Impact Evaluation of Feeder Roads: Phase 1, Baseline Findings Report Report submitted to: Government of Ghana Millennium Development Authority 4th Floor Heritage Tower, 6th Avenue Ridge West, Accra, Ghana Report submitted by: Raymond J. Struyk, Joseph G. Caldwell, John Felkner, Kareem Kysia, and Shova KC National Opinion Research Center 4350 East-West Highway Bethesda, MD 20814 Tel. (001)(301)634-9300 September 20, 2010 Executive Summary Under the Agricultural Project being implemented by Ghana’s Millennium Development Authority (MiDA) some feeder roads are to be rehabilitated or reconstructed to promote development in the sector. In the first phase, about 336 km of feeder roads in eight (8) districts in two intervention zones are to be rehabilitated to reduce transportation costs and time, and increase access to major domestic and international markets. The feeder roads activity will also facilitate transportation linkages from rural areas to social service networks (including hospitals, clinics and schools). The purpose of this project is to conduct an impact evaluation of the MiDA’s Feeder Roads Activity. As stated in the Terms of Reference of the request for proposals, “the primary data for the impact evaluation will be a series of surveys similar in scope to the Consumer Price Index (CPI) survey, examining changes in prices over time Findings from the market surveys will contribute to the overall impact evaluation conducted by the Institute of Statistical, Social and Economic Research (ISSER). The Ghana Living Standards Survey (GLSS) 5+ is the primary instrument used in the overall evaluation, and ‘Difference in Difference’ is the proposed method of evaluation of data.” Thus, this study focuses on how prices of goods sold at local markets (that are transported on improved roads) change over time. It is also to document the changes in goods transport tariffs and passenger fares to market places served by the feeder roads. This report presents a description of the evaluation design, the method employed in drawing the sample and sample description, the implementation of baseline survey, adjustments made to the data initially collected, and spatial patterns in the prices and transportation tariffs. The sample design uses a carefully tailored algorithm employed to match 154 localities that will benefit from the road improvements with an identical number of control localities that are comparatively far from the improvements. The sample size is sufficient to provide robust estimates of price effects associated with the road improvements. The minimum population for a locality to be included in the sample is 1,000, a condition imposed to help ensure that most designated items could be found in most localities. Beginning in August 2009 interviewers began visiting the sample localities to obtain three price observations for each item in the defined “basket” of goods and transportation services. The final “basket” contains 39 fresh food items, 24 packaged food items, 19 non food items and 6 transportation tariffs—3 for the locality’s residents’ most frequent passenger destinations and 3 for the most frequent freight destinations. About half of the price observations were not obtained because the good was not available in a locality or it was only sold by one or two sellers. There were a significant number of observations where the description of the item priced differed from that specified on the pricing sheet (1 cigarette was priced rather than a pack that had been specified). In cases where only invalid data were gathered in a locality for an item, imputation procedures were employed to generate one valid observation for the place. Chapter 4 presents item-by-item price and tariff information for control and treatment localities and summary data on the extent of price variance across localities. Prices demonstrate varying Phase 1, Baseline Report Impact Evaluation of Feeder Roads i degrees of variation. Because this is the baseline report it is not possible to compute price changes associated with the road improvements. To what degree can price variations be associated with simple measures such as the size of the locality, measured by population, or the distance from the next large community? Simple regression models are estimated for aggregate prices of goods’ groups: fresh foods, packaged foods, and other goods. (They cannot be estimated for transport tariffs because the destinations differ for each locality.) The models show that accessibility matters least for fresh goods, since a good share of these are often available locally. But accessibility does have a statistically significant effect for packaged foods and non food goods. The quantitative effects of greater distance and travel times are, however, quite modest. Overall, the models substantially confirm the kind of systematic price variation one would anticipate. This in turn lends confidence to the quality of the data collected. Phase 1, Baseline Report ii Impact Evaluation of Feeder Roads Acknowledgements The work reported here has benefitted substantially from our interaction with MiDA’s Office of Monitoring and Evaluation. In particular, Mrs. Abigail Abandoh-Sam and Patrick Fosu-Siaw provided consistent advice and guidance during the execution of this project phase. They were instrumental in the team acquiring data needed for matching treatment and control localities and the Office’s review of field operations brought important matters to our attention. Dr. Emmanuel Amamoo-Otchere worked tirelessly in helping the NORC team acquire certain data used in matching treatment and control localities and was generally a valuable source on Ghanaian data sets. Pentax Management Consultancy Services carried out the survey whose results are reported here. Dr. Mohammed Muslim was particularly helpful in managing the relations between Pentax and NORC. Phase 1, Baseline Report Impact Evaluation of Feeder Roads iii Phase 1, Baseline Report iv Impact Evaluation of Feeder Roads Table of Contents Executive Summary .................................................................................................................. i Acknowledgements ................................................................................................................ iii Table of Contents ..................................................................................................................... v List of Tables .......................................................................................................................... vii 1. Introduction and Summary ............................................................................................ 3 2. Approach, Sample Design, and Sample Selection ........................................................ 4 2.1 Method ..................................................................................................................... 4 2.2 Sample Frame, Sampling Units and Sample Sizes .................................................. 6 2.3 Sample Design Variables and GIS Data Preparation and Processing .................... 11 2.4 Sample Construction .............................................................................................. 15 2.5 Description of Sample ............................................................................................ 17 3. Conducting the Survey .................................................................................................. 27 3.1 Methodology .......................................................................................................... 27 3.2 Preparation ............................................................................................................. 29 3.3 Training .................................................................................................................. 31 3.4 Field Period ............................................................................................................ 34 3.5 Challenges & Adjustments .................................................................................... 36 3.6 Summary of Data Set Quality ................................................................................ 40 3.7 Double Difference Analysis ................................................................................... 45 4. Findings .......................................................................................................................... 47 4.1 Variations in Observed Prices and Tariffs ............................................................. 47 4.2 Price Variation with Local Conditions .................................................................. 55 Annex A: GIS Systems and GIS Accessibility .................................................................... 59 References ....................................................................................................................... 68 Annex B: Survey Instruments.............................................................................................. 69 Annex C: Selected Results of Regression Models for Imputing Values for Invalid Observations ......................................................................................................... 93 Phase 1, Baseline Report Impact Evaluation of Feeder Roads v Phase 1, Baseline Report vi Impact Evaluation of Feeder Roads List of Tables Table 2.1 List of Variables Derived from GIS Variables Used for Sample Delineation ............................ 13 Table 2.2 Coding of