In ternationa l Scholars Journa ls African Journal of Environmental Economics and Management ISSN 2375-0707 Vol. 5 (6), pp. 341-347, November, 2017. Available online at www.internationalscholarsjournals.org © International Scholars Journals Author(s) retain the copyright of this article. Full Length Research Paper Willingness to Pay (WTP) for pipe-water connection in Makululu compound of Kabwe, Zambia Yordanos Gebremeskel1, Nyambe Mubita2, Bupe Simuchimba1, James Mulenga1 1 Mulungushi University. School of Social Sciences. Department of Economics. Box 80415. 2 University of Barotseland. School of Natural Resources. Accepted 05 October, 2017 The availability of safe drinking water is dependent on better and improved water sources for households. Access to safe drinking water especially piped water is one of the challenges faced by households in Zambia. This study aims to understand the factors that determine the willingness to pay for pipe water connection on low income peri-urban settlements in Zambia. Double bounded questions which have the advantage of including a follow up dichotomous question after the first dichotomous choice question are used. An interval regression model is used to analyze the collected data. The variables of interest are purification, family size, water quality, education and income. The results obtained show that most of these variables have the expected signs. Further, the results suggests that an increase in income will lead to an increase in the willingness to pay to have access to safe drinking water sources. The results also suggest that the distance travelled to fetch water from a community tap and level of household education play a role in the willingness to pay for access to pipe water connection. We estimated that the sample household mean WTP is K283.77 ($38.14) which is roughly 33% of the commercial connection charge. Keywords: Contingent valuation, double-bounded, interval regression, WTP, Zambia. INTRODUCTION Water is essential for human survival. It is for this reason that in 2010, the United Nations General Assembly declared safe and clean drinking water and sanitation a human right essential to the full enjoyment of life and all other human rights (UNICEF, 2014). Notwithstanding its importance, access to improved water still remains a challenge to access safe drinking water. As a result, most people continue to be exposed to waterborne diseases such as cholera, diarrhoea, dysentery, typhoid fever among others. According to the World Health Organisation (WHO), water borne diseases are among the major causes of morbidity and mortality in most developing countries particularly among children under five years old. Contaminated drinking water is estimated to cause 502, 000 diarrhoea death each year (WHO, 2016). Corresponding author E-mail: yorsink@yahoo.com; yordanosge@gmail.com Zambia has made progress in increasing access to improved drinking water. The proportion of the population with access to an improved source of drinking water increased from 24 percent in 2007 to 65 percent in 2014 (ZDHS, 2015). Table 1 below shows that, the percentage of total population piped into dwelling/yard/plot stands at 16 percent in 2014. The progress has been necessitated by the various programmes that the government together with its cooperating partners have implemented such as the National Rural Water Supply and Sanitation Program (NRWSSP) and National Urban Water Supply and Sanitation Program (NUWSSP). Despite this progress, access to improved drinking water in general and piped water in particular continues to be a challenge in Zambia. In comparison to its neighbouring countries, as is shown in Table 2 (Annex), Zambia has among the lowest levels of access to drinking water. The health and time saving benefits of access to clean water, as private or communal, by households in developing Gebremeskel et al. 342 countries have been widely documented. Our main objective in this study is to assess their willingness to pay for private pipe-water connection. This is because of the fact that households in our study site already had access to clean water from communal taps supplied by the water utility company. Thus study sought to estimate willingness to pay for improved water supply through pipe water connection in Bwacha constituency, Kabwe, and to determine the factors (variables) that influence households’ WTP for improved water supply. Makululu is an unplanned settlement characterised by lack of social amenities such as piped water among others. The Contingent Valuation Method (CVM) is used to estimate monetary measure of preference by directly asking respondents’ willingness to pay for changes in quantity or quality of a good or service ((Haab & McConnell, 2002)). The Dichotomous Choice Contingent Valuation Method (DC–CVM) can be presented in different forms. The single bounded DC–CVM asks respondents for a predetermined bid amount for their vote or rejection. Alternatively in double bounded format we offer a follow-up price (bid) for the respondent after getting a response for the initial bid. The follow-up bid in double bounded take a higher value if the respondent’s answer for the first bid was YES and a lower value if the answer was NO. Estimation of household’s demand for water usually conducted by capturing key indicators of household characteristics. Nauges and Whittington (2010) well documented the various empirical researches made on water demand in various countries. Recent notable studies in African countries, using various methodologies, have established that income, family size, time spent to fetch water from existing sources, level of education, age of respondent, bid value, perceived quality of current water supply are the main factors influencing households WTP for improved water supply services (Gulyani et al. 2005; Tarfasa and Brouwer, 2013, 2015; Genius et al 2008.) Despite the fact that various studies have been undertaken to determine the willingness to pay and the factors that determine the willingness to pay (WTP) for quality water in developing countries around the world, we didn’t find enough literature relating to WTP for improved water in Zambia particularly employing the CVM. One particular study done in the North Western Province of Zambia by Aymar (2004) using descriptive analysis concluded that willingness to pay depends to a large extent on income level. The higher the income, the higher willingness to pay. The paper proceeds as follows. In the next section we present the model of willingness to pay. Section 3 discusses study site and sampling technique. Sections 4, 5 and 6 are for results, discussion and conclusion respectively. METHODOLOGY Willingness to Pay Model An application of a contingent valuation questionnaire using the dichotomous choice model renders two answers. These are dichotomous answer ( if the individual answers yes and if the answer is no) given a question about paying a predetermined amount of . The WTP can be estimated by modelling as the following linear function: (1) Where is a vector of explanatory variables, is a vector of parameters to be estimated and is the error term. The respondent will answer yes when his WTP is greater than the suggested amount that is when . WTP and Double-bounded contingent valuation We can assume that an individual is asked if he is willing to pay an amount for a given change in the provision of a given good or service. If the individual answers no then we can infer that , if he answers yes then . But Hanemann et al. (1991) suggested that the follow-up dichotomous question can be asked after the first dichotomous question. This approach, called double- bounded models, is assumed to generate more efficient estimation than a single question or a single question followed by an open ended question. In the case of double bounded model, a follow-up dichotomous question is asked after the first dichotomous choice question. If the individual answers yes to the first question then he/she is asked about his WTP for a higher amount. If he answers no to the first question then a lower amount is offered. The second questions is endogenous in the sense that the amount asked depends on the answer obtained for the first question which is exogenous. The method renders with two answers for each respondent. If we call the first bid amount and the second one , where then i. The individual that answers yes to the first question and no to the second, then . This implies . ii. If the answers are yes for both questions, then . iii. If the individual’s answer are no to the first question and yes to the second, then . iv. Finally, if the individual’s responses are no for both questions, then we will have Econometric estimation of double-bounded or interval data model Let’s define and as the dichotomous variables that capture the response to the first and second closed questions, then the probability that an individual answers yes to the first question and no to the second can be 343 Afr. J. Environ. Econ. Manage. Table 1. Use of drinking water sources (percentage of population) as of 2015: selected countries. Country Urban Rural Total Total improved Piped on premises Total improved Piped on premises Total improved Piped on premises Botswana 99 96 92 45 96 74 South Africa 100 92 81 38 93 73 Namibia 98 69 85 34 91 51 Malawi 96 33 89 3 90 8 Zimbabwe 97 74 67 5 77 28 Swaziland 94 75 69 27 74 37 Zambia 86 36 51 2 65 16 Congo, Dem. Rep. 81 17 31 1 52 8 Mozambique 81 25 37 1 51 9 Source: UNICEF and WHO, 2015. Table 2. Socio-economic characteristics of surveyed households. Variable Name Description Frequency Percent Gender Gender of the respondent Female 63 42 Male 87 58 Education Educational level of the respondent No formal education 8 5.3 Primary 46 30.7 Secondary 68 45.3 Tertiary 28 18.7 Family size Number of family members living in the household Less than 5 84 56 5 up to 10 62 41.3 More than 10 4 2.7 Income Average monthly income Less than K100 9 6 K100 – K150 30 20 K150 – K200 30 20 More than K200 80 53 expressed as .Given this and under the assumption that and , the probability of each one of the cases is given by: Using symmetry of the normal distribution: The other three probabilities, the probability that an individual answers are also captured in a similar fashion. Then, a log-likelihood function can be used to estimate 𝛽 and σ using maximum likelihood estimation. Where , , and, are the dummy variables 1 and 0 denoting the group to which the i th respondent Gebremeskel et al. 344 belongs. A given respondent contributes to the logarithm of the likelihood function in only one of its four parts. Description of study area Our study is conducted in Kabwe, one of the 11 districts of the Central province of Zambia. The specific locality, Makululu Compound in Bwacha constituency, is an unplanned settlement located in Kabwe district of lying between latitude 14° 27' S and longitude of 28° 27'E. It is along the Great North Road, 139 kilometers North East of Lusaka. Kabwe was established as a Lead and Zinc mining town around 1902 and almost all of the mines were closed in mid 1990s that created high unemployment and poverty level on sizable residents. The total population of Bwacha constituency in 2010 was85, 397that belongs to 17,185 households (CSO, 2011:25). The settlement can be described as poor housing units made of mud, with poor sanitation and safe drinking water supply. Poor sanitary practices, like disposal of human excreta usually leads to contamination of water and therefore water-borne diseases Phiri (2016). Majority of the population are not in the formal employment sector with low levels of professional skills and education. We have applied a stratified and systemic sampling techniques for the primary data collection. There are five wards in Makululu compound and three wards were considered to reach 150 respondents. These were Makululu ward, Chililalila and Moomba Ward. A pre-test was conducted on 5 households to confirm the applicability of the questionnaire in general and the contingent valuation technique in particular. It was found that all the questions including the WTP bids were easily understood by the respondents. For the final data collection, proportionate stratified sampling was applied and each ward had fifty households sampled. RESULTS Descriptive statistics This section discusses both the descriptive statistics and the regression estimation. It also includes explanations of the variables which are used for describing and estimating the willingness to pay for pipe water connection in Bwacha. Furthermore, the section discusses the relevant variables used for the WTP estimation. The respondents were given a brief introduction of the benefits from connecting to a commercial water pipe line. A total of 150 respondents were asked double bounded questions with K200 being the initial bid and K100 and K500 were the follow-up bids. We did not find anyone opposed the proposed project of private pipe water connection. Table 3 (see Annex)shows frequency and percent of selected variables to describe the sample respondents. The sample respondents are more male, educated, and moderate family size & income. Table 4 (see Annex) shows that majority, 70%, of the respondents are fetching drinking water from communal taps built by the utility company Lukanga Water and Sewerage. The responses on the assessment of water quality shows that 65(43.3%) and 83(55.4%) of the respondents feel that the water quality is good and very good respectively. This perception of the quality of drinking water also reflected by high percentage, 86.7%, of respondents who are not using any water purification method. For three quarter (75%) of the respondents it takes more than 7 minutes to fetch water from the nearest communal tap. Out of this percentage, 20% of the respondents need to walk for more than 30 minutes to access the communal tap. A more than 10 minutes walking is usually estimated as more than 1 kilometre (0.6 miles) with a normal walking style of an adult person. Majority of our respondents have access to communal taps (70%) as their primary source of drinking water and they perceived the quality as good and very good (98%). The high rating of the water quality correlate with less tendency to use water purification techniques (13%). Though the majority of the households have access to communal tap water, they need to walk more than a kilometre. Given most of them need to fetch water more than once per day, the time cost of fetching water is high. Table 5 below shows frequency and percentage of the double bounded responses as ‘yes-yes’, ‘yes-no’, ‘no- yes’ and ‘yes-yes’. The bids were preceded by WTP question stated as: ‘It is difficult to state the actual cost of connecting your household to the nearest water network. The prices stated below are chosen simply for the purpose of our study. If you are willing to be connected to the water system, are you willing to pay a one-off out of pocket payment?’ Table 6 shows summary statistics of each independent variable used in the interval regression model which is presented in table 7. These variables are expected to affect the WTP for safe drinking water. Water quality, family size, education level and income are expected to have a positive sign in the regression model. Regression Estimation Table 7 presents results from a double bounded interval- regression estimation. We made the estimation using Stata’s intreg censored regression command. The table presents the coefficient (marginal WTP), standard error, significance level, and confidence interval. We found that the overall model is statistically significant as measured by log likelihood and wald chi square. Most of the estimated coefficients have the expected signs. According to the estimation result, education level and income level of the household are positive in sign and statistically significant at 5% and 1% respectively. 345 Afr. J. Environ. Econ. Manage. Table 3. Water source, perceived quality, purification and estimated time to fetch. Variable Name Description Frequency Percent Source Source of drinking water Hand dug well 2 1.3 Borehole 43 28.7 Communal tap 105 70 Water quality Respondent’s perception of current water quality Poor 2 1.3 Good 65 43.3 Very Good 83 55.4 Purification Water purification method used Chlorination 11 7.3 Boiling 7 4.7 Filtration 2 1.3 No purification 130 86.7 Time Time taken to fetch water from the nearest source Less than 7 minutes 38 25.3 15 – 20 minutes 51 34 20 – 30 minutes 30 20 30 – 40 minutes 19 12.7 More than 40 minutes 12 8 Table 4. Distribution of double bounded responses for initial amount of K200 and follow-up K500. Responses Number of responses Percentage Yes –Yes 33 22.0 Yes – No 46 30.7 No – Yes 33 22.0 No-No 38 25.3 Table 5. variables used in the regression and their expected signs. Variable Type Description Mean Std. Dev Sign Educational level Ordered categorical 0 if primary; 1 if basic; 2 if secondary; 3 if tertiary 1.782 .066 + Income per month Ordered categorical 0 if K200 2.217 .079 + Family size Ordered categorical 0 if 1 – 5; 1 if 5 – 10; 2 if >10 0.60 .057 -/+ Distance from the nearest community tap (in minutes) Ordered categorical 0 if <7 ; 1 if 15-20; 2 if 20-30; 3 if 30-40; 5 if >40 1.414 .098 + Purification Nominal 0 = no purification; 1=purification .92 .021 + Whereas distance has a negative sign and is statistically significant at 5% Mean WTP Estimation The estimation of mean WTP is made using the mean values and the marginal WTP values of the variables used in the regression estimation. The estimated mean WTP for the 147 sampled households is K283.7748 ($38.14). The estimate of the mean WTP for the pipe water connection can be used to estimate the total benefits in the specific locality. Consequently, we attempted to expand the sample WTP estimate for the population of Bwacha. According to the Central Statistics of Zambia (CSO, 2011), there were 17,185 Gebremeskel et al. 346 Table 6. Interval regression estimation. Interval regression Number of obs = 147 Wald chi2(5) = 47.31 Log pseudolikelihood = -218.0028 Prob> chi2 = 0.0000 Coef. Robust Std. Err. z P>|z| [95% Conf. Interval] educ 50.81731 22.35321 2.27 0.023* 7.00583494.62879 income 78.70924 14.73437 5.34 0.000** 49.83042107.5881 famsize 29.74863 23.38308 1.27 0.203 -16.0813775.57863 distance -98.20083 41.97465 -2.34 0.019* -180.4696-15.93202 purify -73.84673 79.13106 -0.93 0.351 -228.940881.2473 _cons 143.1545 97.65233 1.47 0.143 -48.2405 334.5496 /lnsigma 5.220299 .0648312 80.52 0.000 5.093232 5.347365 Observation summary: 0 left-censored observations 0 uncensored observations 32 right-censored observations 115 interval observations *Significant at 5%, ** significant at 1%. Table 7. Estimated mean willingness to pay (WTP). Mean WTP [95% Conf. Interval] 283.7748 (USD 38.14) 248.7927 318.7568 Table 8. Aggregate willingness to pay (WTP) of tap water connection in Bwacha, Zambia. Bwacha residents household mean WTP [95% Conf. Interval] 4, 876, 669 (USD 655, 466) 92308.54 93210.62 households in Bwacha in 2010. Multiplying the 810 households by the mean WTP and it yields a total of approximately K4, 876,669 ($655,466), as shown in the following table. DISCUSSION OF RESULTS Interval regression model was used to investigate factors determine household’s WTP for tap-water connection. The analysis result also includes the estimation of WTP for sample households and the population. The interval regression model shows that education level and household’s monthly income are the significant explanatory variables at 5% and 1% respectively. This is consistent with other studies (Jianjun et al. 2016; Parveen et al 2016; Twerefou, et al. 2015; Mezgebo and Ewnetu, 2015; Coster and Otufale 2014; Kwak et al. 2013; Khan et al. 2010; Mbata 2007) which showed that education and income are significant determinants of WTP. Education enlightens people about the importance clean water while income provides the households with ability to pay for the water. The type of purification technique used, the assessment of current water quality and family size found to be statistically insignificant to WTP for private water pipe-line connection. It is found that the mean WTP for sample households is K283.77. This amount is a little higher than the estimated average households’ income that falls in the range K150 – K200. This confirm that the household’s expression of willingness to pay is within their capability to pay. Depending on the sign and magnitude of the coefficient of the variable, it is possible to forecast the change in WTP for a unit change of the variable concerned. For instance, the income variable is measured as ordered category between 0 and 3, as one unit change represent a change in income of K50. Therefore, based on the mean WTP estimation, a one- unit increase in income category adds K78.70 to WTP. The major contribution of this research is that it adds to the literature on willingness to pay estimation for Zambia where there remain evidences of WTP are scarce. To the best of the authors’, this study is the first of its type to 347 Afr. J. Environ. Econ. Manage. apply double bounded discrete choice model with interval regression to pipe water connection in Zambia. CONCLUSION AND POLICY IMPLICATIONS In this study we examine to what extent characteristics of the respondents (the household head) determine the likelihood that they are willing to pay a given amount of money for pipe water connection. The study used a double bounded contingent valuation method to estimate willingness to pay (WTP) for pipe water connection in Bwacha constituency, Zambia. The main objective of this study was to understand the factors that determine the willingness to pay for safe drinking water in a low income urban settlement in Zambia. An interval regression model is used to estimate coefficients of the variable and the WTP. The regression estimation result shows that most of the variables found to have expected signs. Among the five variables considered, two variable, namely income and education level of the respondents are statistically significant in determining the WTP for safe drinking water connection. The mean WTP of income is estimated to be K283.77 for a household. The Lukanga water company, the sole utility company in Kabwe, has a fixed pipe water connection rate based on the distance between a house and the nearest main pipe-line. Those houses within 10 meters distance from the main pipe-line are charged K852 and the rate increases as the house located more than the 10 meters distance. Our estimation of mean WTP is roughly 33% of the connection charge. The findings of the study are vital for utility company, development partners and local government authorities. The connection charge for most households is ‘expensive.’ Nevertheless, our research found that, with some financial or reduced connection charge, it is possible to provide access to safe drinking water to many urban dwellers who are deprived of these essential life sustaining resource for many years. Attainment of more years of formal education plays a paramount role in creating awareness about the many advantages of living a healthy life by committing resources by acquiring utilities like water. The study demonstrated the applicability of contingent valuation method to assess the demand and willingness to pay for access to safe drinking water in developing country like Zambia. Future research works should attempt to use other WTP estimation techniques to extend our findings. DISCLOSURE The authors declare no conflict of interest. REFERENCES Aymar, M. (2004). Ability and Willingness to Pay Study NWWSSC, Commercialisation of Water Supply and Sewerage Utilities in the North-western and Southern Provinces. Zambia: GTZ. Celine Nauges, D. W. (2010). Estimation of water demand in Developing countries. World Bank. Central Statistical Office. (2011). Zambia 2010 Census of Population and Housing. Lusaka, Zambia. Central Statistical Office. (2015). Demographic and Health Survey . Lusaka, Zambia: CSO. Genius, M., Hatzaki, E., Kouromichelaki, E. M., Kouvakis, G., Nikiforaki, s., & Tsagarakis, K. P. (2008). Evaluation consumers' willingness to pay for improved potable water quality and quantity. Water Resources Management, 22(12), 1825 - 1834. GUHA, S. (2007). Valuation of clean water supply by willingness to pay method in a developing nation: a case study in Calcutta, India. Journal of Young Investigators,17. Gulyani, S., Talukdar, D., & Kariuki, R. M. (2005, July). Universal (NOn)service? Water Markets, Household Demand and the Poor in Urban Kenya. Urban Studies, 42(8), 1247 - 1274. Haab & McConnell. (2002). Valuing Environment and Natural Resources. Massachusetts, USA: Edward Elgar Publishing, Inc. Hanemann, W. M., Loomis, J., & Kanninen, B. J. (1991). Statistical efficiency of double bounded dichotomous choice conitngent valuation. American Journal of Agricultural Economics, 73, 1255 - 1263. Lopez-Feldman, A. (2012). Introduction to contingent valuation using Stata. MPRA. Toluca, Mexico: Centro de Investigacion Docenicia Economicas (CIDE0. Phiri, A. (2016). Risks of domestic underground water sourses in informal settlment in Kabwe. Environmental and pollution, 5(2). Solomon, T., & Roy Brouwer. (2013). Estimation of the public benefits of urban water supply improvements in Ethioia: a choice experiment. Applied Economics, 45, 1099-1108. Twerefou, D. T. (2015). Willingness-to-Pay for Potable Water in the Accra-Tema Metropolitan Area of Ghana. Modern Economy, 6, 1285-1296. UNICEF. (2014). The rights to safe water and sanitation: current issues. No 3. USA. World Health Organization. (2016). World Health Statistics. France. Zambia Demographic and Health Survey (ZDHS). (2015). Demographic and Health Survey. Lusaka, Zambia. .