**************************************************************************************************************************************************** In this document you will find a description of the materials uploaded and the information required to reproduce econometric model results. We uploaded a read-me file plus the dataset in comma separated value (csv) format. **************************************************************************************************************************************************** Data collection title: Choice experiment for repairing rural waterpoints (1560 households, Kwale County Kenya, 2013-2014) Data collection abstract and methodology: This record contains water service preference data collected from 1,560 households in Kwale county on the south coast of Kenya. A sample of 531 handpump locations was used as a sampling frame for a household survey administered in late 2013 and early 2014 (see UKDS Reshare record: http://reshare.ukdataservice.ac.uk/853667/). 3,500 households took part in this survey of which a random draw of 1,560 households were selected to take part in a choice experiment on water service preferences. Choices included (1) maintenance service provider (public, private), (2) guaranteed days for repairs (2, 4, 6, 8), (3) cash management (treasurer/cash, bank account, mobile money), (4) monthly household payment (USD 0.5, 1.0, 1.5, 2.0). An orthogonal, main effects design generated 10 choice cards, each with two alternatives and a status quo option eliciting 10 choice responses. Participating households could also select a status quo option reflecting community maintenance and the local payment arrangements (commonly cash). The data is presented as prepared for a conditional logit model estimating the main attributes followed by interactions across four hypotheses of behavioural change: (a) multidimensional wealth, (b) education, (c) gender of respondent, and (d) household concerns. The read-me file describes steps required for estimation of the econometric latent class model specified by a discrete distribution of preferences to estimate heterogeneity. Ethical procedures Prior to data collection, research permits and approvals were obtained from the Government of Kenya’s National Council of Science and Technology and Oxford University’s Central University Research Ethics Committee. Disclosure assessment was conducted as part of the ethical procedures by Oxford University. In terms of ethical considerations, participation in the data collections campaigns was voluntary. Information about the study and its objectives was provided in all cases. Consent was obtained verbally (data collection campaigns were conducted in areas of low literacy in Kwale County). Consent discussion contained information about 1) the right to decline participation or withdraw participation at any point in time, 2) confidentiality procedures, 3) purposes of the research study structure, and 4) contact details for follow-up information. ***************************************************************************************************************************************************** Here we list the STEPS required to carry out the estimation of the econometric model. ***************************************************************************************************************************************************** ******************************************************* *********STEPS to carry out estimation***************** ******************************************************* To estimate the model reported in Table 1 of this paper you need to open the dataset labelled Base-web.csv in NLOGIT version 5. Once you have imported the data, please run the code described in STEP 2 below. ******************************************************* ********************STEP 1***************************** ******************************************************* Import the dataset Base-web.csv to NLOGIT. This includes 1,560 households with 10 choice cards per household, comprising three choice options per card, leading to a total of 46,800 rows. Variables are labelled as follows: *days_2: a value of 1 denotes preference for householdīs handpump repair in 2 days, 0 otherwise. *days_4: a value of 1 denotes preference for householdīs handpump repair in 4 days, 0 otherwise. *days_6: a value of 1 denotes preference for householdīs handpump repair in 6 days, 0 otherwise. *days_8: a value of 1 denotes preference for householdīs handpump repair in 8 days, 0 otherwise. *MAIN_DWO: a value of 1 denotes preference for a public (DWO) maintenance provider, 0 otherwise. *MAIN_PRI: a value of 1 denotes preference for a private maintenance provider, 0 otherwise. *STO_BANK: a value of 1 denotes preference for storing payments in a bank, 0 otherwise. *STO_MPES: a value of 1 denotes preference for storing payments through Mpesa, 0 otherwise. *PAY_50: a value of 1 denotes preference for payment equal to 50 Kenyan Shillings, 0 otherwise. *PAY_100: a value of 1 denotes preference for payment equal to 100 Kenyan Shillings, 0 otherwise. *PAY_150: a value of 1 denotes preference for payment equal to 150 Kenyan Shillings, 0 otherwise. *PAY_200: a value of 1 denotes preference for payment equal to 200 Kenyan Shillings, 0 otherwise. *lwpaym: denotes the interaction effect between wealth (index) and (log)payment for water. *lwdays: denotes the interaction effect between wealth (index) and householdīs handpump (log)days of repair. *mwealth1: denotes the interaction effect between wealth (index) and public maintenance provider. *mwealth2: denotes the interaction effect between wealth (index) and private maintenance provider. *SWEALTH1: denotes the interaction effect between wealth (index) and bank as storage option for payments. *SWEALTH2: denotes the interaction effect between wealth (index) and Mpesa as storage option for payments. *ledpaym: denotes the interaction effect between years of education and (log)payment for water. *ldedu: denotes the interaction effect between years of education and householdīs handpump (log)days of repair. *meduc1: denotes the interaction effect between years of education and public maintenance provider. *meduc2: denotes the interaction effect between years of education and private maintenance provider. *educyr1: denotes the interaction effect between years of education and bank as storage option for payments. *educyr2: denotes the interaction effect between years of education and Mpesa as storage option for payments. *lwseaspy: denotes the interaction effect between seasonality as a water concern and (log)payment for water. *lwseasdy: denotes the interaction effect between seasonality as a water concern and householdīs handpump (log)days of repair. *wseaso1: denotes the interaction effect between seasonality as a water concern and public maintenance provider. *wseaso2: denotes the interaction effect between seasonality as a water concern and private maintenance provider. *lwrelpy: denotes the interaction effect between reliability as a water concern and (log)payment for water. *lwreldy: denotes the interaction effect between reliability as a water concern and householdīs handpump (log)days of repair. *wreli1: denotes the interaction effect between reliability as a water concern and public maintenance provider. *wreli2: denotes the interaction effect between reliability as a water concern and private maintenance provider. *lwdistpy: denotes the interaction effect between distance to water point as a water concern and (log)payment for water. *lwdistdy: denotes the interaction effect between distance to water point as a water concern and householdīs handpump (log)days of repair. *wdista1: denotes the interaction effect between distance to water point as a water concern and public maintenance provider. *wdista2: denotes the interaction effect between distance to water point as a water concern and private maintenace provider. *lwlonpy: denotes the interaction effect between long queues as a water concern and (log)payment for water. *lwlondy: denotes the interaction effect between long queues as a water concern and householdīs handpump (log)days of repair. *wlongq1: denotes the interaction effect between long queues as a water concern and public maintenance provider. *wlongq2: denotes the interaction effect between long queues as a water concern and private maintenace provider. *lwsafepy: denotes the interaction effect between safety as a water concern and (log)payment for water. *lwsafedy: denotes the interaction effect between long queues as a water concern and householdīs handpump (log)days of repair. *wsafet1: denotes the interaction effect between long queues as a water concern and public maintenance provider. *wsafet2: denotes the interaction effect between long queues as a water concern and private maintenace provider. *lwcostpy: denotes the interaction effect between cost as a water concern and (log)payment for water. *lwcostdy: denotes the interaction effect between cost as a water concern and householdīs handpump (log)days of repair. *wcostt1: denotes the interaction effect between cost as a water concern and public maintenance provider. *wcostt2: denotes the interaction effect between cost as a water concern and private maintenance provider. *lgenpm: denotes the interaction effect between being a female respondent and (log)payment for water. *lgenday: denotes the interaction effect between being a female respondent and householdīs handpump (log)days of repair. *mgender1: denotes the interaction effect between being a female respondent and public maintenance provider. *mgender2: denotes the interaction effect between being a female respondent and private maintenance provider. *gend1: denotes the interaction effect between being a female respondent and bank as storage option for payments. *gend2: denotes the interaction effect between being a female respondent and Mpesa as storage option for payments. *lmtspay: denotes the interaction effect between being satisfied with current maintenance service and (log)payment for water. *lmtsdy: denotes the interaction effect between being satisfied with current maintenance service and householdīs handpump (log)days of repair. *mmts1: denotes the interaction effect between being satisfied with current maintenance service and public maintenance provider. *mmts2: denotes the interaction effect between being satisfied with current maintenance service and private maintenance provider. *msts1: denotes the interaction effect between being satisfied with current maintenance service and bank as storage option for payments *msts2: denotes the interaction effect between being satisfied with current maintenance service and Mpesa as storage option for payments ******************************************************* ********************STEP 2***************************** ******************************************************* *Run the code below using NLOGIT version 5. LCLOGIT ; Lhs = CHOICE ; Choices = 1,2,3 ; ; Rhs = days_2 ,days_4,days_6,days_8, MAIN_DWO,MAIN_PRI,STO_BANK, STO_MPES , PAY_50, PAY_100, PAY_150, PAY_200, lwpaym , lwdays , mwealth1, mwealth2, SWEALTH1,SWEALTH2, ledpaym, ldedu, meduc1, meduc2, educyr1, educyr2, lwseaspy, lwseasdy, wseaso1, wseaso2, lwrelpy, lwreldy, wreli1, wreli2, lwdistpy, lwdistdy, wdista1, wdista2, lwlonpy, lwlondy, wlongq1, wlongq2, lwsafepy, lwsafedy, wsafet1, wsafet2, lwcostpy, lwcostdy, wcostt1, wcostt2, lgenpm, lgenday, mgender1, mgender2, gend1, gend2, lmtspay, lmtsdy, mmts1, mmts2, msts1, msts2, ;LCM=one; ;pts=2; ;pds=10; ;parameters; ;halton$ ******************************************************* ********************END******************************** *******************************************************