Rwanda Rural Feeder Roads Impact Evaluation: Creating a Data Ecosystem Maria Jones 18 July 2017
national scope = unique IE opportunity Unique opportunity: usually road projects smaller in scope. GoR coordinating within financing from WB, USAID, EU and Netherlands. Policy objective: bring a motorable road within 2km of every farm in the country. Enhance market access and reduce transport costs for people as well as goods.
rehabilitation of rural feeder roads Applies to 20% of national road network. Upgrading consists of: 1. Widening to the new 6 meter standard (2 lanes), 2. Adding a base layer and resurfacing with lateritic soil, 3. Constructing drainage structures
research questions What is the impact of feeder road rehabilitation on … … market prices of village imports and exports? (trade economics) … HH adaptations to price changes in terms of goods produced and purchased? (consumer welfare functions) … market valuation of improved road access as measured by aggregate land value changes? (asset pricing approach) … Regional development as measured by total population? (welfare measure from urban economics) improved roads market efficiency faster development
design Event study (at the segment level) Exact timing of rehabilitation of any particular segment as good as random idiosyncrasies of donor calendars, construction delays, permitting, and weather Track exact start and end of construction for each segment Key explanatory variable is road roughness before and after upgrading Identification relies on high-frequency market information in catchment area of each road segment sample restricted to segments located close to an existing market
Data
household sample Sample frame: all villages within 1km of road segment 2 sampled villages per segment Class 1: close to a market (for identification) Class 2: very remote, i.e. those whose transport costs are expected to change most with the road rehabilitation 15 HHs randomly selected in each village
HH surveys necessary but not sufficient IE design Event study design requires higher frequency data than practical through HH surveys Scope of research questions interest in precisely measuring market-wide price changes, which will be difficult for any one household to report Sample size / budget practicalities Difficult to know ex-ante which households will benefit most; catchment areas are large and little pre-existing data impact for any one HH in segment catchment likely small
feeder roads: data ecosystem Multi-module household surveys focused on income, consumption and production Household Data Market structure and composition (annual); availability and price of a basket of goods (monthly) Market Data Maps of all road segments, markets and villages Geo data Population census: sampling frame and migration Population data Land Administration Information System: all land transactions registered and geo-referenced Land Data International road roughness index; traffic counts; travel time Road Data Exact start and end date of construction Project monitoring data
what makes this work? Collaboration with project team started long before road construction (2012) Coordinated monitoring data collection plan across donors so data can be easily merged Rwandan government collects a lot of administrative data, relatively organized Advantage of one national ID number, used for all government interactions
Early Results
market integration At baseline, market integration is poor Most market vendors report being professional traders (not producers), but the majority sell at only one market Product availability and price varies within small geographic areas
trader occupation
trader movement across markets
product availability This map shows a high availability good- tomatoes Source: 2017 Price market survey
product price variance This shows the variance in prices across the country. The markers are sized based on standard deviations away from the mean. The regression looks at the relationship between the price per kg of tomato and the distance to the national roads. Source: 2017 Price market survey
preliminary HH impacts First follow-up survey conducted one year after baseline Few segments actually completed; analysis focuses on initial short-term impacts Follow-up surveys to be conducted every 1-2 years over the project lifetime
preliminary results on HH income Suggestive evidence that investing in feeder roads allows relatively remote HHs to catch up to relatively more connected HHs Being located in a remote village decreases HH income by $73 Mean HH income is $316, so 23.1% decrease But feeder road rehabilitation increases HH income by $74 (23.4% increase) Results in a full catch-up to more connected villages Effect driven by income from HH farm & related activities
Thank you!