WPS6321 Policy Research Working Paper 6321 Challenges and Opportunities of Mobile Phone-Based Data Collection Evidence from South Sudan Gabriel Demombynes Paul Gubbins Alessandro Romeo The World Bank Africa Region Poverty Reduction and Economic Management Unit January 2013 Policy Research Working Paper 6321 Abstract The proliferation of mobile phones in developing much higher for respondents who owned their own countries has generated a wave of interest in collecting phones. Both compensation provided to respondents high-frequency socioeconomic surveys using this in the form of airtime and the type of phone (solar- technology. This paper considers lessons from one such charged or traditional) were varied experimentally. The survey effort in a difficult environment—the South type of phone was uncorrelated with response rates and, Sudan Experimental Phone Survey, which gathered contrary to expectation, attrition was slightly higher for data on living conditions, access to services, and citizen those receiving the higher level of compensation. The attitudes via monthly interviews by phones provided to South Sudan Experimental Phone Survey experience respondents. Non-response, particularly in later rounds suggests that mobile phones can be a viable means of of the survey, was a substantial problem, largely due to data collection for some purposes, that calling people on erratic functioning of the mobile network. However, their own phones is preferred to handing out phones, and selection due to non-response does not appear to have that careful attention should be given to the potential for markedly affected survey results. Response rates were selective non-response. This paper is a product of the Poverty Reduction and Economic Management Unit, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at gdemombynes@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Challenges and opportunities of mobile phone-based data collection: Evidence from South Sudan Gabriel Demombynes1 Paul Gubbins1 Alessandro Romeo2 JEL Codes: C81, C83, D04 Keywords: South Sudan, mobile phone, survey methodology Sector Board: POV Affiliations: 1 World Bank 2 University of Rome “Tor Vergata� The authors greatly appreciate the collaboration of both the South Sudan National Bureau of Statistics and Horizon Contact Centers in Nairobi on the project described in this paper. The authors thank Andrew Dabalen, Johannes Hoogeveen, and Alvin Etan Ndip for helpful comments. The views expressed in this paper are the authors’ alone. They do not necessarily reflect the views of the World Bank or its Executive Directors. Demombynes and Gubbins are part of the Poverty Reduction staff of the Poverty Reduction and Economic Management (PREM) unit of the World Bank’s Kenya office. Corresponding email: gdemombynes@worldbank.org 1. Introduction In the past decade, mobile phones have become a ubiquitous feature of life in developing countries. Between 2005 and 2011 mobile cellular subscriptions nearly tripled in the developing world, increasing from 1.2 to 4.5 billion. In Africa, the region with the fastest mobile subscription growth rate, mobile cellular subscriptions increased from 87 million in 2005 to 433 million in 2011. A number of factors contributed to the luxury-to-commodity transformation of mobile phones. Technological advances along with economies of scale in network equipment and handsets in the developed world led to price declines. Additionally, the introduction of pre-paid billing systems lowered service costs by eliminating the need for operators to send bills and collect debts, thus making mobile phones accessible to lower income consumers. Finally, deregulation created a competitive marketplace, further reducing prices and bolstering mobile phone adoption (Economist 2009). Accompanying the growth in handset ownership has been growth in mobile applications, including digital money transfers and payments, citizen polling station surveillance, remote health care consultation and diagnosis, and transmission of timely market information to farmers. One area in which mobile technology holds promise is for implementing surveys. Despite recent advances, in many developing countries there is a lack of reliable and timely socioeconomic data. Today, there is an opportunity to leverage mobile phones for routine data collection to help fill this gap and support policy decisions, strengthen civil society, and facilitate program management (Croke et al. 2012). This paper examines the opportunities and challenges of using mobile phones as a research tool, based on the experience with the South Sudan Experimental Phone Survey (SSEPS). The paper is organized as follows. Section 2 of the paper describes options for data collection via mobile phones, Section 3 describes the South Sudan survey, Section 4 assesses patterns of non-response and survey results, and Section 5 concludes. 2. Mobiles for Data Collection For mobile phones to be useful as a survey device, they need to satisfy a number of criteria. A mobile- based survey should 1) yield a sample size large enough to draw statistically reliable inferences about a population of interest, 2) be economically viable compared to alternative data collection, and 3) generate data of adequate quality. The third criterion relates to sampling, attrition, and non-response as well as respondent behavior when interacting with a mobile phone interface. 2 Data applications for mobile phones have been used for cross-sectional monitoring and evaluating conditional cash transfer programs in Guatemala (Schuster and Brito 2011), longitudinal research on farmer’s understanding of uncertainty in Tanzania (Dillon 2010, Schuster and Brito 2011), enumeration of households by community health workers (Tomlinson, Solomon et al. 2009), and longitudinal third- party monitoring of public services in Tanzania. 1 Approaches vary in terms of the means of data collection (voice, SMS, USSD, electronic forms) and server side components (databases, data reporting, and management interfaces) (Loudon 2009). These choices relate to budget, socioeconomic context, and available technology services and infrastructure. Figure 1 provides a schematic diagram of data collection approaches that are available in addition to traditional, face-to-face paper-based surveys, and Table 1 lays out the strengths and limitations of various technologies used in the approaches. Figure 1: Survey Data Collection Approaches Notes: See Annex A for a full glossary of abbreviations. SMS refers to standard mobile text messages. GPRS is the channel used for mobile Internet on a smartphone.USSD is a less common channel which allows for two-way data communications on a basic GSM phone. It is typically used for requesting information on airtime balances. For the second approach shown in the figure—face-to-face interview using phone, laptop or tablet computer—data could be transmitted via GPRS or by other means, such as physical transfer of electronic media to a central location. 1 For information on the third-party monitoring in Tanzania, see the program website at www.listeningtodar.org 3 Table 1: Strengths and Limitations of Mobile Data Collection Client Options Data Data Collection Transfer Strengths Limitations Client / Interface Method - Simple voice call (users will be familiar with the interaction) - Operators can troubleshoot during - Requires additional management Voice the call, e.g. clarify questions, correct of call-center operations Call Center misinterpretations) - Requires a reliable network signal - Opportunity to build relationship with respondents over time - Good for setting where literacy is low - Simulates a voice call - Cumbersome nature of keypad Interactive Voice Voice/ - Good for settings where literacy is based menu navigation Response (IVR) Keypad low - Requires a reliable network signal - Requires a network signal - Requires that users have a cue-card to guide text input (changes to questions on an ongoing basis may be difficult) - Possibility of lost or dropped SMS - Works on any GSM mobile phone messages SMS (Does not require installation) - Limited to 160 characters of data - Relatively high cost per byte of data sent compared to GPRS - Data collection efforts with high expected data volumes will require a commercial SMS gateway provider (entailing additional costs) - Requires partnership with mobile - Works on any GSM mobile phone service operators (Does not require installation) - Requires a consistent network - Widely used in East Africa for mobile USSD USSD signal transactions - Relatively high cost compared to - Real time communication between SMS and GPRS user and application - Automated data upload from phone to server - Real-time visualization and - Requires smartphone and Electronic Forms GPRS monitoring of results additional installation of software (Java or other - Branching, skip logic and enforced application) validation at time of data entry - Asynchronous: As data can be saved to the phone, can be used in areas without mobile coverage (data is sent to server once in range of a signal) 4 Mobile phones have been used to collect data using three approaches: (1) Face-to-face interview using phone, laptop, or tablet computer with electronic form. The mobile phone essentially takes the place of paper in a traditional household survey, with enumerators conducting a “live� interview but capturing data provided by the respondent using an electronic device. 2 (2) Direct response to phone using SMS, USSD, IVR, or electronic form. Respondents interact with their own devices to answer questions digitally using the mobile’s keypad. The interface and data transfer technology employed could be SMS, USSD, or electronic forms. An alternative approach is to link the respondent to an automated interactive voice response system (IVR) in which the respondent dials a number and is prompted by a computer to record answers using the keypad. (3) Interview with enumerator via phone. Respondents receive a call from an operator at a call-center, are asked questions and dictate their answers which are then recorded by the operator on a computer. Few systematic studies are available that compare the data quality and data collection efficiency of these alternative approaches. In one India-based study evaluating the accuracy of data collection using electronic forms, voice, and SMS through mobile phones, MIT researchers found that a voice interface (in a setting where respondents dictated answers through the mobile phone to an operator in real time) provided the highest level of accuracy, leading them to favor the voice-operator combination for the implementation of a tuberculosis program (Patnaik et al. 2008). In Tanzania, a research team from Twaweza, a citizen-centered initiative in East Africa, favored a voice-operator interface to collect weekly data on provision of public services, due to implementation challenges associated with USSD, electronic forms, and IVR (MobileDataGathering 2011). For reasons explained below, SSEPS also employed a voice-operator approach. 2 The face-to-face, device assisted approach has also been used with devices other than mobile phones, such as tablet and notebook computers and is more broadly classified as Computer-Assisted Personal Interviewing (CAPI). See IRIS Center (2011) for a detailed review of CAPI software options and Caeyers et al. (2012) for an analysis of the performance of CAPI vs. traditional Pen-and-Paper Interviewing (PAPI). 5 3. The South Sudan Experimental Phone Survey In 2010 the World Bank financed a mobile phone-based survey to monitor conditions in South Sudan on a monthly basis and to contribute to knowledge regarding the feasibility and implementation of mobile phone surveys in low-income settings. The rationale for selecting a mobile data collection approach was rooted in context. In early 2010, South Sudan was a year away from the referendum for independence. Given a dynamic situation, setting up a platform to collect high frequency panel data on perceptions, living conditions, and access to services would help shed light on the evolving situation. Additionally, given the poor state of roads and the high costs of travel, mobile phones would offer the potential of a cost-effective solution to collecting high-frequency data by avoiding costly repeat visits to households. While ownership of modern consumer durables (such as televisions, refrigerators and computers) is uncommon in South Sudan, reflecting low incomes and the lack of reliable access to electricity in homes, ownership of mobile phones has been steadily increasing in both urban and rural areas (Table 2). In 2006 the percentage of households that reported owning mobile phones in South Sudan was 2 percent (5 percent in urban areas and less than 1 percent in rural areas). By 2010 reported household ownership was up to 18 percent, with almost two in three households in urban areas owning at least one phone. (In urban areas, those households that have at least one phone own on average two phones.) The growth in household ownership of mobile phones in the span of four years is quite significant, especially in urban areas, suggesting a growing familiarity with the technology and the likelihood that a good proportion of respondents in urban areas would have used a mobile phone previously. 6 Table 2: Household Ownership of Consumer Durables in South Sudan Percentage of Households Owning Item 2006 SHHS 2008/2009 NBHS 2010 SHHS Consumer Durable Overall Rural Urban Overall Rural Urban Overall Rural Urban Radio 21.9 16.8 32.6 24.6 19.6 54.4 30 27.3 60.2 Bicycle 26.2 19.9 39.2 25.0 23.5 34.0 26.3 24.5 46.9 Mobile Phone 2 0.6 5 15.1 7.8 58.5 18.3 14.7 58.5 Motorcycle/ Scooter 1.9 1.1 3.6 3.3 2.0 11.1 4.3 3.5 13.6 Animal Drawn Cart 1 0.8 1.6 2.9 1.8 3.1 2.4 2.4 3.2 Television 1.1 0.4 2.4 3.3 0.6 19.4 2.2 0.8 17.3 Car or Truck 0.6 0.3 1.3 1.7 1.0 6.3 0.9 0.5 5.5 Refrigerator 0.9 0.6 1.7 0.9 0.2 5.0 0.6 0.2 5 Computer 0.5 0.2 1.3 0.7 0.3 3.3 0.5 0.3 3.4 Boat with motor 0.9 0.6 1.5 1.4 1.5 0.8 0.4 0.4 0.7 Source: Authorss calculations using 2006 Sudan Household Health Survey, 2008 National Baseline Household Survey, and 2010 Sudan Household Health Survey. Notes: The table is sorted by highest to lowest overall ownership in 2010. Design, Implementation and Timeline Implementation of SSEPS involved two major phases: (1) Distributing mobile phones to individuals in households in South Sudan’s ten state capitals and training them on their use and (2) using a Nairobi- based call center (Horizon Contact Center) to call respondents on a monthly basis and conduct a 15-20 minute survey. To provide an incentive for participation, individuals who successfully completed a survey were given pre-paid airtime credit for the mobile phone. Planning for the survey began in January 2010, a pilot survey was conducted in June and July of 2010, and the main implementation of the project was conducted between November 2010 and March of 2011 (Figure 2). 3 Specific design choices were influenced by context. South Sudan’s population is widely dispersed over rural areas without good connections to urban centers, less than 25 percent of the population has any formal education, half of households are poor, and very little infrastructure exists outside of state capitals. Given poor network coverage in rural areas, the sample was limited to households in urban settings. To avoid systematically omitting households without a phone, a phone was given to individuals in sampled households, regardless of whether they previously owned a phone or not. It was also hoped that distributing phones would help vest in participants a sense of ownership in the project. 3 After March, the project was handed off to Cordaid (an international NGO) which continued making calls on a monthly basis to participants through early 2012. 7 Figure 2: SSEPS Timeline Year 2010 2011 J F MA MJ J A S O N D J F MA MJ J A S O N D Project Component Duration Project Planning/ Equipment Procurement 5 months Pilot Enumerator Training Phone Delivery & Baseline Survey 1 month Data collection (monthly) 4 months Households targeted for the SSEPS consisted of those that had been surveyed in round 4 of the Household Budget Survey (HBS), which were a representative sample drawn from the ten state capitals. These households were drawn from 10 enumeration areas in each of the state capitals, with 12 households per enumeration area for a total of approximately 1200 households. Of these households, 1007 were located and agreed to participate in the SSEPS. Given that the bulk of South Sudan’s urban population lives in the state capitals, results from the SSEPS are broadly representative of the conditions in urban areas. Selection of the individual respondent within each household to receive the mobile phone and answer calls was performed during an initial household visit made by enumerators from NBS. To assess the effect of incentives on participation, sampled enumeration areas were randomly assigned to two incentive groups: one receiving pre-paid credit of 5 Sudanese Pounds (SDG)—approximately US$2.17—for each call leading to a completed survey and another receiving pre-paid credit of 10 SDG (US$4.35) for each call leading to a completed survey. 4 This credit was transferred on a weekly basis: the call center would send a list of participants that had completed the survey in the previous week to partners at the mobile phone operator (Zain), which would then transfer the credit to households. In the month of December, some of the transferred credits were delayed by two weeks as the purchased credit had not been activated. In addition, enumeration areas were randomly assigned to receive either a basic Nokia phone or a Safaricom solar-rechargeable phone. Given these combinations, four groups defined by phone type and incentive amount were formed: (1) Nokia, 5 SDG incentive group, (2) Nokia, 10 SDG incentive group, (3) Safaricom, 5 SDG incentive group and (4) Safaricom, 10 SDG group. A last incentive which was offered to all respondents came in the form of a lottery. All sampled households 4 US Dollar equivalent figures are based on annual exchange rate average for of 2.3 SDG per USD for 2010. 8 were told that they would be eligible for the chance to win a 230 SDG credit transfer in a monthly survey should they complete the survey. Prior to visiting households, the mobile phones were fully charged, enabled with SIM cards, tested to ensure they could make and receive calls and pre-loaded with 2 SDG of credit. 5 A team of 20 enumerators from the National Bureau of Statistics were trained in the World Bank office in Juba on the procedure for the household visits. This involved an orientation about the purpose of the survey and the protocol for household visits. Visits to sampled households were carried out in 10 teams of 3 beginning the first week in November 2010. Each team covered one state capital. Most visits occurred in the morning or late afternoon. Initial plans called for randomly selecting one respondent from the household roster and alternating between male and female respondents to ensure gender balance. The pilot revealed this would be logistically challenging since randomly selected individuals not present at the time of the visit would require coordinating additional visits to the household until the respondent was found. In addition, as the survey focused on gathering information on general household and community-level characteristics rather than very specific or personal characteristics, the risk of systematically biased responses was not deemed large. As such, when enumerators visited each household, rather than taking the list of household members and randomly selecting one member from the list, the first adult member of the household encountered at the time of the visit was selected for participation. Each selected participant was trained in the use of the phone and given a calendar displaying the specific day he or she would receive a call, with the first call occurring four weeks after the initial household visit. Enumerators collected an alternative phone contact number if available, the language preference of the participant (Arabic, English, Dinka, or Nuer) and the participant’s preferred time of day for calls. Each month’s SSEPS questionnaire consisted of two modules: a set of core questions that were repeated each month and a set of “special� questions that were changed on a month-to-month basis. The core questions fall into three categories: past changes and outlook, material deprivation, and security. The wording for the core questions was taken from Afrobarometer surveys conducted in 2008 and 2009, in 5 Facilities for mobile charging through community charging stations, usually powered by generators or car batteries, are common in urban areas. 9 order to make it possible to compare results from Southern Sudan to other countries. 6 The core set of questions, as well as the special set of questions for each month are provided in Annex B. 4. Results Table 3: SSEPS Sample Characteristics Sample Characteristics 1 2 Table 3 displays key characteristics of the Characteristic N Mean Gender: Female (%) 1004 62.5 sample of SSEPS participants. Sampled Gender: Male (%) 1004 37.5 Age (years) 975 33.6 participants were disproportionately female, Age: Under-20 (%) 975 7.2 averaged 34 years of age at the time of the Age: 20-29 (%) 975 34.8 Age: 30-39 (%) 975 30.9 household visit (two-thirds of participants were Age: 40-49 (%) 975 15.6 between the age of 20 and 40), and primarily Age: 50 plus (%) 975 8.9 Language: Arabic (%) 1007 44.4 requested interviews in Arabic or Dinka. Language: Dinka (%) 1007 37.4 Language: Nuer (%) 1007 8.8 Thirty-six percent of respondents had access to Language: English (%) 1007 9.3 another mobile phone. Top-up value: 5 SDG (%) 1007 51.9 Top-up value: 10 SDG (%) 1007 48.1 Phone Type: Nokia (%) 1007 52.5 Response Rates and Attrition Phone Type: Solar (%) 1007 47.5 Alternative phone available (%) 1007 36.6 At the end of the first round of data collection Preferred call time: morning (%) 1007 57.9 Preferred call time: afternoon (%) 1007 17.3 in December 2010, 68 percent of SSEPS Preferred call time: evening (%) 1007 24.8 participants completed the mobile-phone Town of residence: AWEIL (%) 1007 11.3 Town of residence: BENTIU (%) 1007 7.5 administered questionnaire. Overall response Town of residence: BOR (%) 1007 11.6 Town of residence: JUBA (%) 1007 9.7 rates declined to 62 percent in January, 55 Town of residence: KWAJOK (%) 1007 8.1 percent in February, and 52 percent in March Town of residence: MALAKAL (%) 1007 9.8 Town of residence: RUMBEK (%) 1007 10.0 (Table 4 and Figure 3). Both the level and Town of residence: TORIT (%) 1007 10.6 trend of response rates varied substantially Town of residence: WAU (%) 1007 10.2 Town of residence: YAMBIO (%) 1007 10.9 across town of residence (Figure 4). In Wau, 1 N = Observations with non-missing values 2 N = Unweighted for example, although response rates declined with each successive round, relative to other towns, response rates were high: 84 percent in December 6 The reference period for the Afrobarometer questions is one year. Because the survey was conducted on a monthly basis, a one-month reference period was of greater interest for the time series data. For this reason, in December 2010 the poverty and security questions were asked with both a one-year and a one-month reference period. In subsequent months, they were asked with only a one-month reference period. The outlook questions are only asked with a one-year reference period. 10 and 64 percent in March. In Kwajok on the other hand, response rates were both declining and consistently low. After a drop in February, response rates in Malakal increased from their initial level of 73 percent in December to 77 percent in March. Table 4: Attrition and Non-response 1 2 Characteristic N Mean Overall response rate: Round 1 (December) (%) 1007 68.8 Overall response rate: Round 2 (January) (%) 1007 62.4 Overall response rate: Round 3 (February) (%) 1007 55.2 Overall response rate: Round 4 (March) (%) 1007 51.5 Total surveys completed (out of 4): None (%) 1007 16.9 Total surveys completed (out of 4): One (%) 1007 12.8 Total surveys completed (out of 4): Two (%) 1007 16.3 Total surveys completed (out of 4): Three (%) 1007 23.5 Total surveys completed (out of 4): Four (%) 1007 30.5 Attrition: Dropped out after round 1 (%) 1007 7.2 Attrition: Dropped out after round 2 (%) 1007 6.2 Attrition: Dropped out after round 3 (%) 1007 10.4 1 N = Observations with non-missing values 2 N = Unweighted Participants can be categorized by response pattern: (1) “full-compliers,� who completed all four rounds of the survey, (2) “intermittent compliers,� who completed between one and three surveys intermittently over the four rounds, (3) “drop-outs,� who dropped out after one or more successfully completed surveys and (4) “non-compliers,� who did not complete a single survey. Approximately 17 percent of the sample did not complete a single survey, while 31 percent completed all four surveys between December and March. Approximately half of all households (52 percent) completed between one and three surveys over the data collection period. Of these households, close to half (24 percent of the sample) dropped out after at least one successfully completed interview, 7 percent after round 1, 6 percent after round 2, and 10 percent after round 3. 11 Figure 3: Response Rates by Month, Type of Phone, Incentive and Participants by Response Pattern 100 Full Intermittent Non compliers compliers Drop-outs compliers 90 Full Sample 80 30.5 28.8 23.8 16.9 70 Nokia + 10 SDG 27.2 26.5 24 22.3 60 Nokia + 5 SDG 50 37 24.8 24 14.2 40 [Response Rate (%)] Solar + 10 SDG 30 31.3 28.8 19.5 20.4 Nokia + 5 SDG 20 Nokia + 10 SDG Solar + 5 SDG Solar + 5 SDG 10 27.4 34.7 26.7 11.2 Solar + 10 SDG 0 [Month] Full Sample [Percent] Dec Jan Feb Mar 0 20 40 60 80 100 Among respondents receiving Nokia phones, those offered 10 SDG credit were about 10 percent less likely in each round to complete the survey compared to respondents offered the 5 SDG credit (Figure 3). There was not a similar pattern among respondents using solar-rechargeable phones. Both Nokia and Safaricom solar phone users completed surveys with the same frequency: among 4,028 possible completed surveys across all four survey rounds the overall response rate for users of both phones was 59 percent. This suggests that the constraint of having to charge the Nokia phone using an electrical outlet did not influence survey completion. Three probit regression models were used to explore in more detail factors associated with successful completion of the survey. Model 1 in Table 5 assesses the likelihood that a respondent completed the questionnaire given personal characteristics and SSEPS design features. Models 2 and 3 in Table 5 assess the association of these same factors with the likelihood a participant completed all four surveys and with the likelihood a participant completed none. The decline in response over the course of the four month implementation period is evident in the regression results. Controlling for other variables, respondents were 7, 16, and 20 percent less likely to have completed a questionnaire in January, February, and March relative to December. Operational data tracked by the Horizon Contact Center suggests that this result was in part due to declining network availability over the four month period. 12 Survey completion was most likely among participants over the age of 40, and female participants were close to 10 percent more likely to complete the questionnaire than male participants. Participant language was not associated with response rate, suggesting that efforts Figure 4: Response Rates by Town to ensure interviewers at the call center could address participants in their preferred language were successful. As noted previously, the 72.7 75.8 77.8 value of the incentive offered was negatively associated with survey Malakal 56.6 completion: participants who were offered the 10 SDG top-up were 6.4 Yambio 74.5 69.1 63.6 71.8 percent less likely to complete the survey, and the type of phone received did not affect response rates. 7 Participants who provided an Juba 67.3 67.3 59.2 62.2 alternative mobile phone contact number were 24 percent more likely Bentiu 52.6 63.2 55.3 to complete a survey. Monthly reports from the call center contracted to 43.4 conduct the interviews suggest that this is partially explained by the Rumbek 58.4 46.5 36.6 44.6 fact that the respondents who previously owned a phone had given their 83.5 71.8 70.9 phones to relatives, thus giving agents at the call center more Wau 64.1 opportunities to locate the respondent. Additionally, when the Zain 80.7 72.8 Aweil 68.4 57.9 network used on the phones distributed by the project was not functioning, the call center could use an alternative service provider Kwajok 58.5 45.1 37.8 (mainly MTN and Vivacel) on the alternative phone. Participants with 28.0 Torit 66.4 access to another phone were also more likely to be familiar with the 51.4 44.9 35.5 technology, and participants with their own phones probably choose Bor 65.8 57.3 53.8 service providers with the best local coverage. 26.5 7 Respondents who completed surveys in the first two weeks of December experienced a delay in the transfer of air-time credit. Analysis not presented here shows that controlling for regression covariates, these households were 9.4 percent less likely to complete surveys in rounds 2, 3 or 4 compared to households that did not experience the delay. 13 Table 5: Marginal Effects from Probit Regression Assessing Relationship between Survey Participation and Respondent Characteristics Model 1 Model 2 Model 3 Dependent Dependent Dependent variable: variable: variable: Explanatory Variables Participant Participant participant completed completed all 4 completed 0 survey in round surveys surveys Survey Round: December [reference] Survey Round: January -0.0741*** Survey Round: February -0.156*** Survey Round: March -0.198*** Age: Below 20 [reference] Age: 20-29 0.0872** 0.0578 -0.0275 Age: 30-39 0.118*** 0.127** -0.0474 Age: 40-49 0.143*** 0.108 -0.0807*** Age: 50 plus 0.141*** 0.139* -0.0492 Gender: Male [reference] Gender: Female 0.0973*** 0.0336 -0.0745*** Language: Arabic Language: Dinka -0.0168 -0.00764 -0.00198 Language: Nuer 0.0834 0.0525 0.0170 Language: English -0.0115 -0.0258 -0.0209 Top-up value: 5 SDG [reference] Top-up value: 10 SDG -0.0637*** -0.0274 0.0700*** Phone Type: Nokia [reference] Phone Type: Solar -0.0252 -0.0433 0.00230 Alternative phone not available [reference] Alternative phone available 0.237*** 0.188*** -0.145*** Tow of residence: JUBA [reference] Town of residence: AWEIL 0.0805 -0.0511 -0.0923*** Town of residence: BENTIU -0.175** -0.248*** -0.0829* Town of residence: BOR -0.0799 -0.212*** -0.00464 Town of residence: KWAJOK -0.146* -0.191*** 0.0346 Town of residence: MALAKAL 0.0386 -0.0453 -0.1000*** Town of residence: RUMBEK -0.0594 -0.0434 0.00538 Town of residence: TORIT -0.116** -0.236*** -0.0332 Town of residence: WAU 0.101* 0.0577 -0.0732** Town of residence: YAMBIO 0.0987* 0.0471 -0.0849*** Observations 3,900 975 975 Number of households 975 975 975 *** p<0.01, ** p<0.05, * p<0.1 Participants who completed all four survey rounds were more likely to be above the age of 30 and more likely to have access to an alternative phone. Gender, the type of phone, and the incentive value were not associated with the probability a respondent would complete all four rounds of the survey. The very strong geographical pattern associated with completing all four survey rounds suggests that the differential reliability of network coverage across the ten urban centers was a strong determinant of 14 response rates. Participants who did not complete a single survey were less likely to be female, less likely to have access to an alternative phone, and more likely to have been offered a pre-paid credit of 10 SDG. As additional socioeconomic characteristics of households were available from the NBHS for about half of the SSEPS households, an additional set of regressions was run to explore the association between survey participation and the level of educational attainment of the household head, poverty status, and television ownership. Results from this set of regressions (Annex A) show that respondents belonging to households in which the head member completed post-secondary education were more likely to complete at least one survey. The poverty status of the household and ownership of a television were not associated with the rate of survey completion. Call Center Outcomes Horizon Contact Center kept records of the attempts to reach respondents (Figure 5). In December, among 3617 dials, 36 percent resulted in a contact with a respondent and 19 percent resulted in a completed survey. The efficiency of the calling process improved significantly in January—for every 100 dialed numbers, 34 resulted in a completed survey—but subsequently fell off in February and March where out of every 100 calls, only 17 and 15, respectively, resulted in a completed survey. Major challenges related to reaching the correct respondent involved an unreachable number (37 percent of calls) and no network coverage (18 percent). 8 Problems with the network increased over the course of the four month period, peaking at 32 percent of all dialed numbers in March. Calls to Bentiu and Torit in December and January and Aweil and Bor in March were cited as being particularly and consistently problematic due to poor network connectivity. The fraction of unreachable numbers was particularly high in February, when 47 percent of all dialed numbers were unreachable. 8 Unreachable numbers exist when a number is out of service or a connection cannot be made to the number being called. No network refers to the situation where all numbers fail to go through because of a hitch on the network either the originating network or the destination network. 15 Figure 5: Calling Efficiency Indicators Outcomes December January February March 3617 total calls 1861 total calls 3320 total calls 3394 total calls 3.6 peravg calls respondent 1.8 peravg calls respondent 3.3 peravg calls respondent 3.4 peravg calls respondent Contacts per 100 calls 35.7 48.5 24.7 22.3 Completed surveys per 100 calls 19 34.1 17.1 15.3 Distribution of calls by outcome: Unreachable number 39.2 26.2 47.6 33.4 Correct Respondent 21.1 34.1 17.6 15.9 Call Back 10.6 11.6 5.3 4.6 No Network 10 11.4 18.7 31.8 No Answer 8 8.5 7.1 8 Disconnected Number 6 2.7 1.6 3.7 Incorrect Respondent 1.7 1.1 .5 .3 Contact Hangup 1.2 .6 .5 .7 Number is busy 1.1 2.7 .2 .8 Language Barrier .7 .4 .4 .2 Wrong number .4 .7 .5 .5 Deceased 0 .1 0 .1 Other challenges during survey implementation related to content. For example, in February the rotating set of questions focused on governance and leadership, and call agents asked respondents to share perspectives on leadership quality. Many were reluctant to do so, citing concern that the information would be used against them. In January, when respondents were asked to list all member of the household in order to collect information on educational attainment, many were cautious and wanted reassurance the answers would be kept confidential. Call agents at Horizon also reported many respondents complained about faulty batteries (among solar phone users) and the costs of recharging (among Nokia phone users). There was also some evidence that respondents lost track of their call date and that some research phones were sold. In some cases when interviewers called the lines and made a contact, the person on the other end was not the listed respondent and mentioned having bought the phone on the market. Unfortunately, Horizon did not keep exact track of the extent of these issues so it is difficult to gauge how influential they were in determining overall response rates. Survey Results 16 Figure 6 displays responses to the core Figure 6: Summary of Responses questions by month. (The corresponding to Core Questions, by Month table is provided in Annex A). Overall, the response trends can be divided into two 100 In last month, gone without Food: 100 In last month, gone without Clean Water: groups characterized by their level of 80 80 volatility over the four survey rounds. 60 Once, Twice Four questions yielded consistent response or Several times 60 patterns: access to food, ability to access 40 40 medicines or medical treatment, frequency Never 20 20 Many Times of illness, and personal security. Several or Always 0 0 salient trends emerge. First, the percentage Dec Jan Feb Mar Dec Jan Feb Mar 100 In last month, gone without Fuel: 100 Ability to access medicine in last month: of households reporting food shortages 80 80 “many times or always� declined from 15 60 60 Yes percent to less than 2 percent between December and March. This downward 40 40 No trend most likely reflects the onset of the 20 20 harvest season which spans mid- 0 0 September to mid-January. Second, the Dec Jan Feb Mar Dec Jan Feb Mar 100 In last month, 100 In last month, gone without Medicines: something was stolen: percentage of households reporting 80 80 inability to access medicines doubled from 60 60 18 percent to 36 percent. Third, the frequency of reported illness in the 40 40 household did not change significantly 20 20 and, fourth, reported personal security 0 0 Dec Jan Feb Mar Dec Jan Feb Mar improved somewhat as the percentage of households not having had something stolen from their homes in the last month increased from 67 to 82 percent. Given that response rates gradually fell from December through March, some of the trends noted above could have plausibly resulted from a dynamic selection effect rather than changes in the economic, social, or political environment. However, the response patterns (both levels and trends) are very similar when comparing all responses, as presented above, with only responses from full compliers, i.e. the 17 subset of respondents that completed the survey in all four rounds. A plot of responses from full compliers is shown in Table A4 in Annex A. This suggests that election due to non-response does not appear to have markedly affected survey results. The two questions with most volatility on a month-to-month basis were access to clean water and access to cooking fuel. Among these two questions, two trends are observed: (1) In February, there appears to have been a one-time shock that made it easier for households to get access to fuel, and (2) beginning in February, a shock lasting through March made it more difficult for households to access clean water. Given that access to fuel depends on market conditions that can fluctuate rapidly (and therefore change the availability and price of fuel commodities), and access to clean water depends on environmental conditions (such as the onset of the dry season) and public or private services that can be disrupted, the higher volatility of responses to these two questions on a month-to-month basis relative to others appears plausible. 5. Discussion The fact that response rates dropped by about 5 percentage points per round, starting at 68 percent in December and ending at 52 percent in March, is cause for concern that selective response may bias the survey results. This worry is compounded by the finding that those responding were disproportionately women, older participants, and those with alternative phone numbers. At the same time, the fact that trends for responses from full-compliers alone are similar to trends based on all responses suggests that selective attrition did not ultimately bias the survey results. 9 The largest contribution to non-response resulted from individuals who participated intermittently over the course of the survey period. Future studies using mobile data collection might mitigate non-response by sending simple reminders (possibly using SMS) to reduce the number of intermittent compliers. Targeting older, female household members could reduce the number of non-compliers, but such targeting would not be desirable for surveys seeking a representative sample of individuals, e.g. labor force surveys. 9 It is not possible to determine, however, whether responses from the 17 percent who never responded would have differed from those of those who did response. 18 Notably, poverty status was not associated with survey completion. An important contributor to non- response over the course of the survey period seems to have been differential network coverage between geographic areas and disruption to network coverage within geographic areas over time. Having multiple options to call respondents by recording an alternative contact number helped significantly. This suggests that survey designs that use existing household mobile phones might yield higher response rates than survey designs that hand out mobile phones. The downside of this strategy is representativeness in areas with low household ownership of mobile phones. Larger incentives in the form of pre-paid calling credit did not work to encourage survey completion. In fact, participation rates were slightly lower for those who received the greater incentive. The underlying mechanism of this result is not clear. Given that the design of the study did not incorporate a group without an incentive payment, it is not possible to say whether offering some incentive increased participation versus offering no incentive at all. In terms of implementation, a critical feature of this study was partnering with the National Bureau of Statistics. The NBS provided the staff, knowledge and experience that enabled the sample selection and successful distribution of phones. Being able to tap the NBS’s experience in collecting data in South Sudan—given obvious logistical and operational challenges in the country—was a tremendous asset to the SSEPS project. Overall, the SSEPS experience is encouraging as to the potential for mobile phone-based surveys while at the same time offering a cautionary tale about the potential hurdles to the approach. In particular, variability in the functioning of the mobile network, which was not anticipated by the project team beforehand, turned out to be a substantial problem. At the same time, selection due to non-response turned out to be less of a concern than expected, as it does not appear to have markedly affected survey results in this case. The overall success of the project in an extremely difficult environment suggests that further efforts at mobile data collection in developing countries are worth exploring. 19 References Croke, Kevin, Andrew Dabalen, Gabriel Demombynes, Marcelo Giugale, and Johannes Hoogeveen. (2012). "Collecting High Frequency Panel Data in Africa Using Mobile Phone Interviews." World Bank Policy Working Paper 6097. Dillon, Brian (2011). "Using Mobile Phones to Conduct Research in Developing Countries." Journal of International Development. Volume 24, Issue 4. Economist (2009). "Eureka moments: How a luxury item became a tool of global development." The Economist. September 24. MobileDataGathering (2011). "So here we are." Blog post at http://mobiledatagathering.wordpress.com/2011/07/22/415/ IRIS Center (2011). "Comparative Assessment of Software Programs for the Development of Computer-Assisted Personal Interview (CAPI) Applications. University of Maryland at College Park." http://siteresources.worldbank.org/INTSURAGRI/Resources/7420178- 1294259038276/CAPI.Software.Assessment.Main.Report.pdf Caeyers, Bet, Neil Chalmers, and Joachim De Weerdt (2012). "Improving Consumption Measurement and other Survey Data through CAPI: Evidence from a Randomized Experiment." Journal of Development Economics. Volume 98, Issue 1. Loudon, Melissa (2009). "Mobile Phones for Data Collection." http://www.mobileactive.org/howtos/mobile-phones-data-collection. Patnaik ,Somani, Emma Brunskill, and William Thies (2009). "Evaluating the Accuracy of Data Collection on Mobile Phones: A Study of Forms, SMS, and Voice." Paper presented at IEEE/ACM International Conference on Information and Communication Technologies and Development. Schuster, Christian and Carlos Perez Brito (2011). "Cutting Costs, Boosting Quality and Collecting Data Real-Time - Lessons from a Cell Phone-Based Beneficiary Survey to Strengthen Guatemala's Conditional Cash Transfer Program." En Breve. Number 166. World Bank. Tomlinson, Mark, Wesley Solomon, Yages Singh, Tanya Doherty, Mickey Chopra, Petrida Ijumba, Alexander C Tsai, and Debra Jackson (2009). "The use of mobile phones as a data collection tool: A report from a household survey in South Africa." BMC Medical Informatics and Decision Making. Volume 9, Number 51. 20 Annex A Table A1: Mobile Technology Glossary Acronym Name Definition GSM Global System for Mobile A standard for digital cellular networks SMS Short Message Service The text messaging service component of mobile phones that allows exchange of short messages between devices. Messages are delivered using “store and forward� where messages are first sent to a SMS Center before delivering the text to the recipient. USSD Unstructured Supplementary Protocol used by GSM mobile phones to Service Data communicate with service provider’s computers. This is used to provide mobile users with menu-driven, interactive services (e.g. account balances, top-ups). As opposed to SMS, with USSD, a real-time session is initiated between the mobile user and the USSD application platform. IVR Interactive Voice Response A technology that allows a computer to interact with humans through the use of voice and dial tones through the keypad. GPRS General Packet Radio A packet-based wireless communication service used Service to exchange data between a mobile device and the internet. WAP Wireless Application A technical standard for accessing information over a Protocol mobile wireless network that enables mobile phones to access the internet. SIM Subscriber Identity Module A subscriber identity module or subscriber identification module (SIM) is an integrated circuit that securely stores the International Mobile Subscriber Identity (IMSI) and the related key used to identify and authenticate subscribers on mobile telephony devices (such as mobile phones and computers). 21 Table A2: Marginal effects probit regression assessing the relationship between survey participation and respondent characteristics (using subset of sample that had additional information from the NBHS) Model 1 Model 2 Model 3 Participant Participant Participant completed Independent Variables completed completed 0 all 4 survey surveys surveys Survey Round: January -0.0421 Survey Round: February -0.138*** Survey Round: March -0.115*** Household head education: Primary (%) 0.0468 0.0544 -0.00889 Household head education: Secondary or Post-Secondary (%) 0.0844* 0.0841 -0.0648* Household is poor (%) 0.0186 0.0146 -0.00759 Age: 20-29 (%) 0.0904 0.0880 0.000860 Age: 30-39 (%) 0.135** 0.202** -0.0100 Age: 40-49 (%) 0.200*** 0.253** -0.0841*** Age: 50 plus (%) 0.176*** 0.158 -0.0644* Gender: Female (%) 0.104*** 0.0860* -0.0486 Language: Dinka (%) 0.0268 -0.00393 -0.0310 Language: Nuer (%) 0.132** 0.159 -0.0589* Language: English (%) -0.000995 0.0357 -0.0319 Top-up value: 10 SDG (%) -0.0504 -0.00938 0.0643** Phone Type: Solar (%) 0.00787 0.00340 -0.0227 Alternative phone available (%) 0.185*** 0.120** -0.0910*** Town of residence: AWEIL (%) 0.0930 -0.0892 -0.0915*** Town of residence: BENTIU (%) -0.205** -0.329*** -0.0652* Town of residence: BOR (%) -0.112 -0.191** 0.0445 Town of residence: KWAJOK (%) -0.0978 -0.186* -0.0416 Town of residence: MALAKAL (%) 0.0261 -0.0981 -0.0871*** Town of residence: RUMBEK (%) -0.0418 -0.0208 -0.00361 Town of residence: TORIT (%) -0.117 -0.259*** -0.0542 Town of residence: WAU (%) 0.117* 0.0260 -0.0887*** Town of residence: YAMBIO (%) 0.184*** 0.152 -0.103*** Number of households 456 456 456 Observations 1,824 456 456 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 22 Table A3: Responses to Core Questions, by Round Round 1 Round 2 Round 3 Round 4 1 2 Characteristic Mean CI Mean CI Mean CI Mean CI In last month, gone without enough Food: Never 25.3 (17.3-33.3) 14.8 (10.3-19.2) 25.5 (15.9-35.1) 15.3 (8.8-21.7) In last month, gone without enough Food: Once, twice or several times 58.3 (45.5-71.1) 77.1 (71.6-82.7) 68.8 (60.5-77.1) 82.6 (76.4-88.9) In last month, gone without enough Food: Many times or always 14.5 (8.4-20.5) 8.0 (4.8-11.2) 3.9 (0.7-7.2) 1.7 (0.5-3.0) In last month, gone without enough Food: Don't know 1.9 (-0.7-4.5) 0.1 (-0.1-0.2) 1.8 (-1.4-4.9) 0.3 (-0.3-1.0) In last month, gone without enough Clean Water: Never 35.0 (22.6-47.5) 29.8 (24.8-34.8) 32.1 (24.9-39.4) 10.6 (5.8-15.4) In last month, gone without enough Clean Water: Once, twice or several times 34.1 (20.1-48.1) 43.6 (35.1-52.2) 19.8 (11.8-27.7) 51.4 (42.7-60.2) In last month, gone without enough Clean Water: Many times or always 30.7 (22.4-38.9) 25.9 (17.2-34.7) 48.1 (37.2-59.0) 38.0 (29.8-46.2) In last month, gone without enough Clean Water: Don't know 0.2 (-0.1-0.5) 0.6 (-0.5-1.7) 0.0 (0.0-0.0) 0.0 (0.0-0.0) In last month, gone without enough Fuel: Never 33.4 (20.9-45.8) 24.6 (18.0-31.2) 51.4 (38.6-64.3) 22.6 (14.6-30.6) In last month, gone without enough Fuel: Once, twice or several times 49.3 (36.1-62.6) 54.0 (47.4-60.7) 45.1 (33.1-57.1) 69.5 (63.0-76.1) In last month, gone without enough Fuel: Many times or always 17.3 (11.1-23.5) 21.0 (9.6-32.4) 2.1 (0.2-4.0) 7.1 (3.5-10.7) In last month, gone without enough Fuel: Don't know 0.0 (0.0-0.1) 0.4 (-0.1-1.0) 1.4 (-0.1-2.9) 0.7 (-0.3-1.8) In last month, able to access medicines when needed: Yes 81.4 (75.0-87.8) 68.8 (59.9-77.7) 60.7 (47.6-73.8) 61.3 (51.7-70.9) In last month, able to access medicines when needed: No 18.4 (12.1-24.6) 29.9 (21.4-38.3) 37.2 (24.1-50.3) 36.0 (26.2-45.8) In last month, able to access medicines when needed: Don't know 0.2 (-0.1-0.5) 1.3 (0.1-2.5) 2.2 (-0.2-4.5) 2.7 (0.9-4.5) In last month, gone without enough Medicines: Never 14.9 (11.1-18.8) 15.3 (10.3-20.2) 28.7 (15.5-42.0) 26.2 (16.5-36.0) In last month, gone without enough Medicines: Once, twice or several times 80.5 (76.3-84.6) 78.5 (71.8-85.2) 70.3 (56.5-84.2) 72.9 (62.6-83.2) In last month, gone without enough Medicines: Many times or always 4.4 (2.1-6.7) 6.0 (3.4-8.6) 0.7 (-0.1-1.4) 0.7 (0.1-1.4) In last month, gone without enough Medicines: Don't know 0.2 (-0.2-0.5) 0.2 (-0.1-0.4) 0.3 (-0.2-0.8) 0.1 (-0.1-0.3) In last month, something stolen from household: Never 67.2 (63.6-70.8) 74.3 (68.4-80.2) 76.4 (69.8-83.0) 82.0 (77.0-87.1) In last month, something stolen from household: Once, twice or several times 31.9 (28.5-35.4) 21.9 (17.1-26.7) 23.2 (16.8-29.7) 17.2 (12.2-22.2) In last month, something stolen from household: Many times or always 0.8 (-0.1-1.7) 3.8 (1.6-5.9) 0.4 (-0.2-0.9) 0.7 (-0.3-1.7) In last month, something stolen from household: Don't know 0.1 (-0.1-0.2) 0.0 (0.0-0.1) 0.0 (0.0-0.0) 0.1 (-0.1-0.2) 1 Mean = Weighted 2 CI = 95% Confidence Interval Dataset: sample_core.dta Date generated: 4/26/2012 23 Table A4: Responses to Core Questions, Full Compliers 100 In last month, gone without Food: 100 In last month, gone without Clean Water: 80 80 Once, Twice 60 or Several times 60 40 40 Never 20 20 Many Times or Always 0 0 Dec Jan Feb Mar Dec Jan Feb Mar 100 In last month, gone without Fuel: 100 Ability to access medicine in last month: 80 80 Yes 60 60 40 40 No 20 20 0 0 Dec Jan Feb Mar Dec Jan Feb Mar 100 In last month, 100 In last month, gone without Medicines: something was stolen: 80 80 60 60 40 40 20 20 0 0 Dec Jan Feb Mar Dec Jan Feb Mar 24 Annex B: Questionnaires Core Survey Questions 1. Since the last time we called you, how often, if ever, have you or anyone in your family gone without enough food to eat? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 2. Since the last time we called you, how often, if ever, have you or anyone in your family gone without enough clean water (government, private supplier) for home use? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 3. Since the last time we called you, how often, if ever, have you or anyone in your family gone without enough fuel (including charcoal, firewood, gas) to cook your food? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 4. Since the last time we called you, how often were you or anyone in your family sick enough to need medicines or medical treatment? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 5. Since the last time we called you, were you able to access medicines or medical treatment when it was needed? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 6. Since the last time we called you, how often, if ever have you or anyone in your family been physically attacked? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] 7. Since the last time we called you, how often, if ever, have you or anyone in your family had something stolen from your house? [Once, Twice, Several Times, Many Times, Always, Never, Don’t Know] Special Survey Questions December Looking back, how do you rate the personal security of you and your family compared to one month ago? Would you say that it is much worse, worse, the same, better, or much better compared to one month ago? Don't know Much Worse Worse Same Better Much Better Looking back, how do you rate the personal security of you and your family compared to twelve months ago? Would you say that it is much worse, worse, the same, better, or much better compared to twelve months ago? Don't know Much Worse Worse Same Better Much Better 25 Looking back, how do you rate your (household) living conditions compared to twelve months ago? Would you say that they are much worse, worse, the same, better, or much better compared to twelve months ago? Don't know Much Worse Worse Same Better Much Better Looking back, how do you rate economic conditions in Southern Sudan (country level) compared to twelve months ago? Would you say that they are much worse, worse, the same, better, or much better compared to twelve months ago? Don't know Much Worse Worse Same Better Much Better Looking ahead, do you expect your (household) living conditions to be better or worse in twelve months time? Would you say that they will be much worse, worse, the same, better, or much better in twelve months time? Don't know Much Worse Worse Same Better Much Better Looking ahead, do you expect economic conditions in Southern Sudan (country level) to be better or worse in twelve months time? Would you say that they will be much worse, worse, the same, better, or much better in twelve months time? Don't know Much Worse Worse Same Better Much Better Are you aware of the referendum on Southern Sudan to be held in January 2011? Yes No Don't know how to answer Don't want to answer Do you think that Southern Sudan should form an independent country? Yes No Don't know how to answer Don't want to answer Do you think the referendum will lead to long-term peace between Northern and Southern Sudan? Don't know Yes No Don't know how to answer Don't want to answer January 26 List each dependent: 1. Full Names: 2. Gender: [Select Gender]: Male/ Female 3. Age: What grade is currently attending? [SELECT OPTION] P1 P2 P3 P4 P5 P6 P7 P8 S1 S2 S3 S4 S5 S6 Post-Secondary Diploma Program/ University/ Vocational Training/ Khalwa Who owns the school, Vocational-Training Center that is attending? [SELECT OPTION] Government, Religious Institutions, NGO How does ____ usually travel to school, Vocational-Training Center? [SELECT OPTION] On Foot/ By Bicycle/ By Private Vehicle/ By Public Vehicle How long (in minutes) does it take to get to school, Vocational-Training Center? How much did the household spend in the last year on ___________’s education (including tuition, books, uniform, transportation and all other expense)? (In Sudanese pounds) Is a full time or a part time student? [SELECT OPTION] Full time/ Part Time What is _____’s other main activity? [SELECT OPTION]Taking care for ill family member/ Working at home/ Paid employed/ Unpaid employed/ Paid – in – kind employed/ Nothing/ Other Is attending school in this town or in another place?[SELECT OPTION] This Town/ Another Place In which other place is _______ attending school? (Name of place the dependent attends school) Why is not attending school? (If not attending school) [SELECT OPTION]Cultural Reasons/ Illness/ No money for School/ No school facility/ Paid employed/ Paid – in – kind employed/ School closed/ School too far from home/ Taking care of ill family member/ Unpaid employed/ Working at home/ Other February How much do you trust each of the following, or you haven't heard enough about them to say: The president? Don't know don’t want to answer Somewhat Just a little Not at all A lot How much do you trust each of the following or you haven't heard enough about them to say: Parliament? Don't know don’t want to answer Somewhat Just a little Not at all A lot How much do you trust each of the following or you haven't heard enough about them to say: Local government officials? Don't know don’t want to answer Somewhat Just a little Not at all A lot 27 How many of the following people do you think are involved in corruption, or haven't you heard enough about them to say: The President and officials in his office? Don't know don’t want to answer none some of them Most of them all of them How many of the following people do you think are involved in corruption, or you haven't heard enough about them to say: members of parliament? Don't know don’t want to answer none some of them Most of them All of them How many of the following people do you think are involved in corruption or you haven't heard enough about them to say: Local government officials? Don't know don’t want to answer none some of them Most of them All of them In your opinion, what are the most important problems facing this country that the government should address? Don't know Don't want to answer Nothing/No problems Management of the economy Wages income and salaries Unemployment Poverty/Destitution Rates and taxes Loans/credit Farming/Agriculture Food shortage/Famine Drought Land Transportation Communications Infrastructure/Roads Education Housing Electricity Water supply Orphans/Street children/Homeless Children Political Violence Political instability/Political divisions/Ethnic tensions Discrimination/inequality Gender issues/women's rights Democracy/Political Rights War(international)Civil war Agricultural Marketing Other( i.e. some other problem) What is your tribe? How much do you trust other South Sudanese? Don't know don’t want to answer Just a little A lot I trust them somewhat I trust them a lot March How often in the last year, if at all, did you receive money remittances from friends or relatives from NORTHERN SUDAN? (if never, don't know, don't want to answer skip to March 005) Once Twice Never don’t know Between 3 to 6 times Don’t want to answer More than 6 times 28 How do you usually receive remittances from NORTHERN SUDAN? Postal service Private Company (e.g. Western Union) Bank transfer Brought by a family member/ friend(s), Hawala How much does the household usually receive with each remittance (SDG)? How do you mainly spend it? Food and water Education of household member(s), Medical care of household’s member(s) Savings Other household’s expenditure How often in the last year, if at all, did you receive money remittances from friends or relatives OUTSIDE (NOT INCLUDING NORTHERN SUDAN) of the country (SOUTHERN SUDAN)? (if never, don't know, don't want to answer skip to March 009) Once Twice Never don’t know Between 3 to 6 times don’t want to answer More than 6 times How do you usually receive remittances from OUTSIDE (NOT INCLUDING NORTHERN SUDAN) of the country (Southern Sudan)? Postal service Private company (e.g. Western Union) Bank transfer Brought by a family member/ friend(s), Hawala Other How much does the household usually receive with each remittance (SDG)? How do you mainly spend it? Food and water Education of household member(s), Medical care of household’s member(s) Savings Other household’s expenditure How often in the last year, if at all, did you receive money remittances from friends or relatives INSIDE of the country (SOUTHERN SUDAN)? (if never, don't know, don't want to answer skip to question 13) Once Twice Never don’t know Between 3 to 6 times Don't want to answer More than 6 times 29 How do you usually receive remittances from INSIDE of the country (SOUTHERN SUDAN)? Postal service Private Company (e.g. Western Union) Bank transfer Brought by a family member/ friend(s), other How much does the household usually receive with each remittance (SDG) ? How do you mainly spend it? Food and water Education of household member(s), Medical care of household’s member(s) Savings Other household’s expenditure How often in the last year, if at all, did you send money remittances to friends or relatives INSIDE of the country (SOUTHERN SUDAN)? (If never, don't know, don't want to March 016) Once Twice Never don’t know Between 3 to 6 times don’t want to answer More than 6 times How much does the household usually send with each remittance (SDG)? Where do you usually send remittances? Urban areas (out of town but in other urban areas) Rural areas both urban and rural areas Did you or any member of the household save money in the last year? (if no, if don't know, don't want to answer the interview is completed) Don't know Yes No Don't want to answer Where do you mainly save money? NGOs (e.g. BRAC) informal organization (community level -Sanduk) Other At Home Bank Account Why do you have no savings in a bank account? (Ask only if responses to question March 017 is anything but bank account) Not enough money No bank in town Don’ t trust in banks The bank is too far I need (and I don’t have) an ID and a referee Bank account too expensive Other Not enough money the bank is closed 30 Why do you save money? Medical care of household’s member(s) For bad times For household’s members’ education For wedding expenditures For funeral expenditures For other Household’s members expenditures in the future Other 31