Informing Durable Solutions for Internal Displacement In Nigeria, Somalia, South Sudan, and Sudan Volume C: Technical Aspects Informing Durable Solutions for Internal Displacement In Nigeria, Somalia, South Sudan, and Sudan Volume C: Technical Aspects Acknowledgments This report was led by Utz Johann Pape (TTL; Senior Economist, GPV01) and written together with Ambika Sharma (Consultant, GPV01). The study ‘Eliciting Accurate Responses to Consumption Questions by using Honesty Primes’ was authored by Lennart Kaplan (Consultant, GPV01), and Utz Pape (TTL; Senior Economist, GPV01) and James Walsh (Con- sultant, GPVGE). The chapter has been published in the World Bank’s Policy Research Working Paper Series. The team would like to thank Pierella Paci (Practice Manager, GPV01), as well as the peer reviewers Joanna de Berry (Senior Social Development Specialist, GTFDR), Nandini Krishnan (Senior Economist, GPV06), Nadia Piffaretti (Senior Economist, GTFSA), and Quy-Toan Do (Senior Economist, DECPI) for guidance. The team are also thankful for the guidance and support received from Preeti Arora (Country Program Coordinator, AFCTZ), Elsa Araya (Sr Public Sec- tor Specialist, GGOAE), Tom Bundervoet (Senior Economist, GPV01), Giorgia Demarchi (Social Scientist, GPV02), Pablo Fajnzylber (Adviser, GGIVP), Xavier Furtado (Resident Representative, AFMBW), Qaiser Khan (Lead Economist, GSP01), Indira Konjhodzic (Country Program Coordinator, AFCNG), Nicole Klingen (Country Program Coordinator, AFCET), Rebecca Lacroix (Social Development Specialist, GTFOS), Lynne Sherburne-Benz (Director, GSJD1), Varalakshmi Vemuru (Lead Social Development Specialist, GSU07), and Tara Vishwanath (Lead Economist, GPV03). ii Table of Contents ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii EXECUTIVE SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v INNOVATIONS IN SURVEY METHODOLOGY AND ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 A.  Eliciting Accurate Responses to Consumption Questions Using “Honesty Primes” . . . . . . . . . . . . . . . . . . 1 B.  Drawing Typologies of Displaced Populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 APPENDICES Appendix A:  Eliciting Responses from Honesty Primes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Appendix B: Typologies: Country-Wise Methodology Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 LIST OF FIGURES Figure A.1  HFS, CRS, and IDPCSS coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Figure A.2  Value of core food consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure A.3  Number of core food items consumed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure A.4  Calorie consumption p.c. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure A.5  Calorie consumption p.c. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure A.6  Treatment components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Figure A.7  Consumption distribution by population and treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure A.8  Number of items consumed by population and treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure A.9  Treatment effects across quintiles (IDPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure A.10  Treatment effects across quintiles (non-IDPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure B.1  Clusters of households in 2D and 3D for Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Figure B.2  Clusters of households in 2D and 3D for Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Figure B.3  Clusters of households in 2D and 3D for Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Figure B.4  Clusters of households in 2D and 3D for South Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Figure B.5  Clusters of households in 2D and 3D for Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Figure A1  Respondents’ answers to moral priming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure A2  Treatment effects from different specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Figure A3  Consumption shares (SSP values) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Figure A4  Correlation of household size and purchasing prices per kilo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Figure A5  Accessibility rate of urban and rural areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 iii iv  |  Informing Durable Solutions for Internal Displacement Figure B1  Resultant dendrogram for Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Figure B2  Resultant dendrogram for Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure B3  Resultant dendrogram for South Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Figure B4  Resultant dendrogram for Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Figure B5  Resultant dendrogram for Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 LIST OF TABLES Table A.1  Results using poverty thresholds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Table A1  Treatment questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Table A2  Balance treatment and control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Table A3  Treatment distribution by survey strata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Table A4  Balance of accepting or rejecting a lie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Table A5  Results from quantile regressions of different outcome variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Table A6  Quantile regressions—reduced sample (only non-IDPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Table A7  Results from equation (1) without controls and with controls. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Table A8  Results from equation (1) with interacted controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table A9  Channel—UN assistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Table A10  Results from unconditional quantile regressions of different outcome variables . . . . . . . . . . . . . . . . 26 Table A11  Quantile regressions—outcomes in levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Table A12  Quantile regressions—without outliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Table A13  Quantile regression with per capita scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Table A14  Accessibility rate of urban and rural areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Table B1  Summary of the data and contribution to total inertia for Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Table B2  Size of each group of refuges in Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Table B3  Robust check excluding a share of the sample for Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Table B4  Summary of the data and contribution to total inertia for Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Table B5  Size of each group of IDPs in Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Table B6  Robust check excluding a share of the sample for Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Table B7  Summary of the data and contribution to total inertia for South Sudan . . . . . . . . . . . . . . . . . . . . . . . . 35 Table B8  Size of each group of IDPs in South Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Table B9  Robust check excluding a share of the sample for South Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Table B10  Summary of the data and contribution to total inertia for Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Table B11  Size of each group of IDPs in Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Table B12  Robust check excluding a share of the sample for Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Table B13  Summary of the data and contribution to total inertia for Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Table B14  Size of each group of IDPs in Sudan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Table B15  Robust check excluding a share of the sample for Somalia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Executive Summary Using micro-data to inform durable solutions requires innovative data collection methodologies and data analysis techniques. Sub-Saharan Africa has been ravaged by wars, civil conflicts, and disasters for decades, resulting in 18 million forced migrants, including nearly 11 million IDPs.1, 2 The number of forcibly displaced individuals across the world is currently at an all-time high of about 70 million.3 The large majority of them live as Internally Displaced Persons (IDPs) within their home country (41.3 million worldwide);4 others migrate across borders and are likely to become refugees abroad. For the majority of forcibly displaced people, the primary concern with finding safety has turned into a quest to find a durable solution, possibly away from home. Most displaced people remain in developing countries, making the distribution of IDPs skewed toward Sub-Saharan Africa. Forced displacement destroys livelihoods and opportunities for affected people, creates needs to be addressed immediately, and constrains potential solutions to end displacement. The framework of durable solutions defines criteria to determine to what extent a durable solution has been achieved. The core criteria for durable solutions are given as (i) long-term safety, security, and freedom of movement; (ii) an adequate standard of living, including at a minimum access to adequate food, water, housing, health care, and basic education; (iii) access to employment and livelihoods; and (iv) access to effective mechanisms that restore their housing, land, and property, or provide them with compensation. The framework further suggests three lenses to inform durable solutions: the cause-, the needs- and the solutions-based lenses that are linked to before, during, and after displacement. Providing evidence for durable solutions targeting IDPs is often neglected despite their large numbers and the advantage of their status as citizens. About three in four forcibly displaced people are IDPs in Sub-Saharan Africa. Once IDPs become refugees, the context for durable solutions changes rapidly. Refugees face severe limitations with respect to settling down and getting access to work permits, and cultural barriers such as language and clashing habits and traditions. Furthermore, host countries often do not express an interest in integrating refugees but treat them as a temporary nuisance. IDPs instead are still in their home country where they—in principle—have the same rights as other citizens. Thus, finding opportunities for livelihoods can be easier and can potentially lead more quickly to an end of forced displacement. While steps have been taken to make data collection more comprehensive and standardized, some method- ological questions remain unanswered. Several data limitations prevent an accurate assessment of socioeconomic conditions among displaced people and hosts, which hinders efforts to design targeted policy interventions. While many studies aim at understanding forced migration and suggest different solutions, only a few studies incorporate 1. Pew Research Center, “In Sub-Saharan Africa, Total Number of Forcibly Displaced People Increased Sharply in 2017.” 2. Internal Displacement Monitoring Center, “Global Report on Internal Displacement Summary.” 3. UNHCR, “Figures at a Glance. Statistical Yearbooks.” 4. UNHCR, “Figures at a Glance. Statistical Yearbooks.” v vi  |  Informing Durable Solutions for Internal Displacement a development approach that moves beyond humanitarian assistance, providing long-term solutions. Efforts of pol- icy and data community, including ReDSS, JIPS, and IASC, have led to a more coordinated approach to definitions of displacement-related challenges and, especially, solutions. Further, these efforts have produced frameworks that facilitate high quality, standardized, comprehensive, and detailed indicators to measure IDPs’ gap from a solution to displacement. However, approaches to optimize sampling methodologies in volatile, changing environments are yet to be fully developed. Additionally, issues of underreporting on consumption remain relatively unexplored. The current policy focus on more integrated approaches on studying displacement should be accompanied by more research in this direction. This report addresses a key survey methodology question in the displacement context. In response to under- reporting of consumption patterns (in particular, food consumption), the study proposes the adoption of behavioral nudges in survey design. To mitigate potentially spurious responses, including ‘honesty primes’ in the consumption module of the survey’s questionnaire has led to more accurate reporting. The report also proposes clustering approaches to derive typologies of IDPs, to inform the required specificity of programs, and to find durable solutions. Among the displaced, different groups can have different trajectories in displacement. Initial circumstances of displacement can translate into different needs and solutions depending on the displacement trajectory, which is pertinent for policy. Clustering analysis helps to identify the different typologies of the displaced. The aim of the analysis is to exploit the socioeconomic micro-level data to identify different groups or profiles of displaced households across the countries considered in Sub-Saharan Africa. These typologies are drawn using data on the causes of displacement, the current needs of displaced people, and the potential solution to end displacement. A behavioral experiment indicates that using “honesty primes” can increase consumption reporting for the poorest and most vulnerable. Evidence suggests that IDPs underreport consumption. In recent experiences surveying IDP households in Soma- lia and South Sudan, IDP households have tended to report significantly lower levels of consumption than non-IDP households. These low levels are driven by large proportions of IDP households reporting very low or no consumption at all. In IDP camps in South Sudan, almost one-third of respondents (30 percent) report a calorie intake below the daily subsistence level of 1,200 kcal per day. The median of per capita consumption, at 1,589 kcal per day, is well below the recommended 2,100 kcal per day. While there is little doubt that IDPs are considerably more vulnerable to consumption shocks than non-IDP households, such high levels of no consumption over the recall period are unlikely to be true, as they would be associated with high rates of mortality. These responses, therefore, are likely to be spurious and not a useful basis on which to inform policy. Volume C: Technical Aspects  | vii A substantial part of the population reports consumption well below the subsistence level of 1,200 kcal/day Calorie consumption per capita (p.c.) .0006 .0004 Density .0002 0 1,200 1,589 2,100 Calorie consumption IDPs—control IDPs—treat An experiment in South Sudan investigates whether “honesty primes” can increase honest reporting of con- sumption. Insights from behavioral science have often been used as a policy tool to increase honesty and discourage anti-social behavior. Building on this literature, a set of “honesty primes” were experimentally administered to one-half of the survey respondents in IDP camps and in urban areas across South Sudan. If the primes have a differential impact on the consumption reported by IDPs relative to non-IDPs or among the poorest, then they can help to ascertain whether there was indeed some misreporting of consumption and suggest ways to tackle the issue. Honesty primes increased the reported consumption among lower consumption quantiles IDP Non-IDP The results suggest that the “honesty primes” had a positive impact on the consumption reported by the poor- est and more vulnerable respondents. The primes induce higher reporting only for IDPs at comparatively low lev- els of reported consumption. This treatment pattern is driven by aid-reliant IDPs and vanishes when considering the comparison group of non-IDPs. The results are especially marked for consumption quantities (number of items and kilograms), which are most easily subject to intentional misreporting. Overall, this evidence suggests that some IDPs are indeed misreporting consumption and that primes can help to improve data accuracy. viii  |  Informing Durable Solutions for Internal Displacement The report develops typology analysis to dive deeper into the differences among IDPs. Innovative analysis techniques identify the different trajectories among IDPs. IDP populations are not homo- geneous. Differences in the initial circumstances surrounding each household’s displacement can derive in different needs and solutions, which have policy implications. The study uses Multiple Correspondence Analysis (MCA) to draw different profiles of IDPs based on their past conditions (cause-based indicators), present circumstances (needs-based indicators), and future intentions (solutions-based indicators). MCA aggregates the micro-data to identify which IDP households are more alike and which are more different. Organizing IDPs into groups that share similar characteristics allows for tailored programs that reflect their unique circumstances and helps avoid subjecting households to interven- tions that are not appropriate to their needs and aspirations. The scope of this analysis is to illustrate the relevance of the full displacement trajectory and explore how this can point to tailored solutions. Results and policy implications of this analysis are presented in Volume A, while the methodology is detailed in this volume of the report. More research and experimentation are needed to better measure and understand the socioeconomic well-being of IDPs. Chapters A and B of this volume show the distinct contexts of IDPs and how survey and analytical method- ologies can be adapted to measure and apply the durable solutions framework. The ‘honesty primes’ give an example of how data collection needs to be adapted in the context of displaced populations, due to differences in reporting relative to ‘traditional’ survey respondents. The methodology to derive typologies of IDPs serves as a proposal of how information from quantitative data sets can be harvested to provide important distinctions between different groups of IDPs. Further research and experimentation can help to further improve and adapt methodologies for measuring and understanding pathways toward durable solutions. While evidence is sufficient for underreporting of con- sumption among IDPs, a better understanding of the causes can help to optimize and complement ‘honesty primes’ to further improve consumption reporting. Also, additional innovations like using drones for listing in IDP camps can be considered, but they require a thorough vetting process, given sensitivities and privacy concerns. In addition, analysis to better inform durable solutions, as well as applying the durable solutions framework, like the presented typologies, will be helpful in informing programs and targeting. Finally, IDPs as well as refugees are usually only implicitly included in national household surveys, resulting in an insufficient sample size to conduct analysis specifically for displaced pop- ulations. Thus, explicit sample stratification to better understand displaced populations as well as host communities in national surveys is another important step forward. Innovations in Survey Methodology and Analysis A.  Eliciting Accurate Responses to Consumption Questions Using “Honesty Primes” Introduction 1.  Accurate data are essential to understand the economic situation of IDPs and to develop evidence-based policies to support them. Accurate data on the key economic variables affecting people who have been forcibly displaced, such as consumption and assets, are essential to understand their situation and to develop evidence-based policies to support them. Poor information or data inaccuracies can lead to flawed diagnostics or incorrect assessments of impact, which would ultimately result in policy makers seeking to maximize the impact of allocating limited funds to the wrong people or to the wrong programs. Consumption data are particularly important in the context of displace- ment where malnutrition is rife and mortality rates are high, and where interventions are targeted toward supporting the immediate basic needs of vulnerable populations. 2.  IDPs in South Sudan and Somalia tend to report extremely low levels of consumption, significantly lower than non-IDP households. In previous survey rounds of the Somali High Frequency Survey (HFS), approximately 45 per- cent of Somali IDP households reported food consumption below subsistence levels and 80 percent below recom- mended levels (Figure A5 in Appendix A).5 To some degree, lower consumption levels for IDP populations are expected. IDPs often face significant hardships that hinder their potential for generating adequate livelihoods. Such hardships normally involve experiencing the loss of the breadwinners, lack of productive assets, or falling victim to violence. Indeed, IDPs have much less control over their own livelihoods, employment opportunities are scarce within camps, and a large part of their consumption is provided for by NGOs and international organizations through the distribution of aid. 3.  Despite acknowledging the hardships faced by IDPs, there are a few reasons to suspect that the observed low levels of consumption might be due, at least in part, to misreporting. First, such low levels of consumption would be associated with high rates of mortality due to starvation. The observed mortality rates among IDPs suggest that this is not happening systematically across the country at such a scale.6 Second, non-IDP households that are statistically similar on observable characteristics report higher levels of consumption than IDP households. While IDPs and non-IDPs may have different opportunities to generate income, it is unlikely that IDPs choose not to smooth their resources to balance between food and non-food consumption in a way that endangers their lives.7 These responses, therefore, might be spurious and not a useful basis on which to inform policy. 4.  The World Bank and other policy organizations often develop statistics through individuals’ responses to questions in economic surveys; however, self-reported information is vulnerable to reporting inaccuracies. The 5. Wave 1 of High Frequency Surveys in Somalia. 6. Although data from the USAID led Famine Early Warning Systems Network (FEWS NET) suggest a high level of malnutrition, evidence on mortality across the counties is mixed (FEWS NET, 2018). 7. The underlying survey data of this study indicates that IDPs smooth their consumption by having a more calorie intensive budget profile. 1 2  |  Informing Durable Solutions for Internal Displacement potential for biased information has been documented, and there exists a significant body of research concerned with improving the accuracy of self-reported information collected in household surveys.8 In the context of IDPs, the notion that respondents have—or believe they have—an incentive to report dishonestly is particularly relevant. Indeed, sur- vey respondents in IDP camps may operate under the belief that their responses will influence the provision of human- itarian aid, and will thus misreport consumption in an attempt to influence aid distribution. If survey respondents are reporting dishonestly, the inaccuracies generated in the data are highly problematic. At best, it makes the data false and unusable. At worst, it could lead to misallocations of aid, from more vulnerable areas to less vulnerable areas, or from where immediate relief is critical to where immediate relief is unnecessary. 5.  This chapter presents the results of a methodological experiment conducted in South Sudan to investi- gate whether consumption might be underreported in IDP households and whether incorporating “honesty primes” might encourage more accurate reporting. Insights from behavioral science have often been used as a pol- icy tool, and priming respondents can be used to increase honesty.9 Building on this body of research, a set of “honesty primes” were randomly administered to one-half of the survey respondents in IDP camps and urban areas interviewed across South Sudan in the 2017 wave of the HFS. The primes helped ascertain whether there was indeed some misre- porting of consumption, which might be the case if there was a differential impact on the consumption reported by IDPs relative to non-IDPs, or among the poorest. Furthermore, observing an impact on reporting behavior also serves to suggest a method that can be used to address misreporting. 6. The results suggest that there was indeed some underreporting and that the “honesty primes” had an impact on the consumption reported by poorer and more vulnerable respondents. The primes increases reported consumption only for IDPs with a low consumption level. This treatment pattern is driven by aid reliant IDPs and vanishes when considering the comparison group of non-IDPs. The results are especially marked for consumption quantities (number of items and kilograms), which are most easily subject to intentional misreporting. Overall, this evi- dence is taken to suggest that some IDPs are indeed misreporting consumption and that primes can help to improve data accuracy. However, results should be taken with a grain of salt as it is not possible to compare the reported con- sumption outcomes to more objective consumption data. Before adjusting consumption estimates in a systematic way, data should be compared to more objective information from administrative, observational, or anthropometric measures. Although this type of data was not available in IDP camps due to the fragile context, future research could validate this finding in other settings. 8. There are a number of mechanisms through which the validity of self-reported information in surveys can be compromised. Some inaccuracies result from cognitive biases—for example, acquiescence or “yea-saying” (Bachman and O’Malley 1984; Hurd 1999), extreme responding (Cronbach 1946; Hamilton 1968), and question order bias (Sigelaman 1981). Other inaccuracies emerge from conscious but not calculated behavior. Respondents may deliberately misreport information on sensitive subjects, not to distort statistics but to maintain their reputation or to abide by political norms (Gilens, Sniderman, and Kuklinski 1998; Rosenfeld, Imai, and Shapiro 2016). However, some misreporting is purposeful. Individuals may misreport in a calculated fashion to increase earnings in a study context (Mazar, Amir, and Ariely 2008) or to shape the results of the study if they believe that it will inform policy. It is not surprising that this problem might arise in the context of development aid, an area rife with perverse incentives (Bräutigam and Knack 2004; Cilliers, Dube, and Siddiqi 2015). 9. The World Bank 2015; Rasinski et al. 2005. Volume C: Technical Aspects  | 3   FIGURE A.1    HFS, CRS, and IDPCSS coverage Note: The HFS interviewed a representative sample of households in urban centers in the states colored in blue in the map above. The CRS interviewed households in four of the largest IDP camps in South Sudan, denoted by red diamonds in the map. The black dots refer to major urban centers. The IDPCSS was conducted in the Juba PoC1. Motivation 7.  The experiment sample includes 4,145 IDP and 781 non-IDP households interviewed across South Sudan in 2017. The experiment was built into the questionnaires administered in South Sudan in three different data collection exercises rolled out in mid–late 2017. They were: the Crisis Recovery Survey (CRS), which interviewed a representative sample of IDPs in four of the largest IDP camps across South Sudan; (2) Wave 4 of the High Frequency South Sudan Sur- vey (HFSSS), which conducted a representative survey across urban centers in 7 of the 10 former states in South Sudan; and (3) the IDP Census and Sampling Study (IDPCSS), which conducted a census of all households in Juba POC1 as part of the sampling experiment (Figure A.1). The questionnaire for the three surveys was designed to be directly compa- rable, and the consumption modules were exactly the same. The combined sample included 2,204 IDP households interviewed in the CRS after cleaning, 1,941 IDP households interviewed in the IDPCSS, and 781 non-IDP households interviewed as part of the HFSSS. 8.  Poverty among IDP households is high, 9 in 10 IDP households across South Sudanese Protection of Civil- ians (PoCs) live under US$1.90 PPP (2011) per capita per day in 2017. The incidence of poverty among IDPs is greater than for non-IDP urban residents, more than 9 in 10 camp IDPs live below the international poverty line of US$1.90 per capita per day compared to about 7 in 10 non-IDP urban residents (p<0.01). The depth of poverty expe- rienced by IDPs is greater on average than that of urban residents, meaning that poor IDPs tend to live further below the poverty line. The poverty gap, defined as the mean consumption shortfall relative to the US$1.90 per capita per day poverty line, is 54 percent for camp IDPs compared to 35 percent for non-IDP urban residents (p<0.01). Indeed, IDP households consume on average 333 South Sudanese Pounds (SSP) (May 2017) per capita per week compared to 403 for the non-IDP households (p<0.001) (Figure A.2). IDPs also report fewer consumption items, 6.63 consumption items relative to 7.68 for non-IDP households (p<0.001, Figure A.3). These indicators represent about 20 and 23 percent of items asked about in the households, respectively.10 10. For a detailed description of the sampling methodology and the use of core and non-core items, please consult Pape and Mistiaen (2015). 4  |  Informing Durable Solutions for Internal Displacement   FIGURE A.2    Value of core food consumption   FIGURE A.3    Number of core food items consumed .0004 .15 .0003 Density .0002 .1 Density .0001 0 .05 0 500 1,000 Consumption value, May 2017 SSP 0 IDPs Non-IDPs 0 5 10 15 20 0 5 10 15 20 IDPs Non-IDPs Number of core consumption items Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017.11 Graphs by population   FIGURE A.4    Calorie consumption p.c.   FIGURE A.5    Calorie consumption p.c. .0004 .0006 .0003 .0004 Density Density .0002 .0002 .0001 0 0 1,2001,589 2,100 S RM Calorie consumption Calorie consumption IDPs—control IDPs—treat (adult equivalents) IDPs—control IDPs—treat Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. 9.  A substantial part of the IDP population reports food consumption below the subsistence level. Scaled caloric consumption levels are labeled in Figure A.4 as S subsistence equivalent (1,200 kcal p.c.), R recommended daily intake (2,100 kcal p.c.) and M the median (2,314 kcal p.c.). Almost one-third of respondents (30.1 percent) report a cal- orie intake below the daily subsistence level of 1,200 kcal per day, and the median of per capita consumption is below the recommended calorie intake (1,589 kcal per day). Conditioning on adult equivalents, the median shifts well above the recommended daily intake. However, still a substantial part of the distribution of 16 percent reports below the sub- sistence level and 40 percent below the recommended daily intake (Figure A.5).12 As with the number of consumption items, the graph indicates that there is a slight shift in reported consumption among the treated regarding very low consumption levels. The main analysis will build on the more normally distributed adult equivalents. 11. Estimates presented in Figures A.2 and A.3 are not weighted and are representative only of IDP and non-IDP households surveyed in the study sample. 12. Several respondents report overly high consumption levels, which surpass conventional levels by far (> 4,000 kcal per day). Robustness checks take this issue into account by censoring the data at the extremes. Volume C: Technical Aspects  | 5 10.  Survey data are used to construct measures of food poverty and correspond to low consumption levels.13 In line with the hypothesis, the median of per capita calorie intake is well below the recommended daily intake of 2,100 kcal. While monetary poverty headcounts are a key metric when identifying the poor, metrics based on caloric food poverty lines are of equal relevance in our context. These can be set at the subsistence level of 1,200 kcal and the rec- ommended daily intake of 2,100 kcal.14 Honesty Primes and Experimental Design 11.  A prime is an environmental cue that unconsciously induces a subsequent cognition or behavior. For exam- ple, in studies with prisoners and bankers, participants who engage in activities that prime their identity behave more dishonestly in behavioral experiments than participants who have not participated in priming activities.15 Honesty primes have been found to elicit more honest and accurate responses during questionnaires.16 12. To investigate whether consumption might be underreported by IDP populations, households inter- viewed are randomly exposed to a bundle of “honesty primes.” The treatment comprises of three components, which were simultaneously administered in one treatment arm (Figure A.6).17 These include an emphasis on the impor- tance of accurate answers at the beginning of the survey, a short fictional scenario which will require passing judgment on the behavior of one of the characters,18, 19 and additional questions to determine when the last time was that their household had a meal, forcing the respondents to explicitly report that they have not eaten in the last week.20 While the former two target intentional misreporting, the latter addresses classical measurement error. 13.  The honesty primes address three different behavioral processes. Appeals to honesty are a standard tool in surveys to increase data accuracy by relying on social approval.21 Moral primes induce unconscious cognitions, which are intended to affect subsequent behavior. When facing incentives to lie, respondents answer more truthfully to sus- tain self-consistency. This is based on research illustrating that people make decisions on the basis of both external and internal reward mechanisms, meaning that even in cases where people have a material incentive to lie, their internal drive to protect their sense of self-integrity may override that incentive. Investigative probing puts a higher salience on the question. By asking for broader categories first, subsequent subcategories are put under more scrutiny. Self- consistency is reinforced by relating to a longer recall period of seven days. 13. For a more detailed description of the method for estimating caloric consumption, please consult Appendix A. 14. Ravallion and Bidani 1994. 15. Questionnaires confirmed that participants from the finance industry associate their professional identity with dishonesty (Cohn, Fehr, and Marechal 2014; Cohn, Maréchal, and Noll 2010). 16. Rasinski et al. 2005; Vinski and Watter 2012. 17. The “honesty primes” were administered to a random subset of approximately 50 percent of the sample, with 2,459 in the treatment group and 2,467 households in the control group. For a distribution of treatment and control over the different states, please consult Appendix A. The randomization process was built into the CAPI questionnaires administered in the surveys to achieve a balanced assignment of treatment and control groups across states and camps. 18. Mazar and Ariely 2006. 19. One example of this is when individuals’ beliefs regarding the consequences of lying affect their behavior. In a two-person experiment where one participant can increase her payoff by lying but at the expense to her counterpart, Gneezy (2005) finds that individuals’ propensity to lie is sensitive to the cost it imposes on the other person. 20. In line with the methodology of the main survey, a recall period of seven days is used. Approval to the broader seven-day recall period is intended to induce a higher accuracy in more specific categories, as individuals would like to sustain self-consistency. Appendix A provides a description of the corresponding questions (Table A1). 21. Talwar, Arruda, and Yachison 2015. 6  |  Informing Durable Solutions for Internal Displacement   FIGURE A.6    Treatment components • Appeal to Honesty "Thank you for taking the Ɵme to speak to us. We really appreciate the Ɵme you are giving to parƟcipate in the survey. We encourage you to provide honest informaƟon. By 1 parƟcipaƟng in the survey and by providing accurate informaƟon, you are playing an important role in helping us understand the situaƟon in South Sudan." • Moral Prime to Encourage Honesty "John asks his good friend Deng if he has some money that he can lend him to help him pay for medicine for his sick son. Deng 2 has money but was planning to buy cigareƩes with it. He lies and tells John that he has none. Is it okay for Deng to lie to John?" • InvesƟgaƟve Probing •At the start of the survey module concerning food consumpƟon, the respondent will be asked to tell when was the last Ɵme their household had a meal. This quesƟon will then also be asked for 3 each of four major food categories: ‘Bread and Cereals’, ‘Meat’, ‘Fruits’, ‘Pulses and vegetables.’ E.g., "When was the last Ɵme that any of the household members had Bread and Cereals?" Note: A more detailed description of the prime can be found in the technical Appendix A. 14.  Four main variables of interest are used as dependent variables: (i) the number of consumption items con- sumed, (ii) consumption quantity per capita (in kilograms), (iii) monetary consumption value per capita, and (iv) daily caloric intake per capita. Different dependent variables are specified because they have different implications for the respondent’s scope of influence on their value. The impact of the “honesty primes” on the total consumption value, both in terms of money and food intake, is of primary interest. Yet, they are of second-order value that is calculated as a function of other variables, including consumption quantities and calories or prices that are in turn deflated. Thus, these variables are difficult for respondents to falsify because all of this adds noise to the answer provided by the respondent. The consumption quantity in kilograms is a more direct measure of the quantity consumed as expressed by the respon- dent, and may lead to more accurate estimation of the impact of the “honesty primes.” Finally, counting the number of items may lead to an even more accurate measurement, since the variable does not undergo any cleaning at all and is taken at face value. Furthermore, omitting an item is likely to be the easiest and quickest way for respondents to reduce the value of the household’s consumption.22 22. Note that the number of consumption items is not reported per capita as it does not increase proportionally with household size. Volume C: Technical Aspects  | 7 15.  Randomization can be used to simulate a counterfactual. As there is, unfortunately, no information on the “true” consumption data available, randomization across comparable IDPs can create a quasi counterfactual. Moreover, we control for potentially remaining differences between the primed and the non-primed households. Prices might differ during a harvest period or different IDP camps, and the household composition or socioeconomic status might affect consumption composition. Hence, we control in our regression for month and camp of data collection, house- hold size, household heads’ sex, and the share of children in the household, as well as an asset index. The primes might also interact with household characteristics if women respond differently to honesty primes or there is stronger recall bias in larger households. Moreover, larger households are on average more prone to consumption poverty and might react differentially.23 Thus, some specifications also include interactions of the honesty primes with control variables. 16.  One way to investigate whether people report dishonestly is to test whether consumption rates change in response to “honesty primes.” If these primes are effective, they would be expected to particularly affect poten- tial underreporting, hence, poor households. Moreover, as vulnerable populations would have higher incentives to underreport, priming should be stronger for IDPs than for comparable non-IDP populations. We find the primes induce higher reporting among reported consumption poor households. This treatment pattern is driven by aid-reliant IDPs and vanishes when considering the comparison group of non-IDPs. The results are especially strong for consumption quantities (items and kilograms), which are most easily subject to intentional misreporting. This suggests that IDPs are indeed misreporting. Results 17.  The differences between consumption across treatment and control groups give a slight indication that the treatment may have been effective. The consumption levels shown in Figure A.7 show a slight difference in con- sumption between IDP households in the treatment group and the control group, though this is apparent only at lower levels of consumption, i.e., below 400 SSP. In contrast, the distribution of consumption across the two groups matches more closely for the non-IDP population. The distribution of the number of items displays a similar pattern, though the effect is also faint (Figure A.8). Again, a difference is not visible in the non-IDP population. The number of observations for the non-IDP population is much lower than for IDP, and hence the variance of the distribution is expected to be much greater. 18.  Misreporting can be expected at low consumption levels. Hence, heterogenous treatment effects across dif- ferent consumption levels can put “honesty primes” to an empirical test. Figure A.9 depicts priming effects across dif- ferent consumption levels for the four outcome measures of interest.24 The priming significantly increases reported consumption among lower consumption levels, but not for medium and higher consumption levels. Significant treat- ment effects occur mainly for the number of consumption items and the quantities in kilograms. Monetary and caloric consumption measures are less strongly affected. The latter might also be less susceptible to deliberate misreporting as they depend in part on variables over which the respondent has no control (i.e., calories per item; deflators). 23. Lanjouw and Ravallion 1995. 24. Along with the point estimates Figure A.9 provides a band of the statistical 95 percent confidence interval of the estimate. Thus, if the confidence band does not cross zero, there would be a 5 percent chance of indicating significant effects, while the “true” effect would be zero. 8  |  Informing Durable Solutions for Internal Displacement   FIGURE A.7    Consumption distribution by population   FIGURE A.8    Number of items consumed by popula- and treatment tion and treatment .0004 .15 .0003 Density .0002 .1 .0001 0 .05 0 500 1,000 Consumption value, May 2017 SSP 0 IDPs—C IDPs—T 0 5 10 15 20 0 5 10 15 20 Density Non-IDPs—C Non-IDPs—T Control, IDPs Control, Non-IDPs .15 .1 .05 0 0 5 10 15 20 0 5 10 15 20 Treatment, IDPs Treatment, Non-IDPs Number of core consumption items   FIGURE A.9    Treatment effects across quintiles (IDPs)   FIGURE A.10    Treatment effects across quintiles (non-IDPs) IDP Non-IDP Note: Corresponding point estimates are provided in Appendix A. Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017.25 25. All regressions use cluster robust standard errors (White 1980). Confidence bands refer to the 95 percent confidence interval. Consumption quantities, values, and calories are used in per adult equivalent terms. The regression framework is introduced in Appendix A. No sampling weights used as “honesty primes” are expected to affect, specifically, the extremes of the distribution, and the average treatment effect is not a priori of interest. Volume C: Technical Aspects  | 9 19.  The priming has stronger effects among the more vulnerable IDPs. The non-IDP subsample is used to assess the robustness of the main results as we would expect a less significant priming effect among the non-IDPs. Results in Figure A.10 indicate less significant effects, corresponding to the hypothesis that the “honesty primes” are more effective among the more vulnerable IDPs.26 This corresponds to adverse/perverse incentives in the setting of foreign assistance. Specifically, the IDPs, as exposed more intensively to development aid, could be more likely to signal their “neediness” and/or to provide socially desirable answers to signal their “worthiness” for assistance.27 20.  The “honesty primes” would ideally increase monetary and caloric food consumption to more credible consumption levels for certain strata. Four dichotomous indicators are used to assess whether the priming shifts a significant share of respondents above certain reporting thresholds. Those are equal to one if (i) the respondent household surpasses the caloric subsistence level of 1,200 kcal or (ii) the recommended level of caloric intake of 2,100 kcal. Two further dummies are created at (iii) 66.66 percent and (iv) 100 percent of a normalized poverty line, which are scaled by the fact that only core consumption items were assessed consistently across all surveys. Although the coeffi- cients are mostly positive, only two coefficients turn significant in Columns (2) and (3) of Table A.1. The results stress the positive effect of the primes, where 7 percent more respondent households would report above the recommended daily calorie intake level. However, only certain population strata are affected. The bundle of primes addresses different behavioral processes.   TABLE A.1    Results using poverty thresholds (1) (2) (3) (4) >1,200 kcal >2,100 kcal > (2⁄3) Poverty line >Poverty line Treatment 0.010 0.069* 0.063* 0.029 (0.027) (0.037) (0.037) (0.036) Observations 3,955 3,955 3,955 3,955 R-squared 0.067 0.098 0.118 0.135 State FE YES YES YES YES Month FE YES YES YES YES Controls YES YES YES YES Controls interacted YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: *p<0.05. Discussion 21.  “Honesty primes” can increase consumption data accuracy if incentives to underreport exist. In line with the hypothesis, significant treatment effects are found for lower (potentially underreported) consumption levels. More- over, effects are stronger for the number of consumption items which are more susceptible to deliberate misreporting, than for monetary consumption quantities. Thus, “honesty primes” shift certain strata of the treated but not the average population to more credible consumption levels. In this regard, the significant treatment effects are driven mainly by the vulnerable IDP subpopulation, which is more likely to have incentives to underreport to signal neediness. 26. E.g., Cilliers, Dube, and Siddiqi 2015; Bräutigam and Knack 2004. 27. There is weak evidence that the priming is more effective for respondents relying on UN aid (Table A9 in Appendix A). Moreover, IDPs have a significantly lower probability to find a lie acceptable (refer to Appendix A, Table A4). 10  |  Informing Durable Solutions for Internal Displacement 22.  Results suggest primes are a suitable tool to increase data accuracy in challenging contexts of humanitar- ian crises, but further research is needed to increase their effectiveness. Priming can help to increase the reported consumption levels. However, this analysis has two main limitations. First, it can only compare the treated group against an estimate of the “true” rates consumption. Without more objective data, it is not possible to conclusively indicate whether low consumption levels reported in the control group are not partly attributable to food shortages. Although the mortality rates among IDPs suggest that starvation is not happening systematically across the country, the precarious situation calls for further scrutiny.28 Before adjusting poverty estimates, a thorough comparison with more “objective” data from administrative, anthropometric, or observational sources is needed. Second, the intervention is bundled. For this reason, it is impossible to isolate the causal mechanism affecting the observed changes in reporting. However, if only classical measurement error would be affected, treatment effects of the primes should be uniform. In contrast, heterogenous effects across quantiles suggest that the targeting of intentional misreporting via the appeal to honesty and moral prime would be the driver of our results. Further work is needed to identify an estimate of the true level of consumption against which to compare the primed individuals and to isolate the causal mechanisms by which people are changing their behavior. More research, which unbundles the primes in different treatment arms, can contribute to the understanding of the underlying causal pathways and ultimately to developing more durable solutions. B.  Drawing Typologies of Displaced Populations 23.  IDPs and refugees can often have different trajectories in displacement. IDPs and refugees often have dif- ferent reasons for displacement, or are affected by the drivers of displacement, whether conflict, climate, or otherwise, in varying degrees. This can lead to different needs and solutions depending on the displacement trajectory, which is pertinent for policy. Displaced populations can be classified into groups that reflect their trajectory, based on a cause- needs-solution–based lens. 24.  Clustering analysis helps to identify the different typologies of the displaced. The aim of the analysis is to exploit the socioeconomic micro-level data to identify different groups or profiles of displaced households across the countries considered in Sub-Saharan Africa. Typologies of the displaced were drawn using data collected from house- hold surveys and are based on the causes of displacement, the current needs of displaced people, and the potential solution to their situation. Variables capturing similar trends were combined, and a comprehensive base of information structured around cause, needs, and solutions of the displaced was considered for the analysis of each country. 25.  The typology analysis identifies different groups in the data and checks how these groups differ in policy- relevant indicators. A series of indicators ranging from conditions of displacement and pre-displacement outcomes (cause-based lens), current socioeconomic conditions (needs-based lens), and future intentions and support required to settle anew (solutions-based lens) are used as inputs in a Multiple Correspondence Analysis (MCA). The MCA aggre- gates the inputted indicators to identify which IDP and refugee households are more similar to each other, and which ones have a lower resemblance to each other. This results in the formation of distinct groups, or typologies, of the displaced. Once these groups are derived, they are compared on a series of policy-relevant indicators (along the cause- needs-solutions–based approach) to highlight how they differ. Understanding the differences among the groups can allow better targeting of policy and programs. Clusters identified might not always align with ethnic or language groups, or locations. Thus, the scope of this analysis is to illustrate the relevance of the full displacement trajectory and 28. FEWS NET 2018. Volume C: Technical Aspects  | 11 explore how this can point to tailored solutions. However, a policymaker would likely need to rely on targeting mecha- nisms that help identify individual households. 26.  Results and policy implications of the typology analysis are detailed in Volume A of this report. The typol- ogy analysis identifies groups among the displaced populations in Ethiopia, Nigeria, Somalia, South Sudan, and Sudan. Groups derived from the typologies typically varied on key characteristics such as factors driving displacement, house- hold characteristics and living standards, labor force participation and livelihood structures, and future intentions to stay, return, or resettle. The detailed results along with policy implications are specified in the corresponding country case studies in Volume B of the report. Rationale 27.  Implementing development policies that bring a durable solution to forcibly displaced populations requires an evidence-based approach. Forced displacement has become a priority for policy makers and the inter- national community in many conflict-affected states. A shift in the approach from providing humanitarian assistance to searching for development solutions requires detailed information. This report aims to close some of the data gaps to support development policies aimed at ending displacement. 28.  Effective policies conducive to a durable solution should consider the different profiles and needs among the displaced population. A pre-condition for evidence-based planning is data to understand the links between socioeconomic dimensions and the displacement situation of populations, and to address vulnerabilities and support their economic self-reliance. Furthermore, an assessment of the best durable solution for a household or groups of households should inform the design of policies, which requires identifying groups with different displace- ment profiles. 29.  Quantitative socioeconomic micro-level data can be used to identify a typology of the displaced. The analysis is based on the causes of displacement, the current needs of displaced people, and the potential solution to their situation. Data collected from household surveys is used to identify groups of displaced populations with different profiles, with the aim of generating a comprehensive evidence base. Due to the lack of quantitative socioeconomic data on displaced populations, the knowledge generated is of unique value for durable solutions policies. Methodology 30.  Clusters of similar displaced groups are identified using Multiple Correspondence Analysis. A Multiple Correspondence Analysis (MCA) was implemented using a comprehensive set of indicators about the pre-displace- ment outcomes of the IDPs and refugees and their reasons for leaving (cause-based lens), their current living standards and household characteristics (needs-based lens), and their intentions to move and settle anew (solutions-based lens). The Joint Correspondence Analysis (JCA) approach is used to run the MCA, though other approaches such as the Burt matrix approach and the Indicator matrix approach also yield the same clusters.29, 30 The MCA yields a low-dimensional 29. There are two main methods to implement the MCA: using the Burt matrix or the Indicator matrix. Both approaches are equivalent in terms of how they cluster variables and households. However, differences in the estimation of these matrices consistently leads to a lower overall inertia from the Burt matrix, which results in a higher percentage of inertia explained by the principal dimensions. 30. According to Greenacre (2017), the share of inertia explained when using the Burt or Indicator matrix is artificially low. Thus, Joint Correspondence Analysis (JCA) is used to do the analysis and obtain an accurate estimate of the degrees of inertia explained by the model. 12  |  Informing Durable Solutions for Internal Displacement space (typically 2-dimensional or 3-dimensional) that captures the multiple indicators. Visualizing the displaced house- holds in this low-dimensional space shows that the displaced cluster in distinct groups, in all five country datasets. 31.  A Hierarchical Clustering Analysis is applied to the dimensions to obtain the final clusters. Hierarchical Clus- tering Analysis (HCA) is a method which seeks to identify distinct groups of observations from creating clusters in a hierarchical process. The method employed corresponds to the agglomerative clustering which initially considers each observation as a separate cluster, and then merges the closest two groups while moving up the hierarchical structure. The analysis uses the Wards’ linkage method, which joins the two groups and results in a minimum increase in the error sum of squares.31 The linkage method refers to how the approach chooses the similarity or dissimilarity when compar- ing two observations. To decide the optimal number of clusters, the visualization of the MCA household point-clouds, the visualization of the dendograms from the HCA, and an optimal stoppage-rule were studied. The analysis uses the Calinski-Harabasz pseudo-F index stopping rule.32 Stopping rules refer to the cut-off point in the hierarchy to define the number of groups that result from the analysis. 32.  A share of the sample is excluded to check the robustness of the clustering. Varying shares of the sample are randomly excluded, and the analysis is redone considering only that subsample. The results are generally robust to the random exclusion of 15 to 25 percent of the original sample. Most households do not change their allocation of group when restricting the sample, relative to their original allocation when using the whole sample. Key Country Results 33.  Typology analysis for Ethiopia identifies two groups of refugees with different profiles. Group 1 and Group 2 account for 78 percent and 22 percent of the survey population, respectively (Figure B.1). The displacement situation of each group is different according to the conflict context in each country of origin. Group 1 is mainly com- posed of households displaced by armed conflict, coming from South Sudan, Somalia, and Sudan, while refugees from Group 2 come mostly from Eritrea and Somalia, and were displaced for various reasons. Before displacement, Group 1 was more inclined to an agricultural livelihood and Group 2 to depend on aid and remittances. Currently, Group 1 has larger households, which are less likely to have Ethiopian relatives and are usually headed by women without educa- tion. Group 2 is more likely to have better access to services and to rely on aid and remittances; yet Group 1 is poorer, has deeper poverty, and are more likely to face food insecurity. As for the future, most households in Group 1 prefer to stay in the camp for security reasons, while most refugees in Group 2 intend to move to another country, guided by access to land, services, and employment. 31. Ward (1963) presented a general hierarchical clustering approach where groups were joined to maximize an objective function. He used an error-sum- of-squares objective function. 32. Milligan and Cooper (1985) evaluated 30 stopping rules and concluded the best rules were the Calinski-Harabasz index and the Duda-Hart index. Volume C: Technical Aspects  | 13   FIGURE B.1    Clusters of households in 2D and 3D for Ethiopia     Source: Authors’ calculation using SPS 2017. 34.  Two distinct groups of IDPs are identified in the analysis for Nigeria. Group 1 represents 74 percent of the IDP population and Group 2 around 26 percent (Figure B.2). The Boko Haram conflict led to a steep number of deaths and displacements. The place of origin is similar between both groups of IDPs. Even though most households were dis- placed by armed conflict, Group 1 is slightly more likely to cite this reason compared to Group 2. Before displacement both groups had similar living conditions, yet Group 2 was more inclined to an agricultural livelihood, and Group 1 was more likely to rely on wages, salaries, or their own business. Currently, households in Group 1 have less members, a higher dependency ratio, and are more likely to be headed by an unemployed woman. IDPs in Group 2 are more likely to have an agricultural livelihood and to receive assistance, although both groups are equally poor and food insecure. The differences in housing conditions and access to services between groups are determined by their current location, as Group 1 is more likely to live in host communities, whereas Group 2 in settlements or camps. IDPs in Group 2 were more satisfied before displacement, are more dissatisfied today, are less likely to feel safe, and are more pessimistic about the future. Households in Group 1 prefer to stay in their current location motivated by security reasons, while IDPs in Group 2 intend to return to their place of origin guided by access to land, services, and employment.   FIGURE B.2    Clusters of households in 2D and 3D for Nigeria   Source: Authors’ calculation. 14  |  Informing Durable Solutions for Internal Displacement 35.  Two typology profiles can be distinguished among Somali IDPs. The groups differ in their displacement trajectories, particularly from the cause-based or needs-based lens. Group 1, which accounts for about 40 percent of IDPs, had more agricultural livelihoods at the origin, was more likely to be drought-displaced, and had poorer living conditions before displacement and currently. Group 2, while less agricultural and with better housing at the origin, was more likely to be displaced by armed conflict (Figure B.3). The households of Group 2 are less poor, less hungry, and in better housing than those in Group 1. The differences among the groups, while still present, are smaller in the solution-based lens. Most members of both groups prefer to stay in the current location rather than return or relocate, though Group 2 is more likely to be guided by security and Group 1 by basic amenities and livelihoods in addition to security.   FIGURE B.3    Clusters of households in 2D and 3D for Somalia   Source: Authors’ calculation. 36.  IDPs in South Sudan have two distinct typologies. The two groups, Group 1 and Group 2, are of a roughly equal size, and represent 40 and 60 percent of the IDPs, respectively (Figure B.4). Before displacement, Group 2 had more agricultural livelihoods, worse housing, and was more likely to be displaced by armed conflict. Contrary to this, Group 1 had wage- and business-based livelihoods. A majority of Group 1 was also driven by armed conflict, but to a lesser degree than Group 2. Differences in the current conditions of the two groups are possibly driven from their dif- displacement situations. Group 2 has larger households, higher dependency ratios, higher poverty and aid ferent pre-­ dependence, and feels less safe. Group 2 is more likely to be confident of a moving timeline, but also more likely to seek information to decide on a move, while Group 1 is more optimistic about the future. Group 1 is largely in Juba PoC and Bor PoC, and Group 2 concentrated in Bentiu and Bor PoCs. Volume C: Technical Aspects  | 15   FIGURE B.4    Clusters of households in 2D and 3D for South Sudan   Source: Authors’ calculation. 37.  IDPs in Sudan have two different typology profiles. Group 1 and Group 2 account for 39 and 61 percent of the surveyed population, respectively (Figure B.5). Before displacement, the two groups had different sources of income and households from Group 1 were more likely to be displaced in 2003 and 2004. Households from Group 1 are also more likely to live in shelter provided in the camp, and thus are closer to services and more likely to have improved water sources. There are also differences in the current conditions of the two groups. Group 1 has a higher and deeper poverty incidence, and is more likely to face food insecurity and to rely on assistance from development partners or NGOs. As for the future, most households in Group 2 prefer to stay in the camp guided by security reasons, while a majority of IDPs in Group 1 want to relocate based on employment conditions and other considerations. The timeline for moving is clearer for IDPs in Group 2, and they are more likely to have all the information they need to inform their decision.   FIGURE B.5    Clusters of households in 2D and 3D for Sudan   Source: Authors’ calculation. Appendices Appendix A:  Eliciting Responses from Honesty Primes Experimental Design, Measures, and Sampling Table A1 describes the bundle of “honesty primes” and the questionnaire structure.   TABLE A1    Treatment questions Question type Question or answer options Introduction: Food module [Note for Enumerators]: This is the beginning of Module E, collecting information about the food consumption for the household. [Note for Enumerators]: It is very important that you take the time to collect all consumption data (including all quantities and prices). If not, analysts will follow up with what happened during your interview. [Appeal to honesty]: Thank you for taking the time to speak to us. We really appreciate the time you are giving to participate in the survey. We encourage you to provide honest information. By participating in the survey and by providing accurate information, you are playing an important role in helping us understand the situation in South Sudan. [Moral prime]: I will give you a little scenario and would like to know what you think: John asks his good 1. Yes, it is okay for Deng to friend Deng if he has some money that he can lend him to help pay for medicine for his sick son. Deng lie to John. has money but was planning to buy cigarettes with it. He lies and tells John that he has none. Is it okay for Deng to lie to John? 2. No, it is not okay for Deng to lie to John. [Investigative probing]: When was the last time that any of the household members ate some food? 1. Today 2. Yesterday 3. 2 days ago 4. 3 days ago 5. 4 days ago 6. 5 days ago 7. 6 days ago 8. Did not have a meal in the last 7 days. (continued) 17 18  |  Informing Durable Solutions for Internal Displacement   TABLE A1    Continued Question type Question or answer options Item category subsection: Bread and Cereals [Investigative probing]: When was the last time that any of the household members had Bread and 1. Today Cereals? 2. Yesterday 3. 2 days ago 4. 3 days ago 5. 4 days ago 6. 5 days ago 7. 6 days ago 8. Did not have a meal in the last 7 days. Over the past 7 days, did anyone in your household consume any of the items listed below. Please record . . . list of items and yes/ yes or no for each of the items below. no . . . [Consumption module]: questions asked for each item consumed, i.e., quantity consumed, quantity purchased, price of purchase, etc. . . . [Investigative probing]: (If respondent answers no to all items, but had answered [1–7] in the investigative 1. Yes probing question): You had earlier said that the household had Bread and Cereals. However, you did not report any consumption of Bread and Cereals in the last 7 days. Are you sure this is correct? 0. No This structure is repeated for the following item categories: (i) Bread and Cereals, (ii) Meat, (iii) Fruits, (iv) Pulses and Vegetables. The remaining items are asked without the investigative probing. Volume C: Technical Aspects  | 19 Table A2 describes the balance of primed (treatment) and non-primed (control) individuals. Some unbalances occur with regard to household size and the gender of the household head. However, the differences are small and can be controlled in our regression framework.   TABLE A2    Balance treatment and control   Control Treatment Difference, p-value Household size 4.835 5.098 0.003** (0.060) (0.064) Gender of household head 0.492 0.448 0.005** (0.011) (0.011) Literacy of household head 0.507 0.529 0.155 (0.011) (0.011) Household head completed some primary school 0.540 0.563 0.133 (0.011) (0.011) Is the household head employed 0.328 0.319 0.555 (0.010) (0.010) Share of children in household 0.364 0.373 0.309 (0.006) (0.006) Share of elderly in household 0.011 0.010 0.582 (0.002) (0.001) First component of Asset Principal Component Analysis –0.126 –0.194 0.162 (0.037) (0.032) N 2,079 2,066 Proportion 0.502 0.498 Note: Standard errors in parentheses; *p<0.05, **p<0.01.   TABLE A3    Treatment distribution by survey strata Treatment for “honesty primes” State/camp Control no. Treatment no. Total no. CRS Juba PoC 223 263 486 Wau 294 284 578 Bor 292 257 549 Bentiu 294 297 591 IDPCSS Juba PoC1—IDPCSS 976 965 1,941 HFS—Wave 4 Warrap 60 60 120 Northern Bahr el Ghazal 50 61 111 Western Bahr el Ghazal 62 58 120 Lakes 50 54 104 Western Equatoria 54 50 104 Central Equatoria 38 40 78 Eastern Equatoria 74 70 144 Total 2,467 2,459 4,926 Source: Authors’ calculations using HFS 2017, IDPCSS 2017 and CRS 2017. 20  |  Informing Durable Solutions for Internal Displacement   FIGURE A1    Respondents’ answers to moral priming “John asks his good friend Deng if he has some money that he can lend him to help him pay for medicine for his sick son. Deng has money but was planning to buy cigarettes with it. He lies and tells John that he has none. Is it okay for Deng to lie to John?” Yes, it is okay for Deng to lie to John. No, it is not okay for Deng to lie to John. Source: Based on the “honesty primes” implemented in the HFS 2017, IDPCSS 2017, and CRS 2017. An overwhelming majority of respondents disapproved lying in the fictional scenario, suggesting a positive response to the prime. Less than 10 percent of respondents answered that it is ok for the character in the fictional scenario to lie to his friend (Figure A1). The respondents, who find a lie inappropriate, have a higher share of male and unemployed household heads (refer to Table A4). Moreover, the share of non-IDPs is proportionally higher among the respondents, who would find a lie acceptable. This might point to a difference in social values, but might also be affected by social desirability concerns.33 Caloric consumption measures: All surveys provide information on food consumption quantities, which can be mul- tiplied with average energy load to estimate caloric intake per capita.34 The caloric intake is estimated as follows: caloric 1 intakei = hh_sizei Σitemj * caloriesj * quantityij. We account for the fact that 43 percent of household members are children by using adult equivalents for the scaling. Adult equivalents can account for the fact that children have lower consump- tion levels than adults. One common approach is to use the OECD scale, which scales consumption of additional adults per household by factor 0.7 and of children by factor 0.5.35 33. Cilliers, Dube, and Siddiqi 2015. 34. The high frequency surveys focus on core consumption in order to allow for complete data collection despite the unstable context. However, by design these items capture the lion’s share of consumption (on average approximately 99 percent of total consumption in more comprehensive CRS surveys). 35. For an overview on the use of adult equivalents, please consult Haughton and Khandker (2009). Volume C: Technical Aspects  | 21   TABLE A4    Balance of accepting or rejecting a lie Yes, it is okay for No, it is not okay Deng to lie to for Deng to lie to (1) vs. (2), John John Overall p-value Household size 5.041 5.123 5.119 0.696 (0.228) (0.061) (0.059) Gender of household head 0.327 0.456 0.445 0.000*** (0.032) (0.011) (0.010) Literacy of household head 0.544 0.532 0.533 0.734 (0.034) (0.011) (0.010) Household head completed some primary school 0.565 0.568 0.568 0.919 (0.034) (0.010) (0.010) Is the household head employed? 0.184 0.279 0.270 0.003*** (0.026) (0.009) (0.009) Share of children in household 0.315 0.356 0.353 0.042*** (0.019) (0.006) (0.006) Share of elderly in household 0.014 0.015 0.015 0.890 (0.007) (0.002) (0.002) Level of education of household head 2.060 1.967 1.975 0.205 (0.075) (0.022) (0.021) Non-IDPs (2) / IDPs (1) 1.212 1.155 1.160 0.029 (0.028) (0.008) (0.007) N 217 2,238 2,455 Proportion 0.088 0.912 1.000 Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Regression Framework and Point Estimates The model we are estimating can be expressed as such: Yi = b0 + b1Ti + b2Xi + b3Xi * Ti + gs + at + ei, (1) where Yi denotes the log of dependent variables for household i, and ei is the idiosyncratic error term. Ti is the treatment dummy. Xi denotes a vector of control variables generally associated with consumption, including household size, the gender of the household head, and the proportion of children (under 18) in the household. Moreover, we add an asset index based on the first component of a principal component analysis.36 Also, gs state fixed effects, and at month fixed effects are added. As the treatment might interact with the unbalanced covariates, it makes sense to add also Ti * Xi, the inter- action of the unbalanced controls with the treatment variable to the regression.37 Therefore, we also add the interaction of control variables and the treatment dummy Xi * Ti. Household size and households’ head gender were 36. Using the first component of a PCA is also advocated by Filmer and Pritchett (2001) or McKenzie (2005). As assets (bikes, fans, rickshaws, etc.) can be more easily surveyed by enumerators, those are likely to capture part of the household wealth. 37. Baranov et al. 2017; Lin and Green 2016. 22  |  Informing Durable Solutions for Internal Displacement partly unbalanced (refer to Table A2). This might be of particular importance as larger households buy larger quantities and, hence, consume more while paying lower bulk purchasing prices.38 There is also evidence for this in our sample (refer to Table A13). Controlling for those factors, the main treatment effects are less affected by heterogeneous effects of household characteristics. This regression model is then applied in a quantile regression framework (Table A5). The idea of the quantile regression framework is to take the entire distribution of the dependent variable into account. This is done by estimat- ing several regressions, which put more weight to the quantile of interest.39 This way, differential effects conditional on the quantile of the dependent variable are obtained. Further, it has the advantage of being less prone to outliers and non-normality of the error term.   TABLE A5    Results from quantile regressions of different outcome variables (1) (2) (3) (4) Outcome variables ln(cons. num.) ln(cons. quant.) ln(cons. val.) ln(cons. cal.) Q0.1 0.165** 0.342*** 0.079 0.235* (0.064) (0.079) (0.068) (0.127) Q0.25 0.058** 0.201*** 0.198*** 0.140* (0.028) (0.067) (0.053) (0.080) Q0.5 0.018 0.136** 0.119** 0.042 (0.032) (0.056) (0.050) (0.062) Q0.75 0.047 0.114** 0.071 0.032 (0.034) (0.050) (0.051) (0.067) Q0.9 –0.016 0.049 –0.015 0.013 (0.028) (0.050) (0.054) (0.064) Observations 3,955 3,955 3,955 3,955 Month FE YES YES YES YES State FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1-Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. Constraining the sample only on non-IDPs (Table A6), we can estimate our results analogous to Table A4. The pattern of positive and significant treatment coefficients in the lower quantiles vanishes, except for Column (1). In line with our hypothesis, the honesty primes applied are more efficient for the vulnerable IDP population. Those have higher incentives to indicate need than the non-IDPs. However, one should be careful not to draw strong conclusions from these results, as the number of observations is limited in this comparatively small subsample. 38. Deaton and Paxson 1998. 39. Koenker and Bassett 1978. Volume C: Technical Aspects  | 23   TABLE A6    Quantile regressions—reduced sample (only non-IDPs) (1) (2) (3) (4) Outcome variables ln(cons. num.) ln(cons. quant.) ln(cons. val.) ln(cons. cal.) Q0.1 –0.027 –0.069 –0.026 0.032 (0.079) (0.102) (0.110) (0.113) Q0.25 0.148** –0.052 0.012 –0.057 (0.073) (0.095) (0.107) (0.122) Q0.5 0.067 –0.041 –0.032 0.044 (0.067) (0.081) (0.100) (0.100) Q0.75 –0.071 –0.072 –0.015 –0.052 (0.054) (0.080) (0.092) (0.080) Q0.9 –0.041 0.157 0.074 0.119 (0.047) (0.105) (0.144) (0.127) Observations 780 780 780 770 Month FE YES YES YES YES State FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. Robustness and Channels Although no average treatment effect is expected, it is informative to consider the pure treatment effect as a potential falsification test. The unbalanced controls could be linked to heterogenous treatment effects, those inter- actions are considered in separate regressions.40 The pure treatment affects the number of consumption items signifi- cantly positive (Figure A2). For the other outcomes, results look mixed. The coefficients indicate on average a change in reported consumption by –2 to +14 percent. The negative coefficients for consumption values are counter-intuitive, though insignificant. Moreover, once we control for the interaction of the controls with the treatment, the coefficients for consumption values turn positive, suggesting household heterogeneity.41 Heterogenous treatment effects support the choice of quantile regressions as the main specification. 40. This is also suggested by Baranov et al. 2017. 41. The interactions of the treatment and the asset index, as well as household size, have negative and significant coefficients in line with previous work (see Table A8). 24  |  Informing Durable Solutions for Internal Displacement   FIGURE A2    Treatment effects from different specifications Note: Based on regression coefficients from Table A7 and Table A8.   TABLE A7    Results from equation (1) without controls and with controls. (1) (2) (3) (4) (5) (6) (7) (8) ln(cons. ln(cons. ln(cons. ln(cons. ln(cons. ln(cons. ln(cons. ln(cons. Variables num.) num.) quant.) quant.) val.) val.) cal.) cal.) Treatment 0.035** 0.034** 0.028 0.042** –0.018 –0.001 0.019 0.054** (0.016) (0.014) (0.018) (0.018) (0.018) (0.017) (0.028) (0.027) Observations 3,955 3,955 3,955 3,955 3,955 3,955 3,955 3,955 R-squared 0.001 0.273 0.001 0.070 0.000 0.078 0.000 0.123 State FE NO YES NO YES NO YES NO YES Month FE NO YES NO YES NO YES NO YES Controls NO YES NO YES NO YES NO YES Controls interacted NO NO NO NO NO NO NO NO Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Columns (1–2) are measured on the household level. Columns (3–8) refer to per capita OECD adult equivalents. Volume C: Technical Aspects  | 25   TABLE A8    Results from equation (1) with interacted controls (1) (2) (3) (4) ln(cons. num.) ln(cons. quant.) ln(cons. val.) ln(cons. cal.) Treatment 0.061* 0.137*** 0.081* 0.001 (0.033) (0.042) (0.039) (0.067) Treatment*HH size –0.008 –0.015** –0.009 0.006 (0.006) (0.007) (0.007) (0.011) Treatment*HHH gender –0.020 –0.007 –0.008 –0.059 (0.028) (0.036) (0.036) (0.054) Treatment*Share children –0.001 –0.066 –0.107 –0.016 (0.059) (0.074) (0.073) (0.116) Treatment*Asset index –0.017** –0.003 –0.007 –0.003 (0.007) (0.011) (0.010) (0.016) Observations 3,955 3,955 3,955 3,955 R-squared 0.274 0.073 0.080 0.123 State FE YES YES YES YES Month FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. Heterogenous treatment effects by aid reliance Part of the respondents from the CRS and HFS were also interviewed regarding their previous support through UN agencies. This dummy indicator can also be used to assess heterogenous effects for subpopulations that were previously exposed to UN support.42 The model is analogous to equation (1), where we add UN assistance as a further control variable as well as an interaction term of UN assistance with the behavioral treatment. Only for the number of consumption items is a positive significant coefficient found. However, the effect for the other three outcomes of interest is insignificant, suggesting weak evidence that the priming is more effective for aid-exposed IDPs (see Table A9). 42. The results can only be interpreted as an explorative analysis, as UN assistance was not balanced across treatment and control groups, where treatment households have a higher probability of being previously exposed to aid. 26  |  Informing Durable Solutions for Internal Displacement   TABLE A9    Channel—UN assistance (1) (2) (3) (4) ln(cons. num.) ln(cons. quant.) ln(cons. val) ln(cons. cal.) Treatment 0.100 0.195** 0.171* 0.105 (0.066) (0.080) (0.081) (0.087) Treatment*UN assistance 0.104** –0.059 0.016 0.011 (0.051) (0.060) (0.061) (0.064) Observations 2,204 2,204 2,204 2,204 R-squared 0.38 0.086 0.098 0.108 State FE YES YES YES YES Month FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1 Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. Robustness of results using unconditional quantile regression Conditional quantile regressions are sometimes critiqued on the ground, as they would consider the treat- ment effect conditional on the distribution and not on the individual ranking. Therefore, we also replicate the main regressions within an unconditional quantile regression framework.43 Table A10 depicts the results of uncon- ditional quantile regressions. The results indicate a comparable pattern to Table A5. Especially, the quantities of con- sumption items and kilograms experience positive treatment effects in lower quantiles. Although higher quantiles are affected as well in Column (2), the largest effects can be found in the 10 percent quantile, which would be consistent with the hypothesis of more honest answers among potentially underreporting households.   TABLE A10    Results from unconditional quantile regressions of different outcome variables (1) (2) (3) (4) Outcome variables ln(cons. num.) ln(cons. quant.) ln(cons. val.) ln(cons. cal.) Q0.1 0.105** 0.259*** 0.076 0.134 (0.046) (0.090) (0.079) (0.145) Q0.25 0.078** 0.210*** 0.169*** 0.075 (0.032) (0.067) (0.062) (0.077) Q0.5 0.004 0.104** 0.118** 0.071 (0.035) (0.053) (0.056) (0.063) Q0.75 –0.012 0.132** 0.067 0.025 (0.040) (0.066) (0.059) (0.089) Q0.9 0.024 0.075 –0.003 0.062 (0.044) (0.087) (0.077) (0.119) Observations 3,955 3,955 3,955 3,955 Month FE YES YES YES YES State FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. 43. Firpo, Fortin, and Lemieux, “Unconditional Quantile Regressions.” Volume C: Technical Aspects  | 27   TABLE A11    Quantile regressions—outcomes in levels (1) (2) (3) (4) Outcome variables cons. num. cons. quant. cons. val. cons. cal. Q0.1 0.544** 0.741*** 9.440 229.126* (0.254) (0.173) (13.305) (136.013) Q0.25 0.298* 0.675** 60.585*** 179.447 (0.156) (0.224) (16.261) (157.584) Q0.5 0.151 0.638* 49.404** 197.589 (0.194) (0.280) (20.477) (192.043) Q0.75 0.341 0.700** 19.499 281.317 (0.246) (0.339) (30.762) (286.250) Q0.9 –0.077 0.609 –30.117 –279.159 (0.333) (0.540) (41.581) (284.569) Observations 3,955 3,955 3,955 3,955 Month FE YES YES YES YES State FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. The role of outliers In line with hardly credible low consumption levels, misreporting could be considered to be more prevalent at the tails of the distribution, hence, among the extreme values. On the one hand, it makes sense to consider those outliers. On the other hand, one would like to avoid basing the inference mainly on those extreme values. Ideally, one would know how to distinguish the intentionally misreported outliers and the ones that are caused by errors in reporting or data entry. The log normalization in the main analysis is chosen as a compromise of keeping the data, but making it less susceptible to outliers. Two corresponding robustness checks are carried out: (i) in a more liberal setting; the outcomes in levels can be used, and (ii) in a more conservative setting; the outliers at the 5th and 95th percentile are discarded. Regression results using the levels are depicted in Table A11.44 Table A12 describes the results without outliers and indicates a slightly less nuanced pattern. In line with our hypothesis of stronger misreporting tendencies on the extremes, Column (1) indicates significant treatment effects at the 10th and 25th percentile. Although significant treatment effects at higher consumption levels can be found in Columns (2) and (3), the coefficients for the 25th percentile are quantitatively larger. Finally, with regard to caloric con- sumption in Column (4), statistical significance vanishes, but the largest coefficient is to be found in the 10th percentile. Hence, although the pattern gets weakened when excluding outliers, the primes still significantly affect the reported consumption quantities, mainly in the lower quantiles. 44. As scaling of the outcome variables is different—e.g., the outliers with regard to consumption quantity in kilograms might not correspond to the consumption quantity in calories—the outliers for one measure do not always correspond to outliers among the other measure. In order to guarantee that we still base the inference on the same observations, outliers from all corresponding variables are dropped, which explains that the resulting sample is smaller than 90 percent of the full sample. 28  |  Informing Durable Solutions for Internal Displacement   TABLE A12    Quantile regressions—without outliers (1) (2) (3) (4) Outcome variables ln(cons. num.) ln(cons. quant.) ln(cons. val.) ln(cons. cal.) Q0.1 0.124** 0.106 0.085 0.058 (0.049) (0.067) (0.064) (0.091) Q0.25 0.045* 0.139** 0.162*** 0.042 (0.027) (0.055) (0.044) (0.077) Q0.5 0.000 0.065 0.119** 0.037 (0.032) (0.050) (0.046) (0.059) Q0.75 0.028 0.077* 0.086* 0.049 (0.032) (0.043) (0.048) (0.063) Q0.9 –0.027 0.064 0.027 0.039 (0.023) (0.039) (0.049) (0.051) Observations 3,711 3,605 3,576 3,500 Month FE YES YES YES YES State FE YES YES YES YES Controls YES YES YES YES Interacted controls YES YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1. Column (1) is measured on the household level. Columns (2–4) refer to per capita OECD adult equivalents. Robustness to per capita instead of per adult equivalents There is some uncertainty about the per adult equivalent scaling in the data. Ideally, the distribution might be estimated from more fine-grained data on the intra-household consumption distribution. As detailed data is often not available, Deaton and Zaidi (2002) conclude that “no satisfactory” scaling method is identified so far. Therefore, the OECD scaling methodology is still frequently used.45 Yet, potential concerns with regard to the main results’ stability with regard to the scaling by adult equivalents remains. The estimates indicate that the treatment effects remain stable using agnostic per capita scales. Using quantile regressions, Table A13 indicates that respondents would report statistically significantly higher quantities in Column (1) and Column (2) if treated. Hence, scaling does not explain our results, but is a factor to take into account when inter- preting the outcomes. 45. E.g., Euler et al. 2017; Van Den Broeck and Maertens 2017. Volume C: Technical Aspects  | 29   TABLE A13    Quantile regression with per capita scaling (1) (2) (3) Outcome variables ln(cons. quant. p.c.) ln(cons. val. p.c.) ln(cons. cal. p.c.) Q0.1 0.358*** 0.040 0.207 (0.087) (0.068) (0.135) Q0.25 0.161*** 0.160** 0.076 (0.059) (0.053) (0.081) Q0.5 0.124*** 0.079 0.073 (0.057) (0.054) (0.066) Q0.75 0.050 0.055 0.021 (0.049) (0.054) (0.071) Q0.9 0.057 –0.003 0.027 (0.063) (0.051) (0.081) Observations 3,955 3,955 3,955 Month FE YES YES YES State FE YES YES YES Controls YES YES YES Interacted controls YES YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1 Supplementary Figures and Graphs Figure A3 indicates that IDPs show a more calorie intensive food portfolio than non-IDPs. This can explain why IDPs partly consume more calories than non-IDPs despite lower monetary consumption values.   FIGURE A3    Consumption shares (SSP values) Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: The figure only lists the consumption shares of items that constitute at least 1% of household consumption. 30  |  Informing Durable Solutions for Internal Displacement Figure A4 supports the notion that larger households pay lower bulk prices in line with Deaton and Paxson (1998).   FIGURE A4    Correlation of household size and purchasing prices per kilo (1) (2) Variables Per kg price Per kg price cleaned Household size –12.188 –1.025*** (9.352) (0.248) Observations 18,743 24,409 R-squared 0.049 0.580 Item FE YES YES Source: Authors’ calculations using HFS 2017, IDPCSS 2017, and CRS 2017. Note: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Figure A5 indicates the caloric consumption values for a related survey, which was carried out in Somalia. The low levels of calorie food intake motivated the experiment, as the low consumption levels seemed unlikely to be true.   FIGURE A5    Accessibility rate of urban and rural areas .0005 .0004 .0003 Density .0002 .0001 0 ) ) ) (S (M (R 0 0 20 10 1, 2, Calorie consumption—(S = Subsistence; M = Median; R = Recommended) Calorie Profile IDPs Somalia Source: Authors’ calculations using HFS Somalia Wave 1.   TABLE A14    Accessibility rate of urban and rural areas Region Urban areas Rural areas Mogadishu 87% N/A North-East 99% 89% North-West 98% 97% Central regions 77% 52% Jubbaland 64% 26% South-West 50% 34% Source: Authors’ calculations based on the SHFS 2017–18. Volume C: Technical Aspects  | 31 Appendix B: Typologies: Country-Wise Methodology Outputs Ethiopia   TABLE B1    Summary of the data and contribution to total inertia for Ethiopia % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Displacement in or after 2011 3,618 0.8 0.4 0 1 1 Displacement reason 3,612 1.8 1.1 1 4 5 Household members separated 3,603 0.2 0.4 0 1 2 Change in household size after displacement 3,252 0.1 0.4 0 1 2 Housing safe in place of origin 3,627 0.3 0.5 0 1 2 Housing pre-displacement 3,590 2.5 0.9 1 5 8 Tenure pre-displacement 3,627 1.1 0.3 1 2 1 Crowded dwelling pre-displacement 3,627 0.4 0.5 0 1 2 Distance to services pre-displacement 3,583 0.4 0.5 0 1 2 Improved water sources pre-displacement 3,513 1.4 0.5 1 2 2 Improved sanitation pre-displacement 3,567 1.8 0.4 1 2 2 Electricity pre-displacement 3,627 0.1 0.4 0 1 2 Access to agricultural land pre-displacement 3,571 0.6 0.5 0 1 2 Main source of livelihood pre-displacement 3,627 1.5 0.8 1 3 2 Satisfaction with conditions pre-displacement 3,621 0.6 0.5 0 1 1 Household size of 6 or more 3,627 0.5 0.5 0 1 1 Dependency ratio greater than 1 3,564 0.7 0.5 0 1 2 Ethiopian relatives 3,614 0.2 0.4 0 1 2 Crowded dwelling 3,627 0.5 0.5 0 1 1 Gender of household head 3,627 1.5 0.5 1 2 1 Literate household head 3,626 0.4 0.5 0 1 2 Employment status of household head 3,627 0.5 0.8 0 3 2 Main source of livelihood 3,627 2.9 0.4 1 3 1 Received remittances 3,610 0.1 0.3 0 1 3 Received assistance 3,627 0.9 0.3 0 1 2 Food insecurity 3,536 1.7 0.8 1 3 2 Housing 3,623 1.6 0.8 1 3 5 Sanitation 3,620 1.4 0.5 1 2 2 Electricity 3,627 0.1 0.3 0 1 3 Shared toilet 3,627 0.6 0.5 0 1 3 Access to agricultural land 3,614 0.1 0.3 0 1 1 Travel outside the settlement 3,583 0.4 0.5 0 1 1 Social networks 3,627 0.7 0.5 0 1 2 Current satisfaction 3,610 0.4 0.5 0 1 2 Perception of the future 3,437 1.6 0.5 1 2 2 Safe from violence 3,594 1.2 0.4 1 2 1 (continued) 32  |  Informing Durable Solutions for Internal Displacement   TABLE B1    Continued % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Intention to move 3,622 0.6 0.5 0 1 6 Moving timeline 3,627 3.1 0.9 1 4 8 Security main reason for staying/moving 3,622 0.6 0.5 0 1 1 Push reasons 3,627 1.8 0.8 1 3 5 Pull reasons 3,627 1.7 0.8 1 3 3 Have all the information 3,623 0.8 0.4 0 1 1 Other forms for receiving information 3,623 0.3 0.5 0 1 1 Source: Authors’ calculation.   FIGURE B1    Resultant dendrogram for Ethiopia Source: Authors’ calculation.   TABLE B2    Size of each group of refuges in Ethiopia   TABLE B3    Robust check excluding a share of the sample for Ethiopia No. of % of the sample % of households % of the observations in randomly that changed population the sample excluded group Group 1 78 2,669 Group 1 15 6 Group 2 22 958 Group 2 25 8 Source: Authors’ calculation. Source: Authors’ calculation. Volume C: Technical Aspects  | 33 Nigeria   TABLE B4    Summary of the data and contribution to total inertia for Nigeria % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Displaced in 2014–2015 1,436 0.78 0.42 0 1 1 Reason for displacement 1,437 1.07 0.25 1 2 1 Household members separated 1,437 0.09 0.28 0 1 4 Change in household size after displacement 1,405 0.08 0.27 0 1 4 Housing safe in place of origin 1,437 0.11 0.31 0 1 1 Housing pre-displacement 1,434 3.26 0.71 1 5 4 Crowded dwelling pre-displacement 1,437 0.21 0.40 0 1 1 Improved sources of water pre-displacement 1,430 1.24 0.43 1 2 2 Improved sanitation pre-displacement 1,431 1.34 0.47 1 2 4 Electricity pre-displacement 1,437 0.34 0.47 0 1 3 Access to agricultural land pre-displacement 1,419 0.54 0.50 0 1 2 Source of livelihood pre-displacement 1,437 1.57 0.66 1 3 5 Overall satisfaction pre-displacement 1,437 0.76 0.43 0 1 1 Household with more than 5 members 1,437 0.53 0.50 0 1 1 Age-dependency ratio equal or larger than 1 1,409 0.75 0.43 0 1 1 Overcrowded dwelling 1,437 0.54 0.50 0 1 1 Gender of household head 1,437 1.35 0.48 1 2 1 Literate household head 1,429 0.54 0.50 0 1 2 Employment status of household head 1,437 1.04 0.72 0 3 2 Main source of livelihood 1,437 1.91 0.67 1 3 6 Received assistance 1,437 0.33 0.47 0 1 1 Food insecurity 1,406 1.70 0.87 1 3 4 Not enough money for food in past 7 days 1,434 0.60 0.49 0 1 2 Housing 1,437 1.72 0.88 1 3 3 Sanitation 1,428 1.27 0.44 1 2 3 Electricity 1,437 0.37 0.48 0 1 3 Shared toilet 1,437 0.33 0.47 0 1 1 Access to agricultural land 1,435 0.34 0.47 0 1 3 Social networks 1,437 0.34 0.47 0 1 1 Current satisfaction 1,434 0.53 0.50 0 1 2 Perception of the future 1,261 1.17 0.38 1 2 1 Relationship with neighbors 1,435 1.07 0.26 1 2 1 Intention to move 1,436 0.56 0.50 0 1 8 Moving timeline 1,437 3.22 0.88 1 4 9 Security main reason for staying/moving 1,436 0.88 0.32 0 1 1 Push reasons 1,437 1.49 0.69 1 3 6 (continued) 34  |  Informing Durable Solutions for Internal Displacement   TABLE B4    Continued % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Pull reasons 1,437 1.45 0.71 1 3 4 Have all the information 1,437 0.83 0.38 0 1 1 Source of information 1,430 1.68 0.50 1 3 2 Other forms for receiving information 1,437 0.23 0.42 0 1 1 Source: Authors’ calculation.   FIGURE B2    Resultant dendrogram for Nigeria Source: Authors’ calculation.   TABLE B5    Size of each group of IDPs in Nigeria   TABLE B6    Robust check excluding a share of the sample for Nigeria No. of % of the sample % of households % of the observations in randomly that changed population the sample excluded group Group 1 74 496 Group 1 15 12 Group 2 26 941 Group 2 25 8 Source: Authors’ calculation. Source: Authors’ calculation. Volume C: Technical Aspects  | 35 South Sudan   TABLE B7    Summary of the data and contribution to total inertia for South Sudan % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia State of origin 2,396 2.8 1.2 1 5 22 Relative location 2,388 3.6 1.2 1 6 5 Harm due to conflict 2,396 1.5 0.8 0 2 2 Reason for displacement 2,396 1.6 1.1 1 5 5 Reason for arriving 2,396 1.4 1.4 1 8 4 Ownership of livestock pre-displacement 2,396 0.7 0.5 0 1 3 Source of livelihood pre-displacement 2,395 2.3 1.4 1 6 10 Ownership of productive assets pre-displacement 2,396 0.6 0.5 0 1 4 Access to agricultural land pre-displacement 2,393 0.4 0.5 0 1 3 Improved housing pre-displacement 2,396 0.4 0.5 0 1 1 Distance to key services pre-displacement 2,396 0.6 0.5 0 1 1 Household with more than 4 members 2,396 0.6 0.5 0 1 4 Overcrowded dwelling 2,396 0.5 0.5 0 1 7 Source of livelihood 2,388 4.6 1.1 1 6 6 Access to agricultural land 2,395 0.2 0.4 0 1 2 Ownership of productive assets 2,396 0.2 0.4 0 1 3 Ownership of livestock 2,396 0.1 0.3 0 1 2 Household head literate 2,396 0.5 0.5 0 1 2 Distance to key services 2,396 1.0 0.2 0 1 1 Age dependency ratio equal or larger to 1 2,368 0.5 0.5 0 1 2 Gender of household head 2,396 0.5 0.5 0 1 2 Freedom of movement around camp 2,396 0.3 0.4 0 1 2 Time for relocation 2,396 1.4 1.6 0 4 4 Push and pull factor related to security 2,394 1.1 0.3 1 2 1 Info. needed to decide where to settle 2,392 0.6 0.6 0 2 3 Help needed to settle 2,396 1.1 0.3 1 2 1 Source: Authors’ calculation. 36  |  Informing Durable Solutions for Internal Displacement   FIGURE B3    Resultant dendrogram for South Sudan Source: Authors’ calculation.   TABLE B8    Size of each group of IDPs in South Sudan   TABLE B9    Robust check excluding a share of the sample for South Sudan No. of % of the sample % of households % of the observations in randomly that changed population the sample excluded group Group 1 40 1,213 Group 1 15 5 Group 2 60 1,183 Group 2 25 8 Source: Authors’ calculation. Source: Authors’ calculation. Volume C: Technical Aspects  | 37 Somalia   TABLE B10    Summary of the data and contribution to total inertia for Somalia % Contribution Variable Obs. Mean Std. dev Min Max to total inertia Access to assets pre-displacement 1,028 0.19 0.04 0 1 1 Distance from services pre-displacement 1,006 0.45 0.72 0 1 3 Location of origin relative to current location 973 0.69 1.36 1 4 2 Nonphysical harm experienced 1,023 0.28 0.09 0 1 1 Reason for displacement 952 1.95 3.37 1 6 8 Reason for arriving at location 978 2.18 2.28 1 8 6 Access to livestock pre-displacement 1,028 0.50 0.48 0 1 4 Source of livelihood pre-displacement 992 2.62 3.70 1 9 16 Access to agricultural land pre-displacement 1,022 0.39 0.19 0 1 4 Improved housing pre-displacement 1,007 0.47 0.33 0 1 3 Household size greater than 5 1,028 0.47 0.67 0 1 1 Improved housing 1,028 0.45 0.28 0 1 5 Improved water 1,020 0.47 0.66 0 1 1 Improved sanitation 1,015 0.50 0.45 0 1 3 Hunger 1,021 0.49 0.58 0 1 2 Source of livelihood 1,012 2.82 3.77 1 9 15 Receiving assistance 1,028 0.50 0.46 0 1 2 Access to agricultural land 1,023 0.30 0.10 0 1 1 Access to assets 1,028 0.13 0.02 0 1 1 Access to livestock 1,028 0.49 0.39 0 1 2 Household head literate 1,018 0.49 0.41 0 1 2 Distance to services 1,024 0.44 0.74 0 1 2 Dependency ratio 1 or more 1,022 0.49 0.59 0 1 2 Gender of household head 1,028 0.50 0.50 0 1 1 Freedom to move in and out of camp 1,023 0.34 0.87 0 1 1 Time of moving 1,017 1.27 0.79 0 4 3 Security guides moving decision 988 0.46 1.30 1 2 2 Information needed to decide on move 1,017 0.75 0.55 0 2 2 Help needed to settle 964 0.41 0.78 0 1 2 Source: Authors’ calculation. 38  |  Informing Durable Solutions for Internal Displacement   FIGURE B4    Resultant dendrogram for Somalia Source: Authors’ calculation.   TABLE B11    Size of each group of IDPs in Somalia   TABLE B12    Robust check excluding a share of the sample for Somalia No. of % of the sample % of households % of the observations in randomly that changed population the sample excluded group Group 1 40 405 Group 1 15 16 Group 2 60 623 Group 2 25 29.5 Source: Authors’ calculation. Source: Authors’ calculation. Volume C: Technical Aspects  | 39 Sudan   TABLE B13    Summary of the data and contribution to total inertia for Sudan % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Access to agricultural land pre-displacement 1,964 0.86 0.35 0 1 1 Source of livelihood pre-displacement 1,974 0.91 0.29 0 1 1 Dwelling pre-displacement 1,957 2.14 0.57 1 5 11 Tenure of dwelling pre-displacement 1,974 5.73 0.95 1 7 5 Crowding pre-displacement 1,974 0.11 0.32 0 1 1 Water source pre-displacement 1,974 1.43 0.50 1 2 2 Sanitation pre-displacement 1,974 1.68 0.46 1 2 3 Cooking source pre-displacement 1,974 1.43 0.76 1 3 4 Distance to services pre-displacement 1,967 0.25 0.43 0 1 1 Displaced in 2003–2004 1,974 0.57 0.50 0 1 1 District of origin within North Darfur 1,974 1.66 1.28 1 6 3 Displacement/relocation motivated by security 1,974 0.85 0.36 0 1 1 Number of times displaced 1,974 1.42 0.49 1 2 1 Overall satisfaction pre-displacement 1,974 1.31 0.46 1 2 1 Poor household 1,974 0.76 0.43 0 1 2 Household with 5 or more members 1,974 0.55 0.50 0 1 1 Dependency ratio of 1 or more 1,931 0.56 0.50 0 1 1 Female headed household 1,974 0.48 0.50 0 1 2 Literate household head 1,974 0.57 0.49 0 1 1 Decreased in access to agricultural land 1,974 1.52 0.50 1 2 1 Dwelling 1,974 1.17 0.42 1 3 2 Tenure of dwelling 1,974 1.29 0.64 1 3 4 Water source 1,974 1.22 0.41 1 2 1 Sanitation 1,974 1.27 0.44 1 2 3 Shared toilet 1,974 0.55 0.50 0 1 2 Source of lighting 1,974 2.14 0.97 1 4 4 Cooking source 1,974 1.46 0.56 1 3 3 Distance to services 1,969 0.73 0.45 0 1 1 High food insecurity 1,974 0.51 0.50 0 1 2 Household receives assistance from UN/NGOs 1,974 0.20 0.40 0 1 1 Source of livelihood 1,974 2.23 1.21 1 4 4 Overall satisfaction 1,974 0.20 0.40 0 1 1 Perception of relation between IDPs and host 1,974 0.80 0.40 0 1 1 Member has returned to place of origin 1,974 0.30 0.46 0 1 2 Time for relocation 1,974 1.57 0.70 1 4 7 Information needed to decide where to settle 1,974 0.52 0.50 0 1 2 (continued) 40  |  Informing Durable Solutions for Internal Displacement   TABLE B13    Continued % Contribution Variable Obs. Mean Std. dev. Min Max to total inertia Source of information 1,974 2.17 1.13 1 4 4 Pull factor related to security 1,974 1.43 0.76 1 3 4 Push factor related to security 1,974 1.67 0.87 1 3 8 Source: Authors’ calculation.   FIGURE B5    Resultant dendrogram for Sudan Source: Authors’ calculation.   TABLE B14    Size of each group of IDPs in Sudan   TABLE B15    Robust check excluding a share of the sample for Somalia No. of % of the sample % of households % of the observations in randomly that changed population the sample excluded group Group 1 39 701 Group 1 15 11 Group 2 61 1,273 Group 2 25 9 Source: Authors’ calculation. Source: Authors’ calculation. References Athey, S., and Imbens, G. W. 2017. The state of applied econometrics: Causality and policy evaluation. Journal of Eco- nomic Perspectives, 31(2), 3–32. Bachman, J. G., and O’Malley, P. M. 1984. Yea-Saying, Nay-Saying and Going to Extremes: Black-White Differences in Response Styles. Public Opinion Quarterly 48(2), pp. 491–509. Baranov, V., Bhalotra, S., Biroli, P., and Maselko, J. 2017. Maternal Depression, Women’s Empowerment, and Parental Invest- ment: Evidence from a Large Randomized Control Trial. Bonn: Institute for the Study of Labor (IZA). Bräutigam, D. A., and Knack, S. 2004. Foreign Aid, Institutions, and Governance in Sub-Saharan Africa. Economic Develop- ment and Cultural Change, 255–285. Cilliers, J., Dube, O., and Siddiqi, B. 2015. The white-man effect: How foreigner presence affects behavior in experiments. Journal of Economic Behavior and Organization (118), pp. 397–414. Cohn, A., Fehr, E., and Maréchal, M. A. 2014. Business culture and dishonesty in the banking industry. Nature International Journal of Science (516), 86–89. Cohn, A., Maréchal, M. A., and Noll, T. 2010. Bad Boys: How Criminal Identity Salience Affects Rule Violation. Review of Economic Studies 82(4), 1289–1308. Cronbach, L. J. 1946. Response Sets and Validity. Educational and psychological measurement (6), pp. 672–83. Deaton, A., and Paxson, C. 1998. Economies of Scale, Household Size, and the Demand for Food. Journal of Political Economy, 106(5), 897–930. Deaton, A., and Zaidi, S. 2002. Living Standards Measurement Study Working Paper No. 135. Washington DC: The World Bank. Euler, M., Krishna, V., Schwarze, S., Siregar, H., and Qaim, M. 2017. Oil Palm Adoption, Household Welfare, and Nutrition Among Smallholder Farmers in Indonesia. World Development, 219–235. FEWS NET. 2018. Famine Early Warning Systems Network. Tratto il giorno March 22, 2018 da South Sudan Food Secu- rity Outlook—October 2017 to May 2018: http://www.fews.net/sites/default/files/documents/reports/SOUTH%20 SUDAN%20Food%20Security%20Outlook_102017_0.pdf Filmer, D., and Pritchett, L. H. 2001. Estimating wealth effects without expenditure data—or tears: an application to educational enrollments in states of India. Demography, 115–132. Firpo, S., Fortin, N. M., and Lemieux, T. 2009. Unconditional quantile regressions. Econometrica (77:3), 953–973. Gilens, M., Sniderman, P. M., and Kuklinski, J. H. 1998. Affirmative Action and the Politics of Realignment. British Journal of Political Science 28(1), pp. 159–83. 41 42  |  Informing Durable Solutions for Internal Displacement Gneezy, U. 2005. Deception: The Role of Consequences. American Economic Review 95(1), pp. 384–94. Hamilton, D. L. 1968. Personality Attributes Associated with Extreme Response Style. Psychological Bulleting 69(3), pp. 192–203. Haughton, J., and Khandker, S. R. 2009. Handbook on Poverty & Inequality. Washington DC: World Bank Publications. Hurd, M. D. 1999. Anchoring and Acquiescence Bias in Measuring Assets in Household Surveys. Journal of Risk and Uncertainty 19(1/3), pp. 111–36. Koenker, R., and Bassett Jr, G. 1978. Regression quantiles. Econometrica, 46(1), 33–50. Lanjouw, P., and Ravallion, M. 1995. Poverty and household size. The Economic Journal, 1415–1434. Lin, W., and Green, D. P. 2016. Standard Operating Procedures: A Safety Net for Pre-Analysis Plans. PS: Political Science & Politics, 495–500. Mazar, N., and Ariely, D. 2006. Dishonesty in Everyday Life and Its Policy Implications. Journal of Public Policy & Marketing 25(1), pp. 117–26. Mazar, N., Amir, O., and Ariely, D. 2008. The Dishonesty of Honest People: A Theory of Self-Concept Maintenance. Journal of Marketing Research 45(6), pp. 633–44. McKenzie, D. J. 2005. Measuring inequality with asset indicators. Population Economics, 18(2), 229–260. Pape, U., and Mistiaen, J. 2015. Measuring Household Consumption and Poverty in 60 Minutes: The Mogadishu. Washington DC: World Bank/Proceedings of ABCA Conference 2015. Pew Research Center. 2018. “In Sub-Saharan Africa, Total Number of Forcibly Displaced People Increased Sharply in 2017”: https://www.pewresearch.org/fact-tank/2018/08/09/record-number-of-forcibly-displaced-people-lived-in- sub-saharan-africa-in-2017/ft_18-08-08_ssafricamigration_in-sub-saharan-afr-number-forcibly-displaced/ Rasinski, K. A., Visser, P. S., Zagatsky, M., and Rickett, E. M. 2005. Using implicit goal priming to improve the quality of self-report data. Journal of Experimental Social Psychology, 41(3), 321–327. Ravallion, M., and Bidani, B. 1994. How robust is a poverty profile? The world bank economic review, 8(1), 75–102. Rosenfeld, B., Imai, K., and Shapiro, J. N. 2016. An Empirical Validation Study of Popular Survey Methodologies for Sensi- tive Questions. American Journal of Political Science 60(3), pp. 783–802. Siegelman, L. 1981. Question-Order Effects on Presidential Popularity. Public Opinion Qarterly, pp. 199–207. Talwar, V., Arruda, C., and Yachison, S. 2015. The effects of punishment and appeals for honesty on children’s truth-telling behavior. Journal of Experimental Child Psychology, 209–217. The World Bank. 2015. World Development Report: Mind, Society, and Behavior. Washington DC: The World Bank. The World Bank. 2016, October 20. The World Bank in South Sudan. Tratto il giorno February 24, 2018 da South Sudan Overview: http://www.worldbank.org/en/country/southsudan/overview Volume C: Technical Aspects  | 43 United Nations. 2017. UN News. Tratto il giorno March 15, 2018 da Famine declared in region of South Sudan—UN: https://news.un.org/en/story/2017/02/551812-famine-declared-region-south-sudan-un UNHCR, United Nations High Commissioner for Refugees. 2019. “Figures at at Glance. Statistical Yearbooks”: https:// www.unhcr.org/figures-at-a-glance.html Van Den Broeck, G., and Maertens, M. 2017. Moving Up or Moving Out? Insights into Rural Development and Poverty Reduction in Senegal. World Development, 95–109. Vinski, M., and Watter, S. 2012. Priming Honesty Reduces Subjective Bias in Self-Report Measures of Mind Wandering. Consciousness & Cognition 21(1), 451–55. White, H. 1980. A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica, 817–38.