This post is extracted from an article I co-authored with Sandra Paola Alvarez, Olivier Ferrari and Ulrike Rieder in the last issue of the Migration Policy Practice, published by the International Organization for Migration (IOM). 

Information collected on remittance transfer costs during the last decade has shed light on the high costs incurred by migrants around the world when sending remittances, and has contributed to bringing this issue to the forefront of the international development scene. Nevertheless, the data currently available is not accurate and complete enough, neither to assess the true cost of remittances, nor to understand what drives cost fluctuations or monitor this rapidly evolving market.

Current costs estimations methodologies

At present, the most complete data set on remittance transfer costs available is developed and maintained by the World Bank (see Remittances Prices Worldwide). Updated four times a year since 2008, this data set provides information about the cost of sending money on 227 corridors worldwide.

The data from the Remittances Prices Worldwide group at the World Bank is collected solely through mystery shopping. Through this methodology, researchers – presenting themselves as customers – collect the pricing information manually from the money transfer service providers, either by making an actual transaction or by asking the cost of a transaction in person, over the phone, or through a web interface. Cost information is collected for each corridor and for two different sending amounts (the equivalent of 200 and 500 US dollars), from a range of money transfer operators and banks. Using this methodology, the World Bank collects around 20’000 data points each year, which are then used to calculate the global average cost of remittances; that is, the average of the average cost per corridor, weighted by the size of each corridor.


While the information on remittance transfer costs collected by the World Bank constitutes the most accurate global data set currently available, a number of limitations inherent to the data-collection methodology and the way indicators are constructed, deserve to be discussed.

First, because mystery shopping is a resource-intensive data-collection methodology, the scope of the data set must be targeted. This limits the number of corridors that can be monitored, the number of data-points collected on each corridor, and the frequency of data updating.

A snapshot that fails to describe the bigger picture.

Second, in markets where costs fluctuate significantly over time and where costs vary substantially depending on the amount transferred, data collection regarding the cost of sending two amounts (200 USD and 500 USD) every three months can only provide an approximation to real costs. As a matter of fact, operators will often have more than ten different pricing tiers between 10 USD and 5000 USD, with different fees for each tier and often different exchange rates. As a result, the data currently collected is a snapshot that fails to describe the bigger picture.

Finally, the Remittances Prices Group of the World Bank publishes a quarterly report to monitor the evaluation of remittances transfer costs, using the average cost per corridor as main indicator. The main limitation of the corridors’ averages is that they are not weighted by the number of migrants using each money transfer service provider for which data is collected. This means that in some cases, banks offering unfavorable exchange rates (at least for the amounts for which data is collected) but used by a few migrants, will skew the average upwards. Likewise, if a new money transfer operator offering low cost transfers enters a market, the average will drop even though only a small proportion of migrants uses this new service.

Ways forward to get better data

How then to obtain better data on remittance transfer costs? Firstly, we must admit that obtaining better data on remittance costs is extremely challenging. In order to accurately monitor the remittances market, we probably need to collect a hundred times more data-points than what we collect today. Automated or crowd-sourced data-collection systems are likely to enable the development of a more complete database. Whenever possible, integration with money transfer service providers through API or web scrapers should be developed to receive their prices in real-time. For offline agent-based money transfer service providers, proper incentives have to be developed to encourage clients to report the costs in a central data-base.

Once a more accurate global data set on money transfer costs is available (and that is what we plan to do at TawiPay with the Global Remittances Observatory), various indicators can be developed to monitor the evolution of the money transfer service offer. In order to assess the actual costs incurred by migrants, the development of more complex models will be necessary, including detailed information about migrants’ transfer habits (average amount, frequency, type of money transfer used, etc.), to calculate a weighted average cost of remittances for each corridor.


The full version of the article, including the first two sections written by my co-authors on the limitations of the current estimations of global and bilateral remittances flows can be found here