To evaluate the impact of the rehabilitation of the project roads on the socioeconomic development of the communities along the roads, the study will use a combination of propensity score matching and difference-in-differences regression. Communities along the roads scheduled for upgrades (the treatment communities) will be matched with other communities that have similar characteristics but will not benefit from the road upgrades (the comparison communities). This will result in two comparable groups of communities: one group that is situated along the roads and will receive the (benefits of) road upgrades and another group that will not. The study will obtain pre-treatment and post-treatment data on both groups of communities and compare values of key outcome variables, such as travel time, travel cost, and income.
Kind of data
Sample survey data [ssd]
Anonymized dataset for public distribution
Vehicle Operating Costs
The program includes road segements in four regions of Tanzania: Tanga, Rukwa, Mbeya, and Ruvuma. Three sets of road segments will be upgraded: (1) Tanga-Horohoro (in Tanga region, connects with Kenya border), (2) Tunduma-Sumbawanga (in Rukwa and Mbeya regions, connects with Zambia border), and (3) Namtumbo - Songea and Peramiho - Mbinga (in Ruvuma region, connect with Mtwara Corridor roads leading to the Eastern coast).
Unit of analysis
Households located in communities along the upgraded roads
Producers and sponsors
Economic Development Initiatives
Millennium Challenge Corporation
The survey has been administered in 200 communities in the four Regions. In Tanzania, vitongoji - subunits of villagesare good approximations of the concept of a community. Hence, 200 vitongji needed to be selected into the sample. We distributed these 200 vitongoji along the three road projects proportional to the square root of the length of each road project. This design assures representativeness along the shorter roads as these roads are assigned proportionally more vitongoji than the longer roads.
To construct the actual sample, we drew up an exhaustive list of villages along the treatment and comparison roads. Next, we randomly sampled the required number of villages per road project (24 for Tanga - Horohoro, 34 for Namtumbo - Songea - Mbinga, 42 for Sumbawanga - Tunduma), resulting in a sample of 100 villages. Since MCA?Tanzania expressed interest in stratification of the sample according to distance from the road (to examine whether the effects of road upgrades persist as one moves away from the road), we identified all vitongoji within each selected village (this was carried out by the listing team, which visited the selected villages ahead of the actual survey teams). Per village, the listing team divided the vitongoji in two groups depending on whether the distance from the kitongoji to the road is less than 30 minutes on foot (close stratum) or at least 30 minutes on foot (far stratum). We then randomly selected one kitongoji in each stratum. In this fashion, we arrived at a stratified sample of 200 communities (vitongoji), 100 in each stratum. For each road project, we distributed the villages and vitongoji evenly between the treatment and comparison roads (for instance, Tanga - Horohoro has 12 treatment villages and 12 comparison villages)
Dates of collection
Data collection supervision
Just like in Mainland Tanzania, the Listing Team consisted of two Research Supervisors (RS) working independently of each other. Each RS was equipped with a PC for on-field data entry using CWEST programmes and an internet modem for instant data transmission. The Listing Team visited the target Shehias/villages ahead of the Survey Teams. They left EDI HQ on 2nd June, 2009 and started work in Pemba Island on 5th June, 2009. The work of the Listing Team was different in the comparison than in the treatment Shehias: While the sample of treatment villages had already been established before the visit of the Listing team Consultancy Services for the Design and Implementation of Household Survey and Community Profile for the Transport Sector 97 (the Research Director, Survey Manager and a Consultant has sampled the treatment villages during an earlier visit to Pemba in November 2008), this was not the case for the comparison villages. Rather, the Research Director, Survey Manager and a Consultant had identified 14 appropriate comparison Shehias out of which 42 comparison villages had to be randomly selected. Hence, the first duty of the Listing Team was to visit to selected comparison Shehia, draw up exhaustive lists of villages within each Shehia and randomly sample three villages from the list. Once the villages were chosen, the Listing Team visited all 80 villages in the sample (42 comparison and 38 treatment villages). Within each village, they randomly sampled 20 households (15 ‘core’ households and 5 replacements in case a selected household would refuse or be unable to participate) to participate in to be included in the household surveys. The team also organized for 5 knowledgeable persons in each village to participate in the Community Profiles. The Listing Team recorded all information electronically and sent it to HQ and to the Research Supervisors of the survey teams. The Listing Team finished all work in Pemba on August 2nd 2009. 6.2 SURVEY TEAMS There were 5 Survey Teams for Pemba, each consisting of 4 Research Officers working under the supervision of a Research Supervisor. Each Research Officer had an Ultra Mobile Personal Computer (UMPC) for on-field data entry using CWEST programmes, and each supervisor had a PC and internet modem for instant data transmission. Every evening, the Research Supervisors downloaded gathered data from the interviewers’ UMPCs, checked the data and uploaded them on the EDI server. The Data Processing Team started processing the data as soon they had been uploaded. The Survey Teams left EDI HQ on June 17th 2009 and started conducting the first household interviews on June 27th. By August 10th, the Survey Teams had completed all 1200 household interviews in North Pemba. By August 21st they had also finished the Community Profiles.
Questionnaires 1. Household Survey: Household roster (education, literacy and potential illness of household members), Assets/Livestock, Housing materials, Public transport use, Economic activities 2. Community Profile: designed to collect information on characteristics that (are likely to) influence future socioeconomic trends and baseline information on outcome variables that are likely to be influenced by better roads 3. Ladder of Life Focus Group Discussion: designed to collect information on how a particular community defines welfare and to document the situation of the different participants on an ordinal welfare scale (the Ladder)
Economic Development Initiatives
De Weert, Joachim. 2009. Tanzania Mainland Trunk Roads Baseline, 2009. Economic Development Initiatives.
Version 1.1 (March 2014). This version uses the new MCC metadata template.
Version 2.0 (May 2015). Edited version based on Version 01 (DDI-MCC-TZA-MPR-ROADS-2009-v01.1) that was done by Millennium Challenge Corporation.