Utz J. Pape and Gonzalo I. Nunez Chaim To embark on a sustainable pathway toward development, effective policy responses must be implemented quickly and based on evidence. This requires reliable, timely data, which is often unavailable especially in fragile settings. An innovative High Frequency Survey (HFS) infrastructure offers a modern data collection system to fill critical data gaps. It can provide quantitative data to inform programs and policies, often linked to resilience in fragile settings. Using the cases of Somalia and South Sudan, this note describes the design and setup of such a HFS infrastructure and illustrates how high frequency price data can effectively support decision-making even in the event of an economic or humanitarian crisis. Designing effective interventions requires the fly. A crucial challenge was developing an reliable and timely data especially in shocks automated system that can make cleaned and prone, fragile countries like Somalia and South processed data available instantaneously to online Sudan. Fragile and conflict-affected countries are users. Therefore, the HFS included a near real-time particularly vulnerable to shocks disrupting and dashboard that tracked market prices and currency reversing development outcomes. For example, exchange rates across several locations. Setting up Somalia has suffered from several humanitarian the survey implied a moderate fixed up-front cost for crises linked to conflict and drought. In South Sudan, the development of the software and purchase of the civil war broke out two years after independence in a hardware. Once the survey infrastructure was in challenging macro-economic environment. In such place, frequent collection of data was low-cost and settings, timely humanitarian and development required minimum supervision. In addition, the interventions are key to avoid humanitarian crises flexibility of the system permitted introducing and support the self-reliance of the most vulnerable changes without disruptions. The questionnaire was populations. High frequency price data are essential easily adjusted as needed to include new products to monitor markets and detect economic shocks to and locations. allow evidence-based programming in such 1 contexts. The High Frequency Survey (HFS) allowed for a prompt diffusion of the data to online users, Computer assisted personal interviewing (CAPI) minimized costs and facilitated adjustments on allowed real-time processing of data. The Survey 1 World Bank. "Challenges and Opportunities of High Frequency Data Collection in Fragile States: Lessons from South Sudan". World Bank, 2014. Solutions software was used to collect data. In addition, the architecture of the HFS allowed Computer Assisted Personal Interview (CAPI) makes submission of corrections for wrongly entered collected data instantaneously available in the cloud data without disrupting the automated process. –a necessary feature for a real-time monitoring Data entry errors from enumerators are usually system. Data quality was improved by preventing corrected when data is cleaned. However, manual enumerators from skipping questions and corrections are not possible in an automated system introducing dynamic checks and constraints. For like the HFS. Therefore, a correction questionnaire example, dynamic validation algorithms were used to and submission system were designed as part of the flag suspiciously high or low data entries requiring automated infrastructure to facilitate overwriting confirmation from the enumerators before wrong data entries. Corrections were submitted by proceeding. Downloading and processing of the data analysts to the cloud using tablets with the same data was automated with a scheduled API script to collection software. The processing system auto- download the latest data from the cloud server, and matically read and applied the corrections each time then execute a pipeline of Stata code to anonymize, it produced an update for the dashboard. clean and process the data. Finally, the script updated the data repository access by the Tableau server to visualize the data online. Once the system was in place, enumerators collected data regularly with the In Somalia, the Market Price Survey (MPS) and information becoming automatically available online Currency Exchange Rate Survey (CERS) were within minutes of data collection. easily adjusted to increase the coverage of The data analysis system of the HFS automatically products and markets. The HFS infrastructure downloaded the data from the cloud server to an provides flexibility which facilitates adapting to analysis server and prepared the data before changing circumstance. Using the results from the submitting to the online dashboard. The data first wave of the Somali HFS, the list of products in downloaded from the project’s Survey Solutions the MPS were modified while coverage was extended cloud server was processed using Stata to remove to additional markets. Similarly, the ERS outliers, calculate aggregate statistics and prepare questionnaire was modified to capture exchange the data for visualization in Tableau (Figure 1). The rates from three different money traders in each processing began with anonymization and checks for market: open street hawkers, diverse service security threats like malicious code injections. providers and registered banks or forex traders. Outliers were identified with a moving average Using the ability to push questionnaire revisions standard deviation that does not affect data older remotely to enumerators, these changes did not 2 than three weeks. This allowed outlier detection to disrupt the data collection process. remain accurate despite high volatility in prices. In South Sudan, the questionnaire was modified Outliers were replaced by imputations based on a to track exchange rates offered by commercial linear interpolation of the unit price for the item. banks within 48 hours after the government abandoned the peg of the exchange rate. Between 2 A given unit price is considered an outlier if it is higher or lower than the moving average by more than 50 percent. 2015 and 2017 the value of the South Sudanese to a managed float on the 15th of December 2015. pound (SSP) declined rapidly. The government The questionnaire was modified on-the-fly to track responded by moving from a pegged exchange rate the new commercial exchange rates within 48 hours. Figure 1. Infrastructure of the HFS API data collect Enumerators Cloud server Stata server Tableau server Handheld tablets Survey Solution’s Downloads Repository with 3G/Wifi server and anonymizes, cleans updates data containing the centralized and prepares the source and user survey management data, before sending interface questionnaire system it to the repository Data submission Push data Correction submission Analysts Correction questionnaire for wrong data entries Figure 2. Exchange rate SSP/USD (national average) 200 Exchange rate SSP/USD 150 The online dashboard provided useful insights 100 into the dynamics of the severe drought that 50 affected the Somali population in 2017. In Somalia, the dashboard identified trends in specific 0 products and locations during the drought of 2017, supporting food assistance programs in deciding Data collection Commercial Parallel whether to import aid products or source them locally, as unnecessary imports could have depressed In South Sudan, the dashboard reported a long prices and disrupted livelihoods of local producers.3 period of accelerating devaluation and CPI Meat prices steadily declined in markets of inflation, providing valuable real-time Mogadishu from an average of around US$ 4.65 in information to the government and the March 2016, to just over US$3.00 in May 2017, international community. Only six months after the providing evidence of the poor conditions of government abandoned the peg, prices doubled with livestock and the desperate situation of many the official exchange rate skyrocketing from 2.95 to households selling their livestock as a coping 38 SSP/USD (Figure 2). The gap between the official mechanism to the drought. Aid imports of substitutes and parallel market exchange rates indicated the could have further depressed prices, rendering this coping mechanism less effective. 3 Available at: www.thesomalipulse.com. scarcity of foreign currency at the official rate.4 CPI important piece in the puzzle helping to monitor inflation spiraled further pushed by an insufficient markets, detect and track economic shocks. In turn, domestic production capacity and conflict related this information allows to design and implement disruptions, peaking at an annual inflation of 549 interventions, based on reliable and timey data. percent in September 2016. Rising prices, as seen in The HFS provided information in a timely and Figure 3, forced households to rely on their own food ready-to-use manner, which supported evidence- production, even though its level remained insufficient to prevent growing food insecurity, while based interventions in Somalia and South Sudan. especially urban households fell deeper into poverty. The HFS infrastructure included a near real-time dashboard that tracked market prices and currency The real-time dashboard provided this critical exchange rates across several locations. The information in a timely and high-quality manner to fieldwork strategy involved several innovations to inform programs.5 ensure real-time updates of high-quality data. In Figure 3. High Frequency Price Index for South Sudan addition, the flexibility of the infrastructure allowed 8,000 for adjustments on the fly. Although it was limited to HPI (June 2011=100) 6,000 real-time information on exchange rates and market prices, the innovations are applicable in a much wider 4,000 context of electronic data collection. Furthermore, 2,000 the innovations can be robustly implemented in most challenging field environments including fragility and 0 low-capacity, given that they were already successfully implemented in Somalia and South Sudan. In Somalia, the dashboard provided useful JNG WRP NBG WBG LKS WEQ insights into the dynamics of the drought that CEQ EEQ National affected the country in 2017. In South Sudan, the dashboard reported a long period of accelerating devaluation, providing real-time information on the dynamic of exchange rates and prices. Overcoming data collection challenges in fragile ABOUT THE AUTHOR contexts is a stepping-stone for achieving greater Utz J. Pape is a Senior Economist at the World Bank’s Poverty and Equity Global Practice (GPV). development impacts. To embark on a sustainable upape@worldbank.org pathway toward development effective policy, responses must be implemented quickly and based Gonzalo I. Nunez Chaim is a Consultant for the on evidence. High frequency price data can be one World Bank’s Poverty and Equity Global Practice (GPV). gnunez1@worldbank.org This note series is intended to summarize good practices and key policy findings on Poverty-related topics. The views expressed in the notes are those of the authors and do not necessarily reflect those of the World Bank, its board or its member countries. Copies of these notes series are available on www.worldbank.org/poverty 4 5 World Bank. "South Sudan Economic Update, 2017: Available at: www.thepulseofsouthsudan.com. Taming the Tides of High Inflation". World Bank, 2017.