This study uses disposal as a dependent variable in categorical form. The variable captures the various ways employed by Dar es Salaam residents in disposing their garbage. Specifically, it includes the rubbish pit inside compound, rubbish pit outside compound, rubbish bin, throwing out and other options. As long as the main focus of this study is to do household level analysis, the household related features are of major concern. Table 1 summarizes the descriptive statistics of the variables used in the analysis.
In Table 1, about 3,541 households were interviewed in Dar es Salaam city. Out of that, about 77% were male-headed household, with mean monthly expenditure per adult equivalent of Tanzanian Shillings (TZS) 52,988.45 (44.16 USD), and average household size of 3.7 members. The average age of the household head was 40 years. Only 37.5% out of households in the city own their homes which show that many household heads in the city does not own their homes. Furthermore, many households have family members aged between 15 and 65 years, which implies that majority are in the working age category. The average proportion of family members who are males was 0.53; whereas, on occupation, the majority are paid employees (44%) and self-employed (44%). Table 1 also shows that majority (61%) of the household heads can read and write, as they have attained primary school education; while, 6% only who cannot read and write have not been to school. Relatively small percent (4%) have attained a university education. The walking distance to access public transport (in terms of time) was 10 min on average. And among the three municipalities in Dar es Salaam city, about 41%, 34% and 25% of the households are from Kinondoni, Temeke and Ilala municipality, respectively.
As mentioned earlier, there are five alternatives or ways of disposing garbage employed by the study. Specifically, the “other alternative" includes options like thrown inside compound, informal waste pickers and burning, which in most cases are environmentally unfriendly. The five alternatives used in this study were basically taken as presented in the structure or design of the 2007 Tanzanian HBS questionnaire. In terms of proportion, Figure 1 presents a clear distribution of the five ways to dispose garbage in Dar es Salaam city.
Figure 1 clearly shows the majority of the households in Dar es Salaam city use rubbish bin as their most preferred way to dispose waste, with about 35% of the households. The less preferred alternative with almost 6% of the households was the use of other category to dispose waste apart from the four recognized in the study. And the proportions for the rest ranged between 35% and 6%. As far as SWM is concern, those households who cannot afford to buy the rubbish bins were
ideally supposed to use the rubbish pit rather than throwing or unrecognized alternative. In this case, the likely reason for not using the rubbish pit might be the shortage of land since the city is almost congested with unplanned settlements in most parts. This is evident in the Tanzania population census for 2002, which reveals about 356,286 people were living in Dar es Salaam in 1967 and it was estimated to be 2.5 million with a 4.3% annual growth rate in 2002. In the latest 2012 Census, Dar es Salaam was found to have a population of 4.4 million accounting for 10% of the total Tanzania Mainland population (Tanzania Census, 2013). The population density again, in 2002 and 2012 was 1,793 and 3,133 persons per square kilometre, respectively, in Dar es Salaam. This limits the land availability for rubbish pit establishment.
MNL and marginal effects estimation
Both the MNL model and marginal effects were estimated to identify the determinants of households’ choice on solid waste disposal practices in Dar es Salaam city. The marginal effects basically describe the marginal impact of a certain individual or household characteristics on solid waste disposal choice, and are evaluated at the mean of each characteristic. They also indicate the change in probability of falling in a particular alternative of disposing garbage over a unit change in the given explanatory variable at the mean values, keeping all other explana-tory variables constant. All these estimations were done through econometrics software -STATA version 13. The marginal effects are presented in Table 2, and it was found that some of the variables do not have any significant marginal effect on ways to dispose garbage, although their coefficients were significant in the general MNL estimation (Table 3).
The results show that, the household expenditure positively and negatively influence the choice of rubbish bin and rubbish pit outside compound, respectively, as ways of disposing garbage in Dar es Salaam city. The significant influence, however, is small in margin terms almost negligible as it shows that a unit increase (1 USD) in household expenditure increases the likelihood of using rubbish bin and reduces the likelihood of using rubbish pit outside compound for solid waste disposal by less than 1% in both alternatives, respectively (Table 2). With expenditure as a proxy for household income, those households with higher income are likely to buy the rubbish bins for proper waste disposal. The influence of income on waste management in this study is consistent with other studies, for example Chen (2010), Tadesse et al. (2008) and Tadesse (2009).
The age of the household head had significant marginal effect only in using rubbish bin (p<0.05). So, one-year increase in the age of the household head increases the probability of choosing rubbish bin by 0.4% (Table 2). Therefore, as the age of the household head increases, the likelihood of using rubbish bin also increases. Consistently, the other findings (Abebaw, 2008; Oguntayo and Obayelu, 2013) show a significant effect of age on waste management. Whereas, the marginal effects of distance on choosing rubbish pit both inside and outside compound were found to be positive and statistically significant (p<0.05). Both inside and outside rubbish pit were more likely to be chosen by households staying far from the
main road. However, the expectation was, as distance increase a particular household decides to throw garbage outside, considering that the households is "interior" in terms of accessibility which hinders the vehicle to pass and collect garbage. This expectation is supported by other authors (Oguntayo and Obayelu, 2013; Tadesse et al., 2008; Tadesse, 2009). However, the significant margin term is small, since the probability of choosing rubbish pits both inside and outside compound increases by 0.3% for a unit increase in walking distance, that is, one minute (Table 2).
Looking on the education variable however, it seems to be consistent with a priori anticipation. Post-secondary schooling and university education negatively affected the alternative of throwing garbage outside, and very significant (p<0.01). Meaning that, as compared to no education, those with university education and post-secondary education are less likely to throw outside their garbage, but insignificant for lower education level. In margin terms, those with university education were less likely to throw garbage outside by 15% as compared to those with no education. Similarly, being in post-secondary schooling category reduced the likelihood of throwing garbage by 12% (Table 2). It means that as household head become educated, they tend to choose better ways of disposing waste, also those household head with high level of education like university education are likely to have higher income, and therefore can decide to live in more organized places and opt for better solid waste disposal practices. Consequently, the significant effect of education level on proper waste disposal was also observed in previous papers in Africa (Abebaw, 2008; Oguntayo and Obayelu, 2013; Tadesse, 2009).
The effect of household occupation was positive and statistically significant for only the paid employee category (p<0.10) on choosing rubbish pit outside compound as a major means of solid waste disposal. It implies that paid employees are more likely to use the rubbish pit outside compound as compared to those who are engaged in crop farming and other agricultural activities. Being a paid employee increases the likelihood of adopting rubbish pit outside compound by 7.4% (Table 2). Oguntayo and Obayelu (2013) supports the finding on occupa-tion level affecting proper waste disposal. This can be due to the fact that most of the employed people have attained education and acquired knowledge. Therefore, awareness of the negative environmental impacts can be realized through that knowledge; also through the earnings from, they can employ better ways of disposing garbage like the use of pit and bin.
Among the four proportions of family members according to age, it is only proportion of family members aged above 65 years that have statistically significant marginal effect in using rubbish bin (p<0.05). Suggesting that, a unit increase in proportion of family members aged above 65 years reduces the probability of using rubbish bin by almost 36% (Table 2). In other way, households with many younger members than elders prefer to dispose garbage properly. And perhaps because the old people other way, households with many younger members than elders prefer to dispose garbage properly. And perhaps because the old people are likely to be of less concern with environment as compared to the youths since the youths are still depending on the environment for some years to come. However, it has been observed earlier that the old household head contrarily prefers rubbish bin relative to the rest of the options. Similarly, the marginal effect of proportion of females at the household was statistically significant against throwing (p<0.10). In margin term, as the proportion of females at home increases, the likelihood of throwing out garbage decreases by 8.3% (Table 2). It is consistent with prior expectation as long as females are "traditionally" more sensitive with the surrounding environment at home as compared to males. The household size in this study was not significant as compared to other studies for example (Abebaw, 2008).
Furthermore, the marginal effect for home ownership variable was positive and statistically significant in using rubbish pit inside compound (p<0.01), while negative and statistically significant in using rubbish bin (p<0.01). Those who own a home were more likely to use rubbish pit inside compound than using rubbish bin as compared to those who does not own their home. In marginal impact, those who own their homes are 13% more likely to use rubbish pit inside compound, while reducing the likelihood of using rubbish bin by 10.7% (Table 2). The finding on home ownership was consistent with other studies, for example Oguntayo and Obayelu (2013) on SWM in Nigeria. As expected, those who own their homes are very conscious with the surrounding environ-ment, which makes them to adopt better ways of disposing garbage. The use of a rubbish pit inside compound, also, might be attributed to the space availability at home place. On municipalities, as com-pared to Temeke municipality, the marginal effects on Kinondoni and Ilala gives a similar pattern in choosing rubbish pit outside compound and other alternative. As compared to Temeke municipality, the households in Kinondoni and Ilala are less likely to use the rubbish pit outside compound, but more likely to use other alternative. Also, households in Ilala are likely to throw their garbage as compared to household in Temeke.
Predicted probabilities
This section present and discuss the various figures for predicted probabilities across three explanatory variables. In particular, the explanatory variables include: age of the household head, distance to access public transport and the monthly expenditure per adult equivalent. The predicted probabilities were computed at different values of explanatory variable by keeping all other variables at their mean values.
Age of the household head
Figure 2 shows that the predicted probabilities for rubbish bin increase with the age of the household head. It starts to increase at small rate and later from the age of 40 years and above increases at a high rate. This demonstrates a clear positive relationship with age as age increases household head tend to use rubbish bin, consistently with findings in Table 2. While the predicted probabilities for throwing garbage outside seems to move against that for rubbish bin. It starts by decreasing at lower rate and then after 40 years, starts to fall sharply with age. This negative relationship between age and predicated probabilities for throwing means that as age increases, the predicted probabilities for throwing tend to decrease. Again, the predicted probabilities for using rubbish pit and other option decreases as the age of the household head increases, however, it is at a decreasing rate. It suggests that age of the household head has little effect or does not explain significantly the above alternatives as compared to the use of rubbish bin and throwing out option.
Distance to access public transport
Predicted probabilities for distance in Figure 3 has two sides in terms of direction that are clearly observed. These directions shows that the probabilities for some of the alternatives are increasing with distance in one pattern and the rest are decreasing in the second pattern.
The probabilities of using rubbish pit in both ways are increasing with distance. These patterns support the earlier results in Table 2. In the figure however, the probabilities of using rubbish pit inside compound increases continuously as time to access public transport increases, while that for rubbish pit outside compound increases but later starts to fall down slowly as distance increases. Other pattern is for the probabilities of using rubbish bin, throwing and other which are decreasing with distance and was not significant in Table 2. These predicted probabilities are falling down sharply especially from 200 min onwards. But, the area before 100 min has predicted probabilities that are coming across 0.2 for almost all alternatives (Figure 3). This suggests that when the distance is short from the household to the main road, the probability of choosing ways to dispose garbage does not differ significantly.
Monthly expenditure per adult equivalent
In Figure 4, clearly there is a unique pattern, since it is only predicted probabilities for rubbish bin which increases with monthly expenditure per adult equivalent, but the rest seems to fall slowly approaching zero predicted probabilities. In earlier results (Table 2), the expenditure variable was positively significant in using rubbish bin (p<0.10) and negatively significant in using rubbish pit outside compound (p<0.10). The predicted probabilities of using rubbish bin increases at a high rate towards one, while that for using rubbish pit inside compound, pit outside compound, throwing and other, stays below the predicted probability of 0.2. Among these, which are below 0.2, at least the curve for rubbish, pit outside compound falls down with significant rate before TZS 300,000 (250 USD). The trend of using rubbish bin is increasing with expenditure, since those with high income are likely and capable to buy rubbish bin instead of adopting other ways. Also, the household heads with high education level are likely to have high earnings (high expenditure), which make them to adopt better ways to dispose their garbage.
of household in choosing alternative j; βj is a vector of the estimated parameters associated to the alternative j. Simply by writing in odds ratio will be
as partial derivatives, were computed from Equation 4 for easy interpretation. For instance, the partial derivative with respect to 
predictor is presented as:
