2.1. Study Area
Phnom Penh consists of 9 districts and 36 communes, and the provision of waste collection services varies by location. To determine the study sites, we sought a holistic understanding of current waste management service conditions, such as collection frequency and facility provision. Cintri, a private company in charge of collection services for the whole city, provided information on waste collection in each district in the city [
21]. These data were compiled into a digital map by a geographic information system (GIS; ArcGIS Desktop ver. 21, ESRI, Tokyo, Japan), shown in
Figure 1. As
Figure 1 illustrates, the central part of the city receives daily waste collection services; in contrast, rural areas receive collection services only two or three days a week. Two types of dumpsters with 4 ton capacity (4T) and 10 ton capacity (10T) are provided, and the latter ones are located mainly in the central area.
We assembled the data and calculated population density by population [
22] and area (calculated by GIS). The index “collection frequency/population density” was used to select the study sites. As shown in
Table 1, four sites were selected, HH, LH, HL, and LL, which represent varying levels of collection frequency (H: high, L: low) and population density (H: high, L: low).
2.2. Theoretical Framework and Research Hypotheses
A number of theories have attempted to explain why people do or do not engage in a certain behavior. One of the most applauded was the theory of reasoned action (TRA). TRA, first introduced by Ajzen and Fishbein, proposed that “behavior” was determined by “intention,” which was in turn determined by “attitude toward the behavior” and “subjective norms” [
23]. Attitude refers to the degree that people value the behavior positively or negatively. Subjective norms are the perceived social pressure from those a person is close to, such as family and friends, regarding whether to engage in the behavior or not. This model implied that if people valued a behavior positively and perceived that their friends and family expected them to perform the behavior, they would intend to perform the behavior. However, although many research findings agree with this theory (e.g., Park et al., 2009 [
24]), it has also received a great deal of criticism for its assumption that people’s behavior is determined solely by their intention. Such criticism noted that the model applies only when the behavior is under volitional control. Thus, Ajzen (1991) introduced another theory to predict people’s behavior: the theory of planned behavior (TPB) [
25]. This theory, a revision of the controversial TRA, introduced one more predictor, perceived behavioral control (PBC), which determines intention and behavior. PBC represents how much the target behavior can be controlled by an actor himself/herself. In other words, PBC refers to the degree to which people perceive the behavior as easy or difficult to perform. TPB has become the most popular theory to predict people’s behaviors and it has been widely applied to waste-related behaviors such as recycling [
26,
27,
28] and prevention [
29,
30].
Schwartz introduced the norm activation model, which explained that the activation of personal norms is an important process in determining altruistic behavior [
31,
32]. Personal norms are the “expectations people hold for themselves” regarding whether they should or should not perform a certain behavior [
31]. This model also explained that personal norms are created through the internalization of social norms [
32,
33]. Bortoleto et al. (2012) showed the significant influence of personal norms on waste prevention behaviors [
29].
In addition to the TPB variables and personal norms, other variables that predict waste-related behaviors have been reported. Hornik et al. (1995) showed that besides attitude and intention, laziness and ignorance also played a crucial role in predicting people’s recycling behavior [
15]. Davies et al. (2002) reported that the evaluation of effectiveness was an important predictor of recycling behavior [
11]. Knowledge, concern, and obligation to environmental issues have also been identified in determining recycling and waste prevention intentions [
30,
34]. Chiang et al. (2019) suggested that locus of control is a key determinant to promote pro-environmental behavior [
35]. For waste littering, locus of control has also been proposed as an influential determinant as well as altruism and self-efficacy (e.g., Ojedokun, 2011; Ojedokun and Balogun, 2011) [
3,
36]. Self-efficacy is a belief in one’s ability to accomplish a task or succeed in specific situations. Locus of control refers to whether people believe that they or external powers are in control of their life.
Based on these theoretical frameworks, the present study developed a hypothetical model involving key internal and external factors for waste disposal behavior in public, as shown in
Figure 2. This model set “(a) intention not to dispose of waste in public” (Int_nd) as the target dependent variable. In addition, the variable “(b) intention to keep house and outdoor surroundings clean” (Int_kc) was assumed to have an influence on Int_nd (H1).
The TPB model was adopted with respect to internal factors determining intentions. This model postulates three conceptual determinants, such as “(c) attitude toward the behavior” (Att), “(d) subjective norms” (Sb), and “(e) perceived behavioral control” (Pbc), that predict people’s intentions and behavior. Our model assumed that positive Att has a positive influence on Int_nd (H2), and that Sb and Pbc have influences on both Int_nd (H3, H5) and Int_kc (H4, H6). Our model also assumed some additional internal factors, namely, “(f) social pressure” (Sp) and “(g) personal norms” (Pn). In addition to perception of social pressure from friends and family, perception of institutional and community-level social pressures can influence people’s intentions. Therefore, we involved Sp as well as Sb and assumed that Sp has positive influences on Int_nd (H7) and Int_kc (H8). The model also assumed that Pn has positive influences on Int_nd (H11) and Int_kc (H12), and that social norms such as Sb and Sp are internalized into Pn (H9, H10).
Past experiences and behaviors have been reported to influence current intentions and behaviors. For example, Tonglet et al. (2004) showed that the main determinants of recycling behavior were recycling attitude, previous recycling experience, concern for the community, and consequences of recycling [
37]. Thus, our model assumed the influence of “(h) past experience of waste disposal in public” (Exp) on Int_nd (H13).
Ojedokun (2011) found that personal attributes, such as altruism and locus of control, had impacts on people’s attitude toward littering and responsible environmental behaviors [
3]. Hornik et al. (1995) suggested that the basic problems of recycling were internal barriers such as ignorance, laziness, and inconvenience [
15]. McCarty and Shrum (1994) showed that the inconvenience of recycling strongly influenced people’s recycling behavior [
38]. Based on these previous studies, we assumed that “(i) laziness, ignorance, and locus of control” (Lz, Ig, and Lc) have influences on Int_nd (H14) and Exp (H15).
In considering influences from external factors, two possible aspects were involved. First was the objective condition of waste management, indicated by “(E1) collection frequency” (Cf) and “(E2) facility (bin) provision” (Fp); second was people’s perception of waste management based on current external conditions, indicated by “(j) satisfaction with current waste management services” (Sat) and “(k) concern about waste management” (Con). We assumed that “(E3) population density” (Pd) determines Cf and Fp (H22, H23), Cf and Fp affect Sat (H20, H21), Sat influences Con and Int_nd (H19, H16), and Sat and Con influence Exp (H17, H18).
2.3. Questionnaire Design
The questionnaire consisted of three parts. The first part (I_1–9) measured demographic information: age (I_1), gender (I_2), address (I_3), family income (I_4), education level (I_7), family size (I_8), and house type (I_9). To validate the appropriateness of the answer for family income, car (I_5) and motorbike (I_6) ownership were also recorded.
The second part (II_1–14) included questions about current waste collection services and how people managed the waste they generated: awareness of current waste management (II_1, 2), collection and disposal frequency (II_3, 8), person responsible for waste disposal (II_4), usual number of waste bags disposed (II_5), disposal location (II_6, 9), method of disposal for recyclable wastes (II_7), satisfaction with current services (II_10), any experience of disposing of waste in public open spaces (II_11) and the reason for doing so (II_12), perception of the sufficiency of facility provision (II_13), and free opinions on current waste collection services (II_14).
The third part (III_1–22) included questions concerning the main internal variables (a–g, i): Int_nd (III_16), Int_kc (III_20), Att (III_1, 21), Sb (III_2), Pbc (III_3, 9, 13, 18), Sp (III_4, 7, 10, 11, 17, 19), Pn (III_8), Lz (III_12, 22), Ig (III_6, 15), and Lc (III_5, 9). For each of the 22 statements shown, respondents expressed their opinions using a 6-point scale ranging from 1 “strongly disagree” to 6 “strongly agree”.
2.4. Questionnaire Survey
Target streets were selected for each study site as shown in
Table 1, and respondents were selected by fixed intervals of blocks along the street. For structural equation modeling with more than 10 variables, it is recommended that the sample size be 100 or greater, as a smaller sample could produce unstable and insignificant results [
29,
39]. The total population of Phnom Penh is 1.6 million; according to Yamane’s (1967) formula, 400 samples are needed for this population to obtain a 95% confidence level with significance at the level of 5% [
40]. Thus, a sample size of 100 was used for each of the four sites, resulting in a total sample size of 400.
Waste disposal is a sensitive issue; therefore, face-to-face interviews were considered to be an appropriate method for the survey, allowing an interviewer to give supplemental explanations if the interviewee could not understand a question’s meaning. Nine Cambodian university students were employed to help conduct the interview survey. Prior to the interview campaign, two training sessions were conducted from 10 January to 11 January 2018. In the first session, all questions were explained to ensure that the interviewers understood all materials clearly. In the second session, a mock interview survey was conducted to determine whether the interviewers understood all questions well and how flexibly they could handle a real situation. After these training sessions, the main survey was conducted from 12 January to 19 January 2018, collecting a total of 425 samples.