship between lifestyle and generation of food waste, Parizeau et al. [21] studied the amount of food waste and the attributes and consciousness of households in Canada. They showed that household lifestyle is related to the amount of food waste produced. Evans [22] conducted interviews with households in the United Kingdom to show that food waste at households occurred incidentally as part of the daily eating behavior of consumers, which was influenced by social factors such as lifestyle and material factors such as technology. Williams et al. [23] have shown that consumers who have received environmental education are less likely to generate food waste. Bichop and Megicks [24] interviewed 48 participants in the United Kingdom to identify the relationship between respondents’ psychological factors and their behavior towards food waste. In their study, they found that awareness of the problem of food waste is unlikely to translate into actual action, and improvements in shopping and household habits are needed to sustain reduction behavior in the long term.

However, most of these studies focused on conceptual, normative, or individual contributing factors, and few quantitative empirical studies exist. A few studies comprehensively addressed and analyzed the contributing factors of food waste, one of which is McCarthy and Liu [25], who examined the factors that impacted the generation of food waste in Australia using a 346-strong sample and an ordered probit model; focusing mainly on attribute factors, they pointed out that having children and higher income had a positive effect on food waste. Min et al. [26] analyzed the developing economy in China using a fixed-effects model, focusing on dietary knowledge, and found that the effect of dietary knowledge on food waste differs between urban and rural areas, and between different income groups. Landry and Smith [27] used data from a survey of 2113 American households regarding food stock and household members’ food consumption to estimate the amount of food waste. In their study, they showed that food price, food expenditures, household structure, skills, tastes, and location are related to food waste. Yu and Jaenicke [28] estimated the amount of food waste for 4826 households in the US using data that recorded information about food, including household demographics and quantities acquired. In this study, they revealed that household food security, participation in assistance programs is associated with less food waste, and also that larger households have less food waste. In Japan, Wada and Shinagawa [29] conducted questionnaire surveys to analyze the relationship between the disposal of foodstuff and consumer awareness, but the survey targets were limited to specific universities.

Thus, few quantitative empirical studies have comprehensively addressed attribute factors and behavioral factors in developed countries where household food waste is a critical issue. While Lebersorger and Schneider [30] pointed out that urban residents generate more food waste than rural residents, the existing literature to date has ignored the relationship between regional characteristics and generation of food waste. This regional factor is worth examining, given that household attributes and behavior may vary between urban centers and rural regions.

In this study, we comprehensively address the attribute factors and behavioral factors of household food waste, using Japan as a case study and microdata from 5240 individuals, to clarify the contributing factors. We used an ordered probit model to determine both nationwide and region (large city, small city, village)-specific effects, and estimated the marginal effects of the main independent variables for each region. To the best of our knowledge, our study marked the first attempt to derive from nationwide survey data region-specific variables of food waste and their marginal effects. Such an endeavor—clarifying nationwide and region-specific factors of food waste—should provide valuable insights for national and local food policies. In the remainder of the paper, Section 2 explains the data and method used, Section 3 lists the estimation results, Section 4 discusses key findings and policy implications, and Section 5 concludes.

2. Data and Method

2.1. Data

We use pooled microdata from the national surveys conducted annually by the Japanese Ministry of Agriculture, Forestry and Fisheries (MAFF)—“Awareness Survey on Dietary Education”—from 2016 to 2018. The survey team interviewed individuals aged 20 years and older in 210 municipalities in Japan. For the three surveys, a total of 9000 questionnaires were distributed, yielding a total of 5240 effective responses. The survey sample was representative of the Japanese population and accounted for regional differences across the country. Three thousand samples in each year were proportionally distributed according to the size of the population in each stratum. The number of survey sites was determined such that the sample size of each stratum would be between 10 and 17. Specifically, the sample was selected using stratified, random, two-stage sampling, and the number of people surveyed in each municipality was proportionate to the municipality’s population. The survey included items on nutrition education, diet, and discarding food. The item on discarding food consisted of three questions, translated as follows: 1. How often do you feel disgusted at food being left uneaten or discarded? 2. How often do you discard food that you have bought? 3. Why do you discard the food? (for those who answered “sometimes” to 2). The survey also included numerous items on household attributes and behavior.

2.2. Method

We use the ordered probit model for analysis. From here, we will examine the impact of the explanatory variables, i.e., consumer attributes and behavioral factors, on the dependent variable, i.e., the amount of food waste generated.

First, suppose that the explained variable $Y_i$ corresponds to the following latent variable $Y_i^{*}$.

$$Y_i^{*}=x_i^{\prime }\beta +{\epsilon }_i i=1,2\dots ,N$$

Here, $\beta$ is parameter vector, $x_i$ is vector of explanatory variables for $i$, ${\epsilon }_i$ is error term.

Then, the relationship between $Y_i$ and $Y_i^{*}$ is as follows.

$$Y_i=\{\begin{array}{cc}1& if {\mu }_0<Y_i^{*}\le {\mu }_1\\ 2& if {\mu }_1<Y_i^{*}\le {\mu }_2\\ 3& if {\mu }_2<Y_i^{*}<{\mu }_3\end{array}$$

For $j=1,2,3$, the probability that $Y_i$ takes a certain value is as follows.

$$\begin{gathered} P_{ij}\left[Y_i=j\right]=P\left[{\mu }_{j-1}<Y_i^{*}\le {\mu }_j\right] \\ =P\left[{\mu }_{j-1}<x_i^{\prime }\beta +{\epsilon }_i\le {\mu }_j\right] \\ =P\left[{\mu }_{j-1}-x_i^{\prime }\beta <{\epsilon }_i \le {\mu }_j-x_i^{\prime }\beta \right] \end{gathered}$$

Then, assuming that ${\epsilon }_i$ follows a standard normal distribution, $P_{ij}$ can be expressed as a cumulative density function. By using $F(·)$ as the cumulative density function of the standard normal distribution, we have,

$$P_{ij}=F\left({\mu }_{j-1}-x_i^{\prime }\right)$$

Then, we estimate for parameter $\beta$ and cut point ${\mu }_0\dots {\mu }_3$ using the maximum likelihood method.

For each explanatory variable, the marginal effects are obtained as follows (In this study, we re-estimated $\beta$ in estimating the marginal effects. The explanatory variables with ordinal scales were estimated as continuous numerical variables in the estimation of the ordinal probit model, while in the estimation of marginal effects, $\beta$ was re-estimated and analyzed with each stage of the explanatory variables as independent categorical variables).

$$\frac{\partial P\left[Y_i-j\right]}{\partial x_i}=\left\{F^{\prime }\left({\mu }_{j-1}-x_i^{\prime }\beta \right)-F^{\prime }\left({\mu }_j-x_i^{\prime }\beta \right)\right\}\beta$$

In this model, we defined “level of food waste generation” as an explanatory variable, using the responses to the survey’s second question on food waste (“how often do you discard food that you have bought?”). This variable has three levels, tied to the three-point scale used in the question: “sometimes” (3), “rarely” (2), and “never” (1).

For explanatory variables, we refer to Roodhuyzen et al. [10], who list the contributing factors of food waste; among these, we comprehensively address those that can be extracted from the microdata regarding the “attributes” and “behavior” of consumers.

Table 1. Descriptive statistics and definitions.

Variable		Definition	Obs. 1	Means	Std. Dev. 1
Explained variable
Levels of food waste generation		There are three stages of “do you ever throw away food without eating it?”. “Sometimes” is 3, “rarely” is 2, “never” is 1	5240	2.163	0.724
Explanatory variables (i) Consumer attributes
Gender	Male	Dummy variable with males as 1	2276 2964	0.434	0.496
	Female
Age	Age: 20–29	Dummy variable with people aged 20 to 29 as 1	383	0.073	0.260
	Age: 30–39	Dummy variable with people aged 30 to 39 as 1	593	0.113	0.317
	Age: 40–49	Dummy variable with people aged 40 to 49 as 1	949	0.181	0.385
	Age: 50–59	Dummy variable with people aged 50 to 59 as 1	901	0.172	0.377
	Age: 60–69	Dummy variable with people aged 60 to 69 as 1	1111	0.212	0.409
	Age: >70	Dummy variable with people aged over 70 as 1	1303	0.249	0.432
Children in the household		Number of children in the household	5240	0.522	0.915
Housemates	No	Dummy variable with people with “no housemates” as 1	565	0.108	0.310
	Yes		4675
Income		Five-grade evaluation 2 of “do you have a comfortable life?”	5240	3.338	0.997
Occupation	Management position	Dummy variable with management position as 1	272	0.052	0.500
	Employee (other than management positions)	Dummy variable with employment position (other than management positions) as 1	2313	0.441	0.497
	Self-employed/family worker	Dummy variable with self-employed and family businesses as 1	604	0.115	0.319
	Housewife/househusband	Dummy variable with housewives and househusbands as 1	1186	0.226	0.418
	Student	Dummy variable with students as 1	73	0.151	0.358
Loss intention		Dummy variable with people who “want to reduce food waste” as 1	5240	0.388	0.487
Inheritance	Yes	Dummy variable with people who “carry on the tradition of local cuisine” as 1	3273	0.625	0.484
	No		1967
Agricultural experience	Yes	Dummy variable with people “who, or a family member, had agricultural experience” as 1	2461	0.470	0.500
	No		2779
Knowledge		Five-grade evaluation 2 on “do you know that the food will not become inedible immediately after the expiration date, and make a decision about disposal including other criteria?”	5240	3.978	0.975
Time		Five-grade evaluation 2 on “do you feel comfortable in your daily life?”	5240	3.486	1.261
Safety awareness		Three-grade evaluation 3 on “are you aware of safe eating habits?”	5240	1.959	0.683
(ii) Consumer behavior
Action		Three-grade evaluation 3 on “are you making efforts to reduce food waste?”	5240	2.328	0.719
Checking labels		Five-grade evaluation 2 on “do you check food labels at the time of purchase?”	5240	4.106	0.979
Food management		Five-grade assessment 2 on “do you follow a food preservation method?”	5240	4.033	0.838

Notes. ¹: “Obs.” means “number of observations”. “Std. Dev.” means “standard deviation”. ²: In “five-grade evaluation”, “very applicable” is 5, “applicable” is 4, “neutral” is 3, “not applicable” 2, and “not applicable at all” is 1. ³: In “three-grade evaluation”, “always practiced” is 3, “practiced” is 2, and “not practiced” is 1.

shows the descriptive statistics and definitions of the following variables. As explanatory variables for the “attributes” of consumers, we use “gender”, “age”, “children in the household”, “housemate”, “income”, “occupation”, “intention to reduce food waste” (whether the person wants to reduce food waste; hereinafter, “loss intention”), “inheritance of food culture” (hereinafter, “inheritance”), “agricultural experience”, “knowledge on the best-before date” (hereinafter, “knowledge”), “having time to spare” (hereinafter, “time”), and “awareness of safety in eating habits” (hereinafter, “safety awareness”). As factors related to the “behavior” of consumers, we use “actions for reducing food waste” (whether the person takes concrete actions to reduce food waste; hereinafter, “action”), “checking food labeling” (hereinafter, “checking labels”), and “food management”.

We sought to find the region-specific variables of food waste and the marginal effects of the key variables and divided all regions into two categories: “large cities” and “small cities”. Large cities are the wards of Tokyo, ordinance-designated cities, and cities with a population of 100,000 more, while small cities are cities with a population of less than 100,000. In addition, in order to compare large cities and rural areas, we established villages as a subcategory of small cities, defined as towns or villages in the administrative divisions of Japan.

Table 2. Descriptive statistics by region.

Variable		Large City (N = 3470)			Small City (N = 1770)			Village (N = 509)
		Obs.	Mean	Std. Dev.	Obs.	Mean	Std. Dev.	Obs.	Mean	Std. Dev.
Explained variable
Levels of food waste generation		3470	2.166	0.718	1770	2.157	0.733	509	2.167	0.717
Explanatory variables (i) Consumer attributes
Gender	Male	1474	0.425	0.494	802	0.453	0.498	217	0.426	0.495
	Female	1996			968			292
Age	Age: 20–29	270	0.078	0.268	113	0.064	0.245	35	0.068	0.253
	Age: 30–39	414	0.119	0.324	179	0.101	0.302	51	0.100	0.301
	Age: 40–49	631	0.182	0.386	318	0.180	0.384	89	0.175	0.380
	Age: 50–59	591	0.170	0.376	310	0.175	0.380	93	0.183	0.387
	Age: 60–69	717	0.207	0.405	394	0.223	0.416	112	0.220	0.415
	Age: >70	847	0.244	0.430	456	0.258	0.437	129	0.253	0.435
Children in the household		3470	0.517	0.909	1770	0.532	0.927	509	0.517	0.934
Housemates	No	412	0.119	0.324	153	0.086	0.281	37	0.072	0.260
	Yes	3058			1617			472
Income		3470	3.334	1.000	1770	3.335	0.991	509	3.305	0.980
Occupation	Management position	199	0.057	0.233	73	0.041	0.199	12	0.024	0.152
	Employee (other than management positions)	1530	0.441	0.497	783	0.442	0.497	223	0.438	0.497
	Self-employed/family worker	343	0.099	0.298	261	0.147	0.355	98	0.193	0.395
	Housewife/househusband	824	0.237	0.426	362	0.205	0.403	93	0.183	0.387
	Student	58	0.017	0.128	15	0.008	0.092	3	0.006	0.077
Loss intention		3470	0.403	0.490	1770	0.359	0.479	509	0.375	0.485
Inheritance	Yes	2145	0.618	0.486	1128	0.637	0.481	315	0.619	0.486
	No	1325			642			194
Agricultural experience	Yes	1529	0.441	0.497	932	0.527	0.499	283	0.556	0.497
	No	1941			838			226
Knowledge		3470	3.977	0.933	1770	3.981	0.910	509	3.957	0.906
Time		3470	3.466	1.258	1770	3.527	1.269	509	3.566	1.237
Safety awareness		3470	1.956	0.668	1770	1.963	0.682	509	1.941	0.680
(ii) Consumer behavior
Action		3470	2.327	0.717	1770	2.328	0.723	509	2.314	0.718
Checking labels		3470	4.114	0.965	1770	4.088	1.007	509	4.001	1.024
Food management		3470	4.047	0.835	1770	4.008	0.845	509	3.986	0.848

shows the descriptive statistics by region.

3. Results

3.1. Contributing Factors

As the estimated results show in

Table 3. Estimated results of contributing factors to food waste.

Variable		Coefficient	Z-Value	Large City	Small City	Village
(i) Consumer attributes
Gender (“Female” is 0)		−0.477 ***	−12.56	−0.469 ***	−0.500 ***	−0.610 ***
Age (based on “Age: 20–29”)	Age: 30–39	0.229 ***	2.80	0.200 **	0.275 *	0.411
	Age: 40–49	0.280 ***	3.66	0.212 **	0.402 ***	0.291
	Age: 50–59	0.353 ***	4.64	0.329 ***	0.398 ***	0.330
	Age: 60–69	0.348 ***	4.55	0.290 ***	0.456 ***	0.273
	Age: >70	0.041	0.51	−0.035	0.186	0.190
Children in the household		0.035 *	1.73	0.051 **	0.005	0.052
Housemate (“Yes” is set to 0)		−0.047	−0.89	−0.009	−0.128	−0.001
Income		−0.001	−0.07	0.019	−0.039	−0.017
Occupation (Based on “other than management position”)	Management position	0.023	0.30	0.046	−0.029 *	−0.369
	Self-employed/family worker	−0.131 **	−2.41	−0.063	−0.241 ***	−0.348 **
	Housewife/househusband	−0.066	−1.31	−0.010	−0.172	−0.269
	Student	−0.135	−0.93	−0.233	0.191	0.056
Loss intention		0.063 *	1.92	0.092 **	0.006	0.178
Inheritance (“No” is set to 0)		0.003	0.08	0.012	−0.017	−0.102
Agricultural experience (“No” is set to 0)		0.013	0.41	0.060	−0.092 *	−0.296 ***
Knowledge		−0.013	−0.75	−0.016	−0.013	0.012
Time		−0.033 **	−2.19	−0.034 *	−0.036	0.016
Safety awareness		−0.088 ***	−3.31	−0.101 ***	−0.061	−0.140
(ii) Consumer behavior
Action		−0.380 ***	−16.22	−0.393 ***	−0.353 ***	−0.324 ***
Checking labels		0.076 ***	4.11	0.064 ***	0.088 ***	0.098
Food management		−0.056 ***	−2.63	−0.077 ***	−0.021	0.063
Pseudo R2		0.074		0.076	0.074	0.094
Number of obs		5240		3470	1770	509

Notes. ***, **, and * are statistically significant at 1%, 5%, and 10% levels, respectively.

, we found the following contributing factors.

Consumer attributes. Regarding “gender”, women reported higher losses than men. The data used in this study are subjective evaluations. It is likely that women, who often prepare meals, have more opportunities to recognize the generation of food waste than men in Japan, and the data may have been influenced by this (According to a government survey on the division of responsibilities regarding food preparation between spouses in 2003, the wives were found to prepare meals in around 90% of the cases in Japan (see https://www.gender.go.jp/about_danjo/whitepaper/h15/summary/danjo/html/zuhyo/fig01_00_04_04.html (accessed 26 October 2021, available only in Japanese)). This means that husbands do not prepare meals as often and, hence, have fewer opportunities to recognize food waste. In fact, the average value of the explained variable was 1.99 for men and 2.75 for women, indicating that men are less likely to recognize food waste).

In terms of “age”, in both large and small cities, people aged 30–60 were more likely to generate food waste than people in their 20 s. In other words, working-age people tend to generate food waste. However, no significant effect was found in villages. Owing to the background of such regional differences, we could consider the influence of the presence or absence of housemates (e.g., grandmother or grandfather). From the descriptive statistics in Table 2, we can see that the percentage of housemates is relatively high in villages. Therefore, we added the cross terms of “age” and “housemate” to the estimation, and the results showed that the cross terms were not significant for people aged 30–60 years in both large and small cities. The likely explanation is that working-age people tend to generate more food waste when they do not have a housemate.

Regarding “children in the household”, food waste tended to occur with increasing number of children in large cities. To check whether the presence of a housemate affected the results, as it did in the case of “age”, we added a cross term between “children in the household” and “housemate”, and conducted the estimation. The cross term was not significant even in large cities in this case. Therefore, even in large cities, “children in the household” are less likely to generate food waste when parents live with a housemate.

In terms of “occupation”, “management position” and “self-employed/family workers” had a negative effect on food waste in small cities, and “self-employed/family workers” had a negative effect on food waste in villages, but there was no significant difference in large cities.

“Loss intention” was significantly positive in large cities, and there was a tendency to generate food waste even if intention was high. This point is further discussed in the “Consumer behavior” contributing factor under “action”.

“Agricultural experience” had a negative effect on the generation of food waste at a 10% level in small cities and 5% level in villages. There seems to be a difference in the level of food waste generation, depending on whether or not agriculture is carried out in the vicinity.

“Knowledge” was not found to be significant in any region, which contradicts the study by Nonomura [9] and others, which indicated that consumers confuse the best-before date with the expiration date and dispose of food based solely on the passing of the date, leading to food waste. This study did not confirm that “knowledge” had an effect on the generation of food waste, implying that “knowledge” alone does not lead to a reduction in food waste. A 2020 government survey provides further evidence that the knowledge factor alone cannot explain the increase in food waste; it reported that 87.5% of the 1967 consumers polled understood the difference between the shelf-life and best-before dates (2020 government survey on diet (available only in Japanese). Available online: https://survey.gov-online.go.jp/r02/r02-shokuseikatsu/index.html (accessed on 26 October 2021)). Van Boxstael et al.’s [13] survey in Belgium indicated that, while 69.6% of consumers were aware of the difference in the meaning of shelf-life and best-before dates, only 49.3% considered this difference when disposing of food. In our study, “knowledge” was not significantly related to the occurrence of food waste, probably because there was a gap between knowledge and how people actually dispose of food. This assumption will be analyzed again below, under “checking labels”.

“Time” and “safety awareness” show a significant effect only in large cities, and consumers with “time” and higher safety awareness were less likely to generate food waste. It is possible that busy consumers are more likely to generate food waste. Certain differences in lifestyles, such as working styles, could also influence food waste.

Consumer behavior. Regarding “action”, we found that food waste was less likely to occur in all regions. As mentioned above, “loss intention” did not reduce food waste, but “action” had a great influence on reduction; this highlights the gap between intention and behavior.

Table 4. Estimation results by cross terms of contributing factors to food waste.

Variable	Coefficient	Z-Value
“Loss intention” × “action”	−0.089 *	−0.86
“Loss intention” (re-listed from Table 3)	0.063 *	1.92
“Action” (re-listed from Table 3)	−0.380 ***	−16.22
“Knowledge” × “checking labels”	−0.040 ***	−2.81
“Knowledge” (re-listed from Table 3)	−0.013	−0.75
“Checking labels” (re-listed from Table 3)	0.076 ***	4.11

Notes. *** and * are statistically significant at 1% and 10% levels, respectively. The data refers to the whole country.

shows the results of using the cross terms of “loss intention” and “action”.

Although “loss intention” alone did not reduce food waste, the cross term between “loss intention” and “action” was significantly negative. In other words, intention of the need to reduce food waste alone will not reduce food waste, rather the actual behavior by intended consumers will.

Regarding “checking labels” and “food management”, significant effects were seen mainly in large cities, and the latter factor made food waste less likely to occur, while the former had a positive effect on generating food waste. As “checking labels” leads to food waste, people may be overly concerned about the information on food labeling, without having sufficient and correct “knowledge”, and end up discarding the food. As seen in “consumer attributes”, “knowledge” was not significantly related to the level of food waste generation, while “checking labels” had a significant effect; this again points to the gap between knowledge and behavior.

Table 4 shows the results of using the cross terms, “knowledge” and “checking labels”, which were negatively significant. Therefore, there is a tendency for food waste to be generated simply by “checking labels”, but the same will lead to reduction of food waste if there is adequate and correct “knowledge”. There was no significant effect for just “knowledge”, but a positive significant effect for just “checking labels”. However, when this knowledge is combined with the action of checking, the effect of reducing food waste can be demonstrated.

Regarding “food management”, consistent with the literature, we found that consumers with appropriate storage management were less likely to generate food waste. Studies have highlighted the role of refrigeration management, suggesting that refrigeration errors contribute to food waste and that better refrigeration management can reduce it [15,16]. Our findings provide empirical support for this understanding.

3.2. Marginal Effects

From significant variables in Table 3,

Table 5. Marginal effect of main factors to food waste.

		Explained Variable: Level of Food Waste Generation
Changes in the Level of Variables	Area	1 (Never)	2 (Rarely)	3 (Sometimes)
Agricultural experience
0→1	Large	−0.015	−0.006	0.021
	Small	0.024 *	0.008 *	−0.032 *
	Village	0.070 ***	0.029 **	−0.099 ***
Safety awareness
1→2	Large	0.012	0.005	−0.018
	Small	−0.018	−0.007	0.250
	Village	0.009	0.005	−0.014
1→3	Large	0.049 ***	0.016 ***	−0.067 ***
	Small	0.032	0.007	−0.040
	Village	0.085 *	0.023 *	−0.108 *
Action
1→2	Large	0.029 ***	0.031 **	−0.060 ***
	Small	0.054 ***	0.047 ***	−0.101 ***
	Village	0.057 *	0.052 *	−0.109 *
1→3	Large	0.160 ***	0.079 ***	−0.240 ***
	Small	0.158 ***	0.076 ***	−0.234 ***
	Village	0.138 ***	0.079 ***	−0.217 ***
Food management
1→2	Large	0.042	0.032	−0.074
	Small	−0.118	0.001	0.117
	Village	0.006	−0.000	−0.005
1→3	Large	0.058	0.040	−0.097
	Small	−0.159	−0.013	0.171 *
	Village	−0.090	−0.013	0.102
1→4	Large	0.059	0.040	−0.100
	Small	−0.146	−0.007	0.153 *
	Village	−0.102	−0.018	0.120
1→5	Large	0.090 **	0.051	−0.141 *
	Small	−0.118	0.001	0.117
	Village	−0.087	−0.012	0.099

Notes. ***, **, and * indicate statistically significant at 1%, 5%, and 10% levels, respectively. For the level of variables, see the definitions in Table 1. For “agricultural experience”, 0 means “no” and 1 means “yes”. For “safety awareness” and “action”, 1 means “not practiced”, 2 means “practiced”, and 3 means “always practiced”. For “food management”, 1 means “not applicable at all”, 2 means “not applicable”, 3 means “neutral”, 4 means “applicable”, and 5 means “very applicable”.

shows the marginal effects of each variable, namely, “agricultural experience”, “safety awareness”, “action”, and “food management”. For example, when the level of “action” changes from 1 (“not practiced”) to 2 (“practiced”), the level of food waste generation 1 (“never”) increases by 0.029 in large cities, 0.054 in small cities, and 0.057 in villages.

In particular, the marginal effect of “agricultural experience” shows that, in villages, when “agricultural experience” changes from 0 (“there is no agricultural experience”) to 1 (“there is agricultural experience”), the generation level 1 (“never”) of food waste increases by 0.07, and conversely, generation level 3 (“sometimes”) of food waste decreases by 0.099. It is likely that people who are close to and interested in agricultural production sites are more likely to be aware of the importance of food, and this is reflected in their actual behavior.

Moreover, the marginal effect of “action” is effective even when little effort (1→2) is made in all regions, and a greater effect can be obtained by carrying out more efforts (1→3). As for the impact of the marginal effects, “action” has the largest impact among the variables. By region, especially in large cities, a great effect can be seen when “action” changes from 1 (“not practiced”) to 3 (“always practiced”).

However, “safety awareness” and “food management” proved significant (at 0.05) only in large cities. Moreover, the marginal effect of “food management” only becomes significant when the greatest effort (1→5) is made.

4. Discussion of Key Findings and Policy Implications

We discuss the key findings on, first, the nationwide factors and then the region-specific factors of food waste. In addition to “gender” being a factor that had an increasing effect on food waste in all regions and one that was influenced by subjective evaluation, “action”, a consumer behavior factor, significantly lowered the generation of food waste. “Loss intention” does not have a negative significant effect, and there is a gap between intention and behavior. In other words, intention of the importance of reducing food waste alone does not lead to a reduction; rather, it does only when consumers actually act. This behavioral factor of “action” has the largest marginal effect among all factors. Even with little action, there is an effect on reducing food waste, implying that a greater effect can be obtained with more actions.

Overall, we find that in large cities, working-age people and those who have many “children in the household” are more likely to generate food waste, while those with “time”, “safety awareness”, and “food management” are less likely to generate food waste. “Age” and “children in the household” have no effect in villages, perhaps due to the differences in lifestyles among working-age people. In large cities, “knowledge” does not influence the generation of food waste, while “checking labels” does, reflecting a gap between knowledge and behavior. “Checking labels” alone tends to cause food waste, but when the right knowledge is combined with the action of checking, the effect of reducing food waste can be demonstrated.

In villages, “agricultural experience” has the effect of reducing food waste; this factor has a relatively high marginal effect in villages. People who are close to and interested in agricultural production are more likely to be aware of the importance of food and act accordingly.

Of the above key findings, unlike the qualitative previous studies, our quantitative analysis reveals that “intention” or “knowledge” alone, such as “intention to reduce food waste” and “knowledge on the best-before date”, does not reduce food waste; on the other hand, food waste can be reduced when these are combined with “behavior” to reduce food waste, such as “actions for reducing food waste” and “checking food labeling”.

Based on our results, we can draw three policy implications for reducing food waste. First, policies aimed at reducing household food waste require not only raising public awareness, but also making concrete suggestions on how consumers can implement them in their daily life. Knowledge and loss intention alone will not reduce food waste but will be effective only if they are linked to actual behaviors, based on action and checking labels with intention. For example, along with enlightenment to broaden the range of people who are interested in the food waste issue, opportunities should be provided to gain a deeper awareness that will influence behavior, such as holding workshops in the local institution where people can learn about and think about the food waste issue. In addition, measures can be taken to label food in such a way that it catches the eye during the consumption of food in daily life, for example, noting that it is not the same as the expiration date as a set with the notation of "best before" date.

Second, in large cities, workers are more likely to generate food waste; in contrast, the more time they have, the less they would generate food waste. Thus, suggestions on ways to improve lifestyle, such as work–life balance, should be incorporated.

Third, in villages, those with agricultural experience are less likely to generate food waste. Better interaction with agricultural production sites can reduce food waste in such cases. Such a solution would be useful for cities as well; for example, encouraging consumers living in cities to visit food production sites, or promoting the sale of food while indicating who produced it and where so that people can recognize the sites of food production.

5. Conclusions

In this study, using micro data on 5240 individuals from a nationwide survey conducted by the MAFF, Japan, we conducted an empirical analysis of the contributing factors of household food waste from the perspective of consumer attributes and behavior by classifying regions into large cities, small cities, and villages.

At a nationwide level, “actions for reducing food waste” significantly reduced food waste, while “loss intention” had a limited effect. This finding reveals a gap between attitude and behavior. A desire among consumers to reduce food waste will lead to an actual reduction only if it is accompanied by committed actions. “Actions for reducing food waste” had the largest marginal effect of all the variables, suggesting that food waste could be reduced with even limited action, and reduced significantly with more extensive action.

As for region-specific findings, in large cities, “knowledge” had no notable effect, while “checking labels” facilitated food waste. Here, too, we noted a similar gap between attitude and behavior; these factors together contributed to reducing food waste. Food waste in large cities was also associated with lifestyle. Specifically, working long hours increased the likelihood of food waste, whereas having more free time decreased the likelihood. In villages, “agricultural experience” reduced food waste. Villagers who lived close to farms or were interested in farming sought to avoid wasting food while also putting it into practice.

Our findings above provide insights for national and regional policies for reducing food waste. Nevertheless, the level of food waste is based on the subjective evaluation of consumers in our study, which may be a limitation. In other words, the accuracy of the analysis of household food waste can be further improved using objective indicators.