Syndicate content

What did we learn from real-time tracking of market prices in South Sudan?

Utz Pape's picture

Economic shocks can be painful and destructive, especially in fragile countries that can get trapped into a cycle of conflict and violence. Effective policy responses must be implemented quickly and based on evidence. This requires reliable and timely data, which are usually unavailable in such countries. This was particularly true for South Sudan, a country that has faced multiple shocks since its independence in 2011. Recognizing the need for such data in this fragile country to assess economic shocks, the team developed a real-time dashboard to track daily exchange rates and weekly market prices (click here for instructions how to use it).

An oil-dependent country, South Sudan took a massive hit when oil prices dropped. In responding to the crisis, the real-time dashboard has been of assistance to the Government of South Sudan. Given the severity of the depreciation of the parallel market exchange rate, captured by the dashboard, the government responded by moving from a peg of 2.95 SSP/USD to a managed float on December 15th, 2015. The dashboard continues to track the response of the market indicating an almost continuous free fall reaching 81 SSP/USD in early September 2016. 
 

Integrated Dashboard


The dashboard also provides evidence for the impact of the depreciation on market prices. Prior to the official change of the exchange rate regime, prices had already increased substantially reflected by an annual inflation of 120 percent. This confirms that prices were subject to changes in the parallel market rate, which started to depreciate in early 2015. Most traders did not have access to foreign exchange at the official rate and, thus, priced products using the parallel market rate. In the first six months after the change in the exchange rate regime, prices doubled while the official exchange rate skyrocketed by 1288 percent from 2.95 SSP/USD to 38 SSP/USD. The parallel market rate increased more modestly by 247 percent to 47 SSP/USD as it started from a higher base of 19 SSP/USD in December 2015. The continued existence of the parallel market with a premium of 20 SSP/USD (September 2016) compared to the official rate indicates that foreign exchange is still insufficiently provided at the rates offered by commercial banks. All of this information drawn from the real-time dashboard is without a doubt helpful for the government in making decision and also sheds light on the reality of market prices as it provides not only timely information but also high-quality data.

Documenting such evidence would not have been possible without innovative use of technology, which allowed tailoring mobile data collection to a fragile context like South Sudan. The development of the dashboard involved several innovations to ensure both real-time updates and data reliability. Data on daily exchange rates and weekly prices are collected in 15 locations in South Sudan using Android handheld tablet computers equipped with SurveyCTO, uploaded to a SurveyCTO server in the cloud via 3G/Wifi and made automatically available in the online Tableau dashboard.

Here are a few examples of innovations that we developed. Most of them help to overcome limited field access by remotely monitoring field data collection and quality:

  • Dynamic soft constraints flag unusually high or low data entries and ask data collectors to confirm the accuracy of the entries. This method shifts data cleaning to the field, where the best knowledge is available. 
  • Random sound bites record the conversations of the data collectors at random points during the interview. Suspicious interviews can be verified with replays of those sound bites. 
  • GPS tracking software allows the tracking of locations and trajectory of data collectors, even retrospectively, when the tablet re-enters 3G/WiFi areas. This helps to determine whether interviews are conducted at the correct locations. 
  • A real-time monitoring system can identify challenges in the field such as low-performing data collectors. The early identification is used to support data collectors with specific feedback to improve their work and mitigate any negative impact on data quality. It also allows adjustments to the questionnaire on the fly. For example, the questionnaire was modified to track exchange rates offered by commercial banks within 48 hours after the country moved suddenly from a pegged to a floating exchange rate regime.
Once the data is in the cloud, it must be processed before it can be visualized. We designed a real-time data analysis infrastructure that automatically downloads the data from the cloud server onto an analysis server, processes the data and submits it to the dashboard. Data processing starts with anonymization and checks for security threats (like malicious code injections). We also developed solutions to challenges that emerged from automated data cleaning:
  1. Flagging obvious outliers. Traditional outlier detection algorithms use one or two standard deviations to mark outliers. When applied to dynamic data, the standard deviations change over time and, thus, even historic data is revised. Keeping the standard deviations constant is ill-suited when data volatility increases as in the context of high inflation. Therefore, we implemented a moving window approach that adjusts to higher volatility without changing data older than three weeks. 
  2. Correcting data entry errors. Data collectors make data entry errors that are usually fixed in the pre-processing code. But using an automated system, it is not possible to adjust the pre-processing code as any system update requires IT approval. Therefore, we designed a submission system for data corrections that allows the analyst team to over-write actual data entries from data collectors. The corrections are submitted using a questionnaire implemented with the same software (SurveyCTO) and kept in the cloud in a separate table documenting any corrections. The automated processing system reads and applies the correction table to the data.

Overcoming the challenges in data collection in fragile context, this innovative system is a stepping-stone to achieving greater development impacts. Although it is currently limited to real-time information of exchange rates and market prices in South Sudan, the innovation opens doors to broader use. It can be implemented in almost any context given that it has worked in an extremely fragile and low-capacity environment like South Sudan. In its current implementation, it has served the government of South Sudan and its development partners in designing policies and programs.

Comments

Submitted by Asha Abdel Rahim on

Dear Utz,
How are you doing. Well done paper.
It helps many, reflects the fragility of the country, but shows way forward!

Add new comment