The Effect of Overseas Educational Experience on Pro-Environmental Practices: Evidence from Cambodian Academic Scholars

This paper revealed that after spending years overseas for learning, Cambodian fellows realized that a clean society more likely results from the adopting of a waste management system. The lack of such is the main barrier for waste separation willingness and practices in Cambodia. Self-transcendence values and volunteerism are empirically found as the fundamental factors for any interventions designed to promote pro-environmental intentions and practices respectively. The fellows engaging in volunteerisms appear to do waste separation, ignoring how inconvenient the waste management system or facilities are. Similarly, the fellows concerning the environmental and social issues appear to spare their valuable time to do waste-separation which they find not difficult. This is not the case in this COVID-19 epidemic though. The study results imply that in a society where an inconvenient waste management system is in place the fellows without past volunteering, no matter how knowledgeable they are, will less likely engage in the waste-sorting practice.


Introduction
There has been increasing recognition that human activities have become so pervasive and profound in their consequences that they degrade the environment on a global scale [1]. As of 2015, approximately 6300 million tons of plastic waste had been generated by the continuing growth of the population [2], around How to cite this paper: Ros

Material and Methods
The study was conducted online using Survey Monkey. Two groups, the testing group and a control group were interviewed. In cooperation with the Cambodian Association in Japan (CSAJ) and the Hiroshima Alumni Network (HAN), the online survey was administered using the mailing list of the Association and Telegram Apps of the network to target the alumni, a group of people who had been educated overseas (N = 164). Around 85% of respondents gained their educational experiences in Japan. The remaining 15% experienced overseas education in Thailand, New Zealand, Australia, South Korea, or the Netherlands.
To reach the group of people with only a local education in Cambodia (N = 59), the questionnaires were distributed via social media, Facebook. Within two weeks, starting from mid-November 2020, a total of 224 respondents agreed and participated in the survey.

Outcome Variables
The two outcomes to be tested include pro-environmental behaviors and the environmental knowledge affecting people's motivations towards waste separation.
Waste separation is explored as being a requisite for any campaigns to promote solid waste management in Cambodia [75]. As an indicator of behavior, respondents were asked about the frequency with which they had engaged in two types of waste management practices: 1) waste separation and 2) waste reduction. This is a single-item measure: "How often do you separate waste?" and "How often do you use your own reusable shopping bag?" The frequency of their waste disposal practice is then divided into five categories: never, rarely, sometimes, often, and always.
For environmental knowledge, respondents were asked to rate the 5-point Likert scale on five items: 1) "People in my family don't support waste separation"; 2) "I will not segregate unless the waste collection service requires me to do so"; 3) "My neighboring families never separate their waste"; 4) "Waste separation is difficult" and 5) "I don't have time to separate waste". Participants responded on a 5-point scale (0 = strongly disagree to 4 = strongly agree).

Control Variables
Participants reported their gender, age, and marital status; these three background variables were included in our subsequent analyses as control variables. A binary indicator (1 = volunteerism, 0 = otherwise) was also added as the covariate since it strongly affects waste management practices [47]. Since the survey took place during the outbreak of COVID-19 by which human physical and mental health is affected [76], it is crucial to some extent to understand its relationship with daily practice of waste management. So, respondents were asked to rate the 5-point Likert scale on one-item measure for one's own concerns over COVID-19: "How concerned are you about the possibility of becoming infected by . Participants responded on a 5-point scale (0 = not at all to 4 = extremely concern). Higher scores indicate people are concerned about the possibility of a COVID-19 infection.
For social and environmental concerns, respondents were asked to rank the 7-point Likert scale on nine-item measure (0 = extremely disagree to 6 = extremely agree). These items were adopted from a study conducted to measure human value in environment, self-transcendence [77] [78]. A row mean score of the construct was computed after the reliability and validity of the construct were measured. Higher scores indicate people are more likely altruistic and biospheric. Descriptive statistics of these outcomes and control variables can be found in SI Table B1.
To evaluate the construct reliability and validity, Alpha Reliability ( ∝ ), the internal reliability (rho) [79], the convergent validity of the indicators as well as the composite reliability of the construct [80] were all tested. These tests are used to find out if all indicators belonging to the same construct are highly correlated with each other. The internal reliability (rho) was computed based on the result of Confirmatory Factor Analysis (CFA) analyzed with Stata 15.1.
The CFA results show the goodness of fit with the data set. All the factor loadings are substantial and statistically significant. If you are 1 standard deviation higher on self-transcendence, you will respond 0.90 standard deviations higher on X1 (i.e., it is important to love nature more), 0.57 standard deviations higher on X6 (i.e., it is important to avoid war) respectively (see SI Table B2).

Balance Test
The test assumed that the respondents assigned to either the control or testing group differed only in their access to education. To show that the statement is true, a balance test was conducted on six covariates, including age, gender, spouse, COVID-19 concerns, volunteerism, and self-transcendence. The test used t-tests as a means for comparison to check if there is statistically significant difference between the two groups. As a result, the null hypothesis that the true difference in means of each covariate is equal to 0 is not rejected (Table 1). So, the above-mentioned assumption holds.

Econometrics Analysis
To generate causal estimates of the effects of overseas educational experiences on waste separation behavior, I employed an econometrics analysis, the Instrumental Variable (IV) design. The intent is to test if there is statistically significant difference between the testing group (the group experiencing overseas education) and the control group (the group experiencing only local education), in terms of their behaviors and knowledge towards waste separation. So, the testing group is a binary variable (1 = overseas educational experience, 0 = otherwise). Though this is not the case for the control group, each group is also endogenous to general education. So, I instrumented the testing group (endogenous regressor) by education level (instrument). I estimated a two-stage least squares (2SLS) strategy with the following equation: First-stage regression:  model. An instrumental variable must not be correlated with the equation's disturbance ϵ, and it must be highly correlated with the included endogenous regressor. As recommended by [82], the partial R 2 and F-statistic of the identifying instruments first-stage regression are useful indicators of the quality of the IV estimates and should be reported. It shows the strength of instrument. According to [83] [84], an F-statistic less than 10 is cause for concern or means that our instrument is weak. After being tested, the resulting F-statistic of each model is higher than 10 with p < 0.01, meaning that the instrument is not weak.

Results and Discussion
As a result, overseas educational experience obviously gave Cambodian fellows not only academic experience, but also consciousness of the need for waste recycling and practical ability. Waste separation as a practice which was found statistically significant at 5% level (β = 1.060, p < 0.05) is not surprising (regression 3; Table 2) considering Japan rules, for instance, that waste separation at source is compulsory. However, the practice sharply decreased and become statistically insignificant (β = 0.378, p > 0.05) upon controlling fixed effects (regression 4; Table 2). So, the variation in waste separation practice is more likely explained by situational factor, the access to an effective waste management system particularly.
This result reflects the significant role of waste management system in managing solid waste of either country. In Cambodia, waste separation is neither mandatory at source, nor practical at waste management system. Therefore, people, no matter how knowledgeable they are, will be less likely to practice waste separation while facing inconvenient circumstances.  Waste separation behavior is obviously determined by waste management facilities and social context [89]- [94]. Therefore, the fellows are neither to adhere to responsible waste separation upon their return home since they had realized that formalized and widespread waste collection system in their municipality has not evolved.
Similarly, overseas educational experience could not even facilitate environmental knowledge change which is unexpected and different from previous studies [27] [41] [42]. No statistically significant difference was detected between the testing group and control group in terms of the five items used to assess the knowledge change. The fellows all had similar understanding on waste problems. They appeared to disagree with the statements that waste separation is difficult (β = −0.243, p > 0.05), waste separation is impossible until there is an active participation from their neighbors (β = −0.905, p > 0.05) and waste separation is impossible until there is support from their family members (β = −0.532, p > 0.05). However, the fellows experiencing oversea education seemed to have stronger commitment on time allocated for waste separation (β = −0.773, p < 0.10), particularly as soon as an improved waste collection system is installed.
The fellows realize that a clean society is more likely achieved with an improved  Table 3).
Those fellows who felt concern with COVID-19 infection seemed to find waste-separation difficult and complicated (β = 0.165, p < 0.10) and they appeared to allocate their time for taking care of their health rather than waste separation (β = 0.236, p < 0.01). It is a logic that when both our physical and mental health are severely affected by COVID-19, our daily routine including waste management practice will also be affected [77]. It should be the municipality to take overall responsibilities for municipal solid waste management, particularly in this pandemic, they insisted (β = 0.175, p < 0.10) (regression 4; SI Tables   B4-B6).
On contrary, people with self-transcendence values expressed their opinion in an opposite way. They appeared to be able to allocate their time to do waste-separation (β = −0.422, p < 0.01) which they found not difficult (β = −0.388, p < 0.01). They seemed to be ready to spend their time sorting their garbage for a better environment (regression 4; SI Tables B4-B6). This result is parallel with the previous result, citing that people would spend time on pro-environmental activities if they realized the environmental problems and they could mitigate them [95]. For those fellows who got married, they disagreed waste sorting needs involvement of the neighbors (β = −0.423, p < 0.10) (regression 4; SI Table B7), while those fellows experiencing clean-up activities (i.e., volunteering) disagreed waste sorting needs the involvement of family members (β = −0.284, p < 0.10)(regression 4; SI Table B8). By the way, volunteerism (i.e., past volunteering in clean-up activities) is the only variable found statistically significant (β = 0.497, p < 0.05) in explaining the variation in waste separation practice which is consistent with the previous study of [47]. Noticeably, its interaction effect with the coefficient of overseas educational experience makes the model statistically significant at 5% level with F-statistic = 156.42. The insignificance of overidentification test X 2 (1, N = 203) = 0.077; p = 0.781 confirms the validity of the model (regression 5, SI Table B9).
Thus, this finding implies that people experiencing overseas education and voluntarily engaging in clean-up activities more likely practice waste separation for environment's sake no matter how inconvenient the waste management system or facilities is. So, social factors are unlikely the barrier concerning waste-separation practice which is beyond the facts claimed by previous studies. Nevertheless, Self-transcendence values (i.e., biosphere plus altruism) were unlikely the determinant of waste separation practice, even though it has been regarded as the foundation for any interventions to promote waste separation behavior [51] [52]. Its interaction effect was not significant (β = 0.088, p > 0.05) with small F-statistic = 10.65 (regression 6; SI Table B9). This is likely from the fact that a variety of behaviors are not always explained by human values [54] and a predictor of one behavior is unlikely the predictor of the others [21]. Values appeared to affect behaviors more influenced by personality traits. For instance, a study found that values or identity was one of determinants of plastic consumption behavior [96] and another study found that altruistic people more likely, for the sake of environment and its enjoyment, engaged in beach clean-up activities [56].
Though gender did play a crucial role in predicting pro-environmental intentions (SI Table B10 [100], this study didn't find its effect on waste-sorting practice, but shopping-bag usage (β = −0.507, p < 0.01) (regression 4; SI Table B12). From the correlation matrix (SI Table B13), female respondents are more concerned with environmental and social issues (r = −0.23, p < 0.05) [101] [102] [103] and whereby more engage in environmental and social affairs voluntarily (r = −0.18, p < 0.01). So, any interventions designed to promote pro-environmental behaviors more likely be met with more success if oriented to participants regardless of gender. Different from [41], this study did not find any spousal influence on both waste-separation and shopping-bag usage.
The findings of this study obviously provide insights on how the eco-innovations, a way of creating a successful circular economy [104] as cited by [105], are performed particularly in developing countries where waste management system and recycling facilities are under development. Waste-sorting practices among consumers, known as a type of 3Rs (i.e., reduction, reuse and recycle), is the environmentally responsible behavior required for the success of circular economy [ and Production [110].
Among the three determinants of eco-innovations (i.e., policy and regulation, supply side and demand side determinant), this study suggested to act immediately and as a priority regarding the policy and supply side determinant. The Best Environmental Practices (BEP) provided in the Stockholm Convention National Implementation Plan (NIP) which is being implemented by developing countries [112] will be more effective if attitudinal factors (i.e., volunteering, and self-transcendence values) which is also found in Europe as an inevitable one [113], are considered. Community-based Waste Management approach, an alternative to convince more and more engagement from the public in waste management [114], shall be promoted in the policy as well. It has been found as a solution to raise public awareness as well as to mobilize people to voluntarily participate and ensure the cleanliness in a community (i.e., to monitor and prevent waste disposal).
Nevertheless, at supply side, waste management system and recycling facilities at municipality level shall be upgraded as well. The better the waste management system is, the more convenient people will feel, whereby more public participation in waste management is expected [20]- [25]. For demand side (i.e., consumer needs), in addition to theoretical knowledge and know-how, increasing people interest in the value of nature is important. This study suggested to design a comprehensive environmental education program (i.e., theories, practices, and technology) in educational institutes, especially among children [9] [10] [114].

Robustness Check
Since the waste-separation practice is insignificant, the waste-separation intention among the fellows might be statistically insignificant too, according to the theory of planned behavior [46]. In developing country like Cambodia where waste collection facilities are still conventional and not yet upgraded, waste-separation will be feasible for only two purposes -for compost or for sale. Respondents of either group (i.e., testing group and control group) are unlikely to have intention to separate their garbage for compost or even for sale. Respondents were then asked to rate the 5-point Likert scale on three items for 1): "I plan to separate my garbage into organic and non-organic", 2): "I will sell my garbage separated" and 3): "I will compost my garbage separated". Participants responded on a 5-point scale (0 = most unlikely to 4 = most likely).
From  The shopping-bag usage behavior, where it is more influenced by personality traits than social factors; was tested and found statistically significant at 10% level (β = 0.834, p < 0.10) (regression 4; full results are in SI Table B12). Experiencing overseas education contributes to increasing the probability of using shopping bag (SI Table B15). Thus, the variation in shopping-bag usage behavior was indeed unexplained by situational factor. The fellows that had used shopping bag were not influenced by the new context and they seemed to keep practicing this behavior regardless of the places they reside in. This could result from the existing capacity among the fellows, as claimed previously that people who have environmental knowledge can make more informed decisions on how and what to consume and dispose [11] [12] as cited by [13] [118]. In this sense, the capacity related to waste management or disposal of either group (i.e., testing group and control group) appears to be not significantly different.

Conclusions
The overseas educational experience is partially linked to knowledge change among Cambodian graduates. Understanding that a clean society basically rests on the waste management system governed by the municipality, the willingness to separate waste became less likely upon their arrival at their home country, Cambodia, where solid waste management systems are not convenient or functionally unreliable. Thereby, waste separation practice was not detected as well However, past volunteering in clean-up activities was found as the only predictor significantly associated with waste-separation practice and its interaction effect with the coefficient of overseas education contributes to increasing the probability of sorting garbage even in an inconvenient context. Conversely, human concerns about the environment and social welfare (i.e., self-transcendence) were seen as fundamental to promoting intentions rather than the practice toward waste separation. This finding proves that there is indeed a value-action gap, particularly for social influence-based practices (i.e., waste-sorting). This finding suggested promoting volunteering among children and teenagers which could be made via its application in the NIP, the community-based waste management approach, the Early Childhood Learning Program within the national educational system as recommended by [9] [10] or the experiential learning This study was conducted targeting only those Cambodians with higher education. With an average age of 30 years old and standard deviation of 6 years, the respondents of this study were in the minority at the educational institutions.
This group represents merely 6.7% of those persons aged 25 and above who have attained post-secondary education [120]. So, these study results cannot be gene- There is doubt regarding the effect of "self-transcendence" on waste sorting practice. Since [55] and [56] confirmed altruism has a relationship with pro-environmental behaviors, the combined altruism and biosphere into one-dimension, self-transcendence, should expectedly increase its strength and sensitivity. If so, its effect on pro-environment behaviors is stronger and can be statistically detected more easily. However, it was not the case in this study. So, whether this one-factor or two-factor construct is better at predicting pro-environmental behaviors should be further explored.

Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.

1) How to dispose of Household Garbage
In Japan, depending on the city, town, or district, there may be designated bags that residents are required to use for trash. In other words, the rules for separating and disposing of garbage depend on the local municipality. Many municipal offices provide pamphlets ( Figure A1) that explain the rules of garbage disposal. Waste sorting at household level is strict. Figure A1. Guideline for household garbage disposal in Higashi-Hiroshima city, Japan.   Standard errors are in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.  Table B13. Effect of overseas educational experience on shopping-bag usage.

Interaction Effect:
To make sure if people engaged in social and environmental affairs (volunteerism) will practice waste separation regardless of how inconvenient the waste management system is, I interacted the dummy variable of overseas educational experience with volunteerism variable (regression 5, Table B9). The result is robust to the claim with coefficient (β = 0.497, p < 0.05) with F-statistic = 156.42. On contrary, the interaction effect between the coefficient of overseas educational experience and self-transcendence is statistically insignificant (β = 0.088, p > 0.05) with small F-statistic = 10.65 (regression 6, Table B9).

Average Marginal Effects:
The probability of sorting garbage among the fellows is statistically insignificant upon their arrival in Cambodia (regression 4, Table B9). However, the probability of using shopping bag is statistically significant (regression 4 & 5, Table B12) even after their arrival at their home country. The coefficient of average marginal effect is all negative in the first bottom three categories: never, rarely and sometimes but positive for often and always (Table B15). Therefore, experiencing overseas education decreases the probability of not using shopping bag but increases the probability of using it by approximately 11% and 16% for category "often" and "always" respectively.