Estimating workers remittances using household surveys: Experience and Lessons from Uganda Presented by Kenneth Alpha Egesa BANK OF UGANDA Migration and Remittance Peer-Assisted Learning Network (MIRPAL) Conference May 31-June 3, 2010 Moscow, Russia
Presentation Plan Introduction to compiling worker s remittances in Uganda Some results of the 2008 survey Current methodology for estimating worker s remittances Conducting the survey of worker s remittances: Sample size determination Conducting the survey of worker s remittances: Sample selection Key questions for estimation of worker s remittances Key variables and estimation procedure for the 2006 estimates Key variables and estimation procedure for the 2008 estimates Challenges and lessons learned
Introduction to compiling worker s remittances in Uganda (1) Compilation for the period to 2006 was based on financial sector data Crude estimates were derived from unexplained foreign exchange purchases by foreign exchange dealers Increased foreign exchange transactions due to active financial markets and incomplete information on derivative trading resulted in high increases in workers remittances estimates There was no supporting information on the estimates obtained which was problematic when it came to explaining the estimates. A survey of workers remittances was conducted in 2006 to verify the efficiency of the estimation approach and to collecting supporting information
Amounts in Millions of US $ Introduction to compiling worker s remittances in Uganda (2) Workers remittances estimates 1000 900 800 700 600 500 400 300 200 100 0 2004/05 2005/06 2006/07 2007/08 Workers remittances (residual estimates) Workers remittances (survey estimates)
Introduction to compiling worker s remittances in Uganda (3) The 2006 survey of worker's remittances receipts was conducted as a project co-funded by Bank of Uganda (BOU), Deutshe Gesellschaft fur Technische Zusammenarbeit (GTZ) and Swedish International Development Agency (SIDA). Transmission of ownership to BOU involved budgeting for the 2008 survey and the setting up of a fully fledged section in the statistics department responsible for the survey by BOU. A committee comprised of Bank of Uganda, Uganda Bureau of Statistics (UBOS) and Economic Policy Research Centre (EPRC) staff handled survey design (questionnaire design, sample size, sample selection etc) and implementation (recruitment and training of enumerators, supervision etc). Data processing (editing, data entry and generation of initial tables) was done by UBOS and grossing up of survey estimates and report writing was done by BOU.
Introduction to compiling worker s remittances in Uganda (4) A parallel survey was also conducted on worker s remittances sender s through the main corridors (United States of America (USA), United Kingdom (UK), and Republic of South Africa (RSA) with main funding from the World Bank. The survey targeted sender s of remittances through existing associations and embassies in the sender s economies identified as leading source countries. The results were used to validate the findings of the remittances recipients survey on main source country, type and mode of transmission of remittances, value of cash remittances, use of remittances, age of senders, occupation and education level of senders, relationship with recipients etc). The results were used to attempt estimating the number of
Some results of the 2008 survey Total remittances received in 2008 were estimated at US$732.4 million 42.5% was received through International Money Transfer Operators (MTOs), Banks accounted for 25.1% and 21.7% was through informal channels 67.1% of recipient households received only cash remittances while 13.4% received only items in-kind, the rest received both. Cash Remittances were estimated at US$666 million or 91.0% of the total. 39.9% of remittances were from Europe, followed by Africa (26.4%) and North America (24.5%). Remittance recipient households were estimated at 0.436 million for cash and 0.136 million for remittances in kind. A total of 543,705 remitters was derived with the average
Current methodology for estimating worker s remittances (1) Two pronged approach is used comprised of financial sector data for preliminary estimates and survey data for final estimates Financial sector data estimates are derived from monthly reports of international money transfer agencies such as Western Union, MoneyGram etc Total estimates are then derived using previous survey estimates of shares of remittances through international money transfer agencies to total estimates Revisions are made when survey estimates become available (usually three months after the end of the year). The 2008 survey attempted to obtain the quarterly profile of remittances through progressive inquiries on remittances received per period.
Conducting the survey of worker s remittances: Sample size determination The 2006 sample size was determined at 4100 by setting the confidence interval of the sample estimate to 95%, using estimated sample proportion of 2% (p) for households receiving remittances and estimated standard error of 0.0025 (s.e(p)) from the 2005/06 NHS. The 2008 sample size was determined at 2040 by setting the confidence interval of the sample estimate to 95%, using estimated sample proportion of 4.2% (p) for households receiving remittances and estimated standard error of 0.025 (s.e(p)) from the 2007 survey. c Zc * s. e p n 2 2 1.96 p(1 p)/ c
Conducting the survey of worker s remittances: Sample selection A two-stage sample selection procedure was used where enumeration areas were selected as the first stage units and households were selected as the second stage units. Selection of the enumeration areas was by systematic sampling and households by probability proportional to size determined through a listing exercise. Proportionate distribution between rural and urban households was followed in the first survey but was urban biased in the second survey due to few recipients in rural areas owing to lack of financial services. The respondent is the household head although there cases where there is more than one recipient per household or where the recipient is not the household head.
Key questions for estimation of worker s remittances Cash remittances received in past month, past quarter, past six months and past year Remittance channel used Remittances in kind received Number of senders Source country of the remittances Profile of recipients Profile of senders Use of the funds
Key variables and estimation procedure for the 2006 estimates Estimated sample average of cash remittances received by households Ratio of households in the sample that received remittances Total number of households indicated in the NHS Total workers remittances = {average remittances per household * number of households receiving remittances} Number of households receiving remittances = {number of households that received remittances in the survey / total number of households in the remittances survey) * total number of households reported in the NHS Ratio of funds through money transfer operators
Key variables and estimation procedure for the 2008 estimates Estimated sample average of cash remittances received by households Estimated number of households that received cash remittances in 2008 Estimated sample average of value of in-kind remittances received by households Estimated number of households that received in-kind remittances in 2008 Estimated number of households that received remittances in 2008 was derived from the UBOS mid-year population growth rates applied to the estimated number of households that received remittances in 2006 Ratio of funds through money transfer operators
Challenges and lessons learned Estimation of remittances received in kind (type quality, quantity issues) The cost is high but quality data is required which calls for finding a balance between the two. Dealing with household with more than one recipient Estimation of remittances received in kind (type quality, quantity issues) Memory recall of cash amounts received Revision of previous estimates and methodology for preliminary estimates in absence of survey results Determining sample distribution between rural and urban households Determining the relative importance of different channels of transmission of remittances.
Thank you