Cost effective panel data collection for Kampala
This Kampala based study will pilot methods for collecting low-cost, high-frequency panel survey data that improves on existing data sources in three ways.
- The data will provide information about prices (consumer prices, wages, and rents) at the individual and location level;
- The data will allow for the creation of commuting flows at the individual (or group) level;
- The data will allow for tracking of people when they move as well as places over time.
New approaches to survey data collection (for example mobile phone based surveys) mean that this data may have a higher benefit to cost ratio than existing big data (e.g. satellite or cell phone data) or administrative (e.g. census) data sources.
As existing data sources are often costly or difficult to get, frequently do not measure prices, are not individual specific, do not allow for tracking of individuals, and are available infrequently and sporadically. As a result of this, predictions about the impacts of projects such as transportation infrastructure are often based on models that have to infer key prices from location choices, and must be calibrated to data that is up to 10 years old. Further, evaluations of projects are not able to make use of modern identification techniques such as synthetic controls which require long panels.
The long-term goal is to design and collect a household and location level panel data set that overcomes these difficulties. The feasibility of such a data collection task has, however, not been demonstrated. The short-term goal for this project is therefore to determine the feasibility, cost and utility of such data collection. We will determine feasibility by piloting and checking methods, particularly methods for eliciting location-based data from households at low cost. We will determine cost through intensive piloting. We will determine utility by using the pilot data to understand how survey measured outcomes differ from inferred outcomes used in models. In this regard we are assisted by the fact that another project with the IGC is currently developing a state of the art model of Kampala. For example, we will compare actual commute times by income to those that are currently being used by the Kampala city model and we will compare rent gradients from actual data to those inferred using location wage data and population density data.
Beyond the long-term goal, which is to make the case for this type of data collection, the data will fill current knowledge gaps. For example, understanding the trade-off between commuting time and rents, and how this may differ by income group, will feed into work on an urban CGE model currently being developed by researchers with support from the IGC. If the pilot is successful, we would hope to start rollout of a survey soon that could form the baseline for a study of Kampala’s planned transport infrastructure projects.