How Do Transaction Costs Influence Remittances?
Author:
Mr. Kangni R Kpodar
Search for other papers by Mr. Kangni R Kpodar in
Current site
Google Scholar
PubMedClose
and
Patrick A. Imam
Search for other papers by Patrick A. Imam in
Current site
Google Scholar
PubMedClose
Publication Date:
04 Nov 2022
eISBN:
9798400224270
ISBN:
9798400224270
Language:
English
Keywords:
cost elasticity; remittance flow; remittance provider; remittance service; cost-mitigation policy
Download PDF (761.1 KB)

Using a new quarterly panel database on remittances (71 countries over the period 2011Q1- 2020Q4), this paper investigates the elasticity of remittances to transaction costs in a high frequency and dynamic setting. It adds to the literature by systematically exploring the heterogeneity in the cost-elasticity of remittances along several country characteristics. The findings suggest that cost reductions have a short-term positive impact on remittances, that dissipates beyond one quarter. According to our estimates, reducing transaction costs to the Sustainable Development Goal target of 3 percent could generate an additional US$32bn in remittances, higher that the direct cost savings from lower transaction costs, thus suggesting an absolute elasticity greater than one. Among remittance cost-mitigation factors, higher competition in the remittance market, a deeper financial sector, and adequate correspondent banking relationships are associated with a lower elasticity of remittance to transaction costs. Similarly, remittance cost-adaptation factors such as enhanced transparency in remittance costs, improved financial literary and higher ICT development coincide with remittances being less sensitive to transaction costs. Supplementing the panel analysis, the use of micro data from the USA-Mexico corridor confirm that migrants facing higher transaction costs tend to remit less, and that this effect is less pronounced for skilled migrants and those that have access to a bank account.

Abstract

Using a new quarterly panel database on remittances (71 countries over the period 2011Q1- 2020Q4), this paper investigates the elasticity of remittances to transaction costs in a high frequency and dynamic setting. It adds to the literature by systematically exploring the heterogeneity in the cost-elasticity of remittances along several country characteristics. The findings suggest that cost reductions have a short-term positive impact on remittances, that dissipates beyond one quarter. According to our estimates, reducing transaction costs to the Sustainable Development Goal target of 3 percent could generate an additional US$32bn in remittances, higher that the direct cost savings from lower transaction costs, thus suggesting an absolute elasticity greater than one. Among remittance cost-mitigation factors, higher competition in the remittance market, a deeper financial sector, and adequate correspondent banking relationships are associated with a lower elasticity of remittance to transaction costs. Similarly, remittance cost-adaptation factors such as enhanced transparency in remittance costs, improved financial literary and higher ICT development coincide with remittances being less sensitive to transaction costs. Supplementing the panel analysis, the use of micro data from the USA-Mexico corridor confirm that migrants facing higher transaction costs tend to remit less, and that this effect is less pronounced for skilled migrants and those that have access to a bank account.

I. Introduction

Remittances matter for development. They are a key source of funding for developing countries. Migrants send part of their earnings to family members to provide them with basic subsistence, build and invest in the economy back home. According to World Bank data, remittances to low- and middle-income countries more than doubled during the past 15 years to reach US$550 billion in 2021. Over half of it goes to people in rural areas, and about 75 percent is used to cover basics such as food and medical or school expenses, while the remaining is invested in assets or saved (IFAD, 2021).

Being a major source of income for the poorest households, policy makers in advanced and developing countries are looking at avenues to encourage remittance flows to grow even further. However, a major stumbling block in sending money home is the high, some would argue excessive, transaction costs involved. Despite the commitments from the international community to reduce the cost of remittances, and the inclusion of the 3 percent cost target in the Sustainable Development Goals (SDGs), progress has been slow in recent years.1 The strong resilience of remittances during the Covid-19 pandemic has brought back to the forefront the debate surrounding the stubbornly high transaction costs (see Kpodar et al., 2021).

It is accepted wisdom that individuals sending money home in many parts of the world, particularly sub-Saharan Africa, are paying a very high cost, which may dissuade further flows. Considering fees, exchange rate margin and other costs, between 5 and 15 percent of remittances are “lost” due to high transaction costs, depending on the country and the amounts send home (Ratha, 2021). Because formal remittances involve high fixed costs and hence are expensive to provide, low-income individuals refrain from remitting, or are incentivized to use cheaper informal alternatives (Gibson, McKenzie and Rohorua, 2006; Yang, 2011). Against this backdrop, this paper attempts to address the following questions: how elastic are remittances to changes in transaction costs? what are the factors or policy interventions that may help explain cross-country differences in the cost elasticity of remittances?

This paper is related to several strands of the literature. First, a large body of work looks at the impact of remittances at the macroeconomic and microeconomic level. Many studies find that remittances stimulate growth, although no broad consensus exists yet on that matter (see Cazachevicia, Havraneka, and Horvath, 2020, for a recent discussion). This favorable impact builds on the premise that remittances are an important source of development financing, a significant source of international reserves, and often arrive countercyclically. At the micro-level, remittances respond to income shortfalls and, in that way, have the potential to smooth households’ income, thereby reducing risks to the family (Choi and Yang, 2007). Remittances are also linked to improved economic, health and education outcomes as well as higher financial inclusion. However, there are also potentially some deleterious effects observable at the macro-level (Chami, Fullenkamp, and Jahjah, 2003). Besides producing Dutch Disease-type effects, potentially reducing the quality of institutions, or delaying needed fiscal adjustment, remittances may also complicate the monetary transmission mechanism, making monetary policy harder. There is also evidence that remittances may breed dependency by discouraging receiving household members from working, possibly trapping countries into anemic long-run growth (see Amuedo-Dorantes, 2014, for a summary).

Second, a narrow set of studies investigate the determinants of remittance transaction costs. For instance, Freund and Spatafora (2008) use cross-country data for 66 countries and find that recorded remittances depend positively on stocks of migrants and negatively on transaction costs and exchange rate restrictions. Transaction costs are also lower when financial systems are more developed, and exchange rates are less volatile. Similarly, Beck and Martinez Peria (2011) conclude, based on a dataset covering 119 country corridors, that the size of the migrant population, banking competition and lower barriers to access banking services are conducive to lower remittance costs. More recently, Beck, Janfils, and Kpodar (2022) exploit a richer remittance cost dataset to show that cost- and risk-based constraints as well as market structure hinder affordable remittance transaction costs. Corridor and firm-specific analysis reveals that higher income per capita, easier geographical access to financial institutions, larger remittance market, shorter distance between sending and receiving countries, competition in the remittance market, and pegged exchange rate regime are associated with lower transaction costs. In the same vein, da Silva Filho (2021) underscores that the drivers of remittance costs are multiple and complex, with regulatory issues (such as exclusivity clauses, stringent AML/CFT regulation, and restrictive licensing), a lack of price transparency, higher number of banks and a thin remittance market being likely to have a detrimental effect on remittance costs.

Third, there are several selected studies focusing on the cost elasticity of remittances. These range from cross-country and panel studies (e.g.; Freund and Spatafora, 2008, Ahmed, Mughal, and Martínez-Zarzoso, 2021) to country-specific studies (Ferriani and Oddo, 2019, on migrants in Italy; Kosse and Vermeulen, 2014, on migrants in the Netherlands; Kakhkharova, Akimovb, and Rohdeb (2017) on migrants in Russia; Ahmed and Martínez-Zarzoso, 2016, on migrants from Pakistan; and Gibson, McKenzie, and Rohorua (2006) on migrants from Tonga). Overall, these studies document a significant and negative effect of transaction costs on remittance inflows, partly due to high transaction costs encouraging informal remittances. Nevertheless, these studies do not go beyond the cost elasticity estimate to investigate how this varies across countries with different characteristics.

In this paper, we improve on existing studies along several dimensions. First, we exploit a new quarterly database on remittances, allowing us to investigate the elasticity of remittances to cost with high frequency data in a dynamic setting using local projections (Jorda, 2005). As such, this paper sheds light on the short- and medium-term impact of a shock to transaction costs on remittances and how persistent this effect might be. Second, unlike previous studies, this paper undertakes a systematic analysis of factors that can shape the cost elasticity of remittances. It is important to note that we are not taking a view on whether having a low-cost elasticity to remittances is good—a low elasticity implies less reduction in remittance flows when transaction costs increase, but also lower increase in remittances occurs when transaction costs are reduced. But the reason to focus on elasticities is that a lower elasticity will imply that fewer remittances flow informally. Two broad categories of factors/policies are considered: (i) cost-mitigation policies that directly tackle the root causes of high remittance costs (e.g. policies to promote competition in the remittance market, foster financial sector development, and safeguard correspondent banking relationships) and (ii) cost-adaptation policies that do not primarily address the cost of remittances, but promote more efficient remittance choices (e.g. policies to enhance price transparency, financial literary and information and communication technologies (ICT) development). Third, this paper uses micro data from the USA-Mexico corridor to ascertain the cost elasticity of remittances and its heterogeneity with respect to financial literacy, thus providing more granular results to support the evidence found at the macro level. Studying the USA-Mexico corridor is interesting from several standpoints. It is the largest corridor in the word; has several remittance service providers with a large array of products; and has in-depth publicly available data.

With a sample covering 71 countries over the period 2011Q1-2020Q4, the findings suggest that a 10 percent reduction in transaction costs leads to a 0.9 percent increase in remittances in the first quarter after the shock, with the impact becoming statistically insignificant from zero in subsequent quarters, implying that the response of remittances to transaction costs is essentially of a short-term nature. This result suggests that moving from the 2020 level of transaction costs (6.3 percent) to the SDG target of 3 percent will generate an additional US$32 billion in remittances, much larger than the direct cost savings. Therefore, migrants would not only fully pass on the cost savings to their families, but also send more, implying an absolute elasticity higher than one. The results are robust to an instrumental variable approach.

In investigating the cost-mitigation and adaptation factors, this paper finds that in countries where competition in the remittance market is high, the financial system is developed, and ties with correspondent banks holds up, the elasticity of remittance to transaction costs is much lower than otherwise. This indicates that remittances are less sensitive to transaction costs where alternative informal channels to the repatriation of money exist. Likewise, less opaque remittance transaction costs, improved financial literary and higher ICT development (along its multiple dimensions: ICT use, access, capability and affordability) help explain why some countries may have a lower cost-elasticity of remittances. Finally, evidence from the USA-Mexico corridor, using an annual survey of the Bank of Mexico during 2013-2017 and covering over 37,000 individuals, confirms that migrants who face higher transaction costs tend to remit less, even after controlling for socio-economic characteristics. More importantly, education level or access to a bank account, a proxy of financial literacy, mitigate the cost-elasticity of remittances, consistent with the findings from the panel data.

The paper is structured as follows. Section II presents data and some stylized facts on remittances and transaction costs and lays out the empirical model and methodology employed. Section III discusses the key results of the paper, with additional findings relegated to the annexes, while Section IV concludes with some policy recommendations.

II. Data and Empirical Strategy

A. Cost of Remittances and Remittance Flow Data

The key source of remittance cost data is the “Remittance Prices Worldwide” dataset compiled by the World Bank. This database provides information for 365 corridors (consisting of a combination of 48 sending and 105 receiving countries) and by remittance service provider regarding the fee paid by the sender for sending the equivalents of US$200 and US$500 in local currency. It also contains the exchange rate applied to the transaction, if available, the type of the remittance service provider (e.g., a Bank or Money Transfer Operator-MTO), the time it takes for the money to become available to the receiver, the payment instrument that can be used by the sender and the form of the payment to the receiver (e.g. cash to cash) as well as the access point for the remittance service. The data are available on a quarterly basis from 2011Q1 onwards. In the analysis below, the remittance cost as a share of the amount transferred averaged at the country level is the main variable of interest.2

In addition, the World Bank also compiles a database on remittance inflows covering 215 countries and territories going back to the 1980s. This database is, however, carried out on an annual basis, and therefore does not allow to exploit the quarterly variations in the cost of remittances. Moreover, one would expect remittance flows to respond sharply to short-term variations in transaction costs, which annual data can fail to capture or may “over smooth". To address this issue, we construct, as discussed below, a new and unique dataset of quarterly remittance flows for a sample of 95 countries, consisting of 18 high-income countries, 62 middle-income countries and 15 low-income countries. The data run from 1971Q1 for a handful of countries through 2020Q4 for most countries.3

The new quarterly remittance database builds on the monthly dataset from Kpodar et al. (2021). The approach used in this paper is similar, with the main sources being the detailed balance of payments (BOP) and statistical notes published by national central banks and statistical institutes. For some countries, data are reported in local currency or a different currency than the US dollar, in which case, we use the quarterly average exchange rate from the IMF’s International Financial Statistics (IFS) database or relevant central banks to convert the remittance flows into US dollars. The compilation of the remittance data follows an internationally accepted definition whereby remittances are the sum of personal transfers and compensation of employees.4 For countries that do not report personal transfers in their BOP, data on workers’ remittances have been used instead as a proxy. When data compensation of employers is missing, they are not accounted for given that these flows are typically marginal compared to the size of personal transfers.

The combination of the quarterly remittances flows and costs data results in a sample of 71 countries, covering a period from 2011Q1 to 2020Q4 (Annex 1 provides the sample composition). The other variables in the model will be discussed in the subsequent sections as they are brought in the model specification.

B. Stylized Facts on Remittance Costs

Figure 1 illustrates the average fee as a percent of a remittance of US$200 by country during the period 2011Q1 to 2020Q4. The costs for migrants to send money across borders are extremely expensive in southern Africa (e.g. Angola, Botswana and Namibia), where they reached more than 2.5 times the sample average (7.5 percent), often due to high exchange rate margins. Small islands also exhibit high remittance costs, probably reflecting lack of economy of scale and limited integration into the international financial system. At the other end of the spectrum, the lowest costs are prevalent in eastern Europe. In between, a wide range of countries at different levels of development exist, with different corridor-specific characteristics.

Figure 1.
Figure 1.

Average fee as a share of a $200 USD remittance (2011Q1-2020Q4)

(percent)

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Note: The red line indicates the sample averageSource: Remittance Prices Worldwide (World Bank)

While remaining high, remittance costs have declined on average by about 5 percent during a five-year period (2016Q1-2020Q1). But this picture masks significant heterogeneity across countries (Figure 2). Even though more countries have recorded a decrease than an increase in fees, the magnitudes are striking. Remittance costs increased by more than 40 percent in the Gambia, Afghanistan, and the Kyrgyz Republic, albeit from a lower starting point for the latter.5 The large increase in remittances costs in these countries was mainly driven by higher exchange rate margins on the back of volatile and depreciating currency. On the other hand, many countries such as Latvia managed to achieve a significant reduction in remittance costs, probably due to enhanced competition and the rise of digital remittances. Nonetheless, the decline in costs observed in Lesotho and Eswatini was temporary, as it reflected the reduction in fees implemented by South African banks as a way to provide relief to customers early in the pandemic.

Figure 2.
Figure 2.

Change in Average fee for a $200 USD remittance (2016Q1-2020Q1)

(percent)

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Note: The red line indicates the sample average. The chart excludes Syria and Amenia that recorded an unusually large increase in transaction costs.Source: Remittance Prices Worldwide (World Bank) and authors’ calculations.

Looking at the developments in remittance cost over a decade (2011-20), the 5 percent decline in the second half of the period was modest relative to the large reduction achieved in the first half of the period (Figure 3). While further analysis is required to uncover the factors behind this large decline, including the role of competition, it seems to have coincided with the narrowing of the interest spread, a proxy of financial intermediation cost (associated with the traditional borrowing and lending operations of banks). This may not be surprising as remittance operators, including Fintech companies, often rely on the traditional cross-border payments infrastructure. Figure 3 also shows that the remittance cost has been consistently higher than the interest rate spread, with the gap having narrowed only marginally over a decade.

Figure 3.
Figure 3.

Trends in Average fee for a $200 USD remittance and Interest Rate Spread (2011-20)

(percent)

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Sources: Remittance Prices Worldwide (World Bank), and World Development Indicators.

There is also some heterogeneity in remittance costs regarding the type of remittance providers. Remittance fees charged by traditional banks tend to be more expensive than MTO’s and that of the post office (Figure 4), reflecting higher regulatory costs faced by banks and remittance services not being the most important product of the larger package of services that their clients receive (Beck, Janfils and Kpodar, 2022).

A clear negative correlation is noticeable between quarterly remittance costs and remittance flows (Figure 5). Countries with lower remittance costs tend to have higher remittance flows, and conversely higher remittance costs weigh on migrant transfers. While there are certainly other factors affecting remittances beyond costs, it is still not clear whether high remittance cost result themselves from the smaller remittance market, though Figure 5 offers suggestive evidence that costs matter for remittances. The econometric section will help provide a more rigorous answer.

Figure 4.
Figure 4.

Average fee for a $200 USD remittance by Type of Provider (2011Q1-2020Q4)

(percent)

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Sources: Remittance Prices Worldwide (World Bank) and authors’ calculations.
Figure 5.
Figure 5.

Remittances Flows and Average fee for a $200 USD remittance (2011Q1-2020Q4)

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Sources: Remittance Prices Worldwide (World Bank) and authors’ calculations.

C. The Model and Econometric Approach

To analyze the elasticity of remittances flows to transaction costs, the paper relies on the following model estimated on a sample of 71 countries with data over the period 2011Q1-2020Q4:

ln (Rem)c,t+h=∑i=1nαiln (Rem)c,t−i+∑i=1nβiln (Cost)c,t−i+∑j=0hδjln (Cost)c,t+j+θhXc,t+vt+uc+εc,t+h

         for h=0, …, H         Eq(1)

where:

  • Rem is the amount of remittances expressed in millions of US dollars received by a country c in quarter t

  • Cost stands for the fees per a US$ 200-dollar remittance expressed as a share of that amount

  • X is a set of control variables which include the income per capita of the remittance-receiving country and that of the remittance-sending countries to capture the level of development, the US dollar/local currency exchange rate, and the number of migrants originating from the remittance-receiving countries.

  • v is the time dummy for each quarter, u is the country-specific effect, and e is the error term robust to heteroscedasticity, autocorrelation and cross-sectional dependence (Driscoll and Kraay, 1998)

  • the number of lags, n, is set at 4 considering the quarterly frequency of the remittance data, and the forecast horizon, h, is set at 5 months.

Due to the lack of quarterly GDP data for many countries, income per capita is at annual frequency. Similarly, for the stock of migrants, we have only two data points, one for the period 2011-15 and the other for the period 2016-2020. The size of the population which is available annually is also used as an alternative indicator since it captures well the relative size of the remittance market (from the recipient side). Similar to remittance flows and the transaction costs, the US dollar/local currency exchange rate is available on a quarterly basis. For a given remittance-receiving country, the income per capita of the remittance-sending country is calculated as the weighted average of the income per capita of all countries hosting migrants from that country, with the weight being the host country’s share in total migrants. Annex 2 provides the summary statistics and correlation matrix.

As we are interested in the dynamics of remittance flows in reaction to a change in transfer costs, the local projection (LP) approach developed by Jordà (2005) is an appropriate estimator. The LP allows to assess the impact of a shock at time t on the dependent variable by generating multi-step predictions using direct forecasting models that are re-estimated for each forecast horizon. The LP is shown to robustly handle highly persistent data and the estimation of impulse responses functions (IRFs) at long horizons (Olea and Plagborg-Møller, 2021), and easily accommodate non-linearities (Auerbach and Gorodnichenko, 2013).

Further, Jordà (2005) argues that the LP is robust to misspecification of the lag structure as the impulse responses can be defined without any reference to the unknown data generating process. As such, the LP improves on the conventional Vector Autoregressive models (VARs), which require imposing sufficient identifying restrictions to derive the IRFs. If the VAR specification turns out to be non-representative of the data generating process, this can lead to a bias in the estimation of and inference from the IRFs. Nevertheless, Barnichon and Brownlees (2019) point out that the flexibility of the LP relative to VARs comes at the cost of efficiency. In contrast, Plagborg-Møller and Wolf (2021) note that from an identification and estimation standpoint, the LP is equivalent to VARs. Overall, the ease of the implementation of the LP and the intuitive design makes it a preferred approach for this paper. More importantly, the ability to handle nonlinearities is key for this study as we aim to investigate the factors that affect the elasticity of remittances to transaction costs.

One adjustment, however, to the LP specification is the addition of the correction factor suggested by Teulings and Zubanov (2014). When there are subsequent shocks, which is likely, the derived impulse response function also captures the treatment effect given the usual path of subsequent shocks (and not only that of the contemporaneous shock) and the usual behavior of other variables. Teulings and Zubanov (2014) point out that this could lead to a bias in the results, and therefore the LP specification estimated at horizon h needs to be expanded to control for shocks occurring between t+1 and t+h (embedded in the fourth term of Eq. 1). In doing so, the effect of potential subsequent shocks (change in transaction costs) is sterilized, thus allowing to isolate properly a change in transaction costs at time t on remittance flows a time t and over subsequent horizons.

Before discussing the IRFs obtained from the LP, we also present a static version of the model estimated with a fixed-effect estimator (without considering the lagged variables and horizons). The idea is to start with a simplified and straightforward version of the model before incorporating the more complex dynamic that the LP deals with, allowing to show the robustness of results. The fixed-effect estimator will also be used for testing the elasticity of remittances to transaction costs with the micro data from the US-Mexico corridor.

III. The Results

A. How Elastic Are Remittances to Transaction Costs?

Table 1 shows the results of the static model using the fixed-effect estimator. The findings suggest a clear, and negative association, between transaction costs and remittances flows with the coefficient of the transaction costs being statistically significant at the 1 percent level in all specifications. The magnitude of the marginal impact appears sizeable. For instance, a 10 percent decrease in transaction costs could stimulate remittance by about 2 percent (column 1 and 2). This elasticity is halved once we control for the exchange rate (column 3), most likely reflecting the fact that exchange rate is also an important cost factor driving the amount migrants decide to send home. Using the transaction cost of a US$500 remittance confirms the result obtained with the transaction cost of a US$200 remittance. The results also hold using the ratio of remittances to GDP or the ratio of remittances to the size of the country’s population.6

Table 1.

Transaction Costs and Remittances: Fixed-Effect Estimates

article image
Notes. Fixed effect estimations. Time dummies included. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. USD exchange rate denotes the units of local currency per USD.

As for the control variables, income per capita in the receiving and sending country is positively associated with remittance. One could expect that as income per capita rise in the receiving country, remittances might start to decline beyond a threshold. While the squared coefficient of income per capita of the receiving country enters the regression with a negative sign, this coefficient is only marginally statistically significant at the 11 percent confidence level. The implied threshold is estimated at US$46,000, beyond which remittances decline with increasing income per capita. None of the countries in our sample is close to that level (the maximum income per capita is about US$25,000), which may explain why the coefficient of the square term is barely significant.

The coefficient on the exchange rate is positive and significant at 1 percent in all regressions, suggesting that remittances in US dollar terms increase with a depreciation of the local currency. Mandelman and Vilán (2020) observe this intertemporal substitution in the case of Mexico as a stronger dollar provides immigrants with additional incentives to send more resources back home. This result also makes sense from the cost perspective because a depreciation of the currency of the recipient country means that the cost of acquiring 1 unit of the local currency has dropped, hence the demand for the local currency by migrants increases, ceteris paribus. We also observe in the regressions that, as expected, the stock of migrants (the supply side of remittances) is positively correlated with remittances, and this result holds when the population of the recipient country (the demand side of remittances) is used as an alternative indicator.

Now turning to the full-fledged dynamic model, the IRF obtained from the LP corroborates the findings from the fixed effect estimations. As shown in Figure 6, a 10 percent reduction in transaction costs result in a 0.87 percent rise in remittances in the first quarter (Annex table 1 presents the underlying regressions), close to the estimate in the fixed-effect regressions. The short-term response of remittances to transaction costs is predictable and consistent with the classic price-demand curve, whereby the demand for remittance services declines as the price increases. The impact of the shock in the subsequent quarters is not significantly different from zero, indicating that the effect is not persistent over time. We also use the median cost of a US$200 remittance, and the result is similar. Further, using the average and median cost of a US$500 remittance does not alter the finding above.7

Figure 6.
Figure 6.

Cost-Elasticity of Remittances: Local Projections

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

The estimated elasticity is economically meaningful. Taking the size of remittances to low and middle-income countries in 2020, amounting to US$705.5 billion, and the average transaction cost which stood at 6.3 percent in 2020, and bringing transaction costs to the UN target of 3 percent would results in an additional US$32 billion income for households in the receiving countries.8 This is US$8.9 billion more than the saving in costs for the migrants, suggesting that migrants would return to their families much more than the savings from lower remittance costs. In order words, estimates of remittances lost using the difference between the actual cost and the UN target (implicitly assuming that a US$1 reduction in cost translates to a US$1 dollar increase in remittances) significantly underestimate the true burden of high transaction costs on remittances (one way to think about the “true burden” is the money lost during the transfer process – the difference in the amount of remittances beyond what is redistribution within the migrant family). This result points to a transaction cost reduction as a very powerful tool to boost remittances.

To ensure that our result is not driven by a simultaneity bias between the size of remittance and transaction costs, the IRF is rerun with an instrumental variable (IV) approach. Indeed, a small remittance market may be associated with higher transaction costs, potentially due to the limited number of remittance service providers (and therefore less competition), a lack of economy of scale and the fee structure whereby small amounts are subject to higher transaction costs. We use the share of MTOs in the remittance service market as an instrument of transaction cost as this variable is unlikely to be correlated with the size of remittances. As shown in Figure 4, banks tend to charge higher remittance fees than MTOs, and therefore where MTOs are more prevalent, remittance fees should be lower. This is supported by the first-stage regression (see Annex Table 2) where transaction costs are regressed on the share of MTOs in remittance service providers and the other explanatory variables of the remittance model. The coefficient on the share of MTOs is negative and highly significant. This result is also consistent with Beck, Janfils and Kpodar (2022) who find a positive and significant relationship between the share of banks among market participants in a corridor and remittance costs. The IRF-IV confirms the previous findings that higher transaction costs undermine remittance flows (see Annex Figure 1).

B. Explaining Heterogeneities in the Cost Elasticity of Remittances

To test country characteristics that could explain why the cost elasticity could vary from one country to another, we introduce in Eq.(1) an interaction term between transaction costs and the indicator of the policy of interest to see how the resulting marginal impact of the transaction costs change. The policy indicator is also added in additive term to Eq.(1) to be able to properly identify the coefficient of the interaction term.

Although the LP is well suited for non-linearities, representing in a two-dimensional chart, the IRF between a variable x and another variable y conditional to a third variable z requires an additional step. A common approach is to use a dummy variable taking 1 for a low regime of the variable z and 0 otherwise (for instance below and above the average of the variable), and then plot the IRFs for the two regimes. Alternatively, some studies use a smooth transition function between the two regimes (see Auerbach and Gorodichencko, 2013; Furceri, Loungani, and Zdzienicka, 2018), instead of a dummy variable which assumes an abrupt shift. In this paper, we propose a third approach that exploits the variation in the conditional variable z without any transformation or assumption on the smoothness of the transition function. This consists in introducing the interaction terms between the variable x and z, and plot the IRFs for the 10th and 90th percentile of the distribution of z. Since the marginal impact of x on y is a linear function of z, the IRF for any values of z between its 10th and 90th percentile should lie between the IRF of this lower and upper bounds.

As discussed above, we consider two broad categories of policies.9 First, the cost-mitigation policies that directly tackle the root causes of high remittance costs. These include enhancing competition in the remittances market, deepening financial development, and addressing correspondent banking relationship issues as discussed below. We start with the premise that the elasticity of remittances to transaction costs would be higher in high-cost countries than in low-cost countries, meaning that the elasticity itself is a function of the level of transaction cost.10 This is, because: (i) as average transaction costs decline, the information search cost for the migrant may become higher than the expected savings from a more competitive remittance provider, and as the incentives to search for lower transaction costs decline, remittances would be less sensitive to these transaction costs; (ii) when transaction costs are low, other factors such as the speed of the transaction and the convenience of access to the funds would matter more for migrants and their families, thus outweighing potential effects of a marginal change in transaction costs on remittances. Annex Figure 2 illustrates that the data supports this hypothesis. The direct implication is that by lowering transaction costs, cost-mitigation policies would also likely reduce the elasticity of remittances to transaction costs.

Second, cost-adaptation policies include price transparency, financial literary and lowering information cost through information and communication technologies (ICT). The primary purpose of these policies discussed below, is not necessarily to address the issue of high costs, but instead to reduce information asymmetries in the remittance markets. As such, cost adaptation policies seek to ensure that consumers have all the information available as well as the skills required to make an informed decision on the remittance services they opt for. Ultimately, this could result in a reduction in transaction cost, but the link is not direct. Nevertheless, the distinction between cost-mitigation and cost-adaptation policies is not always clear cut. For instance, ICT development has the unique characteristic of also being a cost-mitigation factor as it enables cost-effective remittance services, through digitalization. In the following paragraphs we will discuss the findings of the IRFs for these different policies,

Cost-mitigation policies

• Competition in the remittance market

The degree of competition is a key determining factor of costs in the remittance market (Beck and Martinez Peria, 2011; Mbiti and Weil, 2011; da Silva Filho, 2021; Beck, Janfils and Kpodar, 2022). In a less competitive market, a remittance service provider may have incentives to mark up prices by taking advantage of its market power. In contrast, where there is a competition, remitters will tend to switch to cheaper remittance service providers, bringing the average cost down.

To test the effect of competition on the elasticity of remittances with respect to transaction costs, we use the average number of remittance service providers as a proxy for competition and interact it with the transaction cost. The IRFs presented in Figure 7 shows that in countries with a small number of market players (remittance service providers, RSPs), remittances react more to transaction costs than in countries with a large number of market players, confirming our conjecture.11 In the first quarter, the elasticity estimated at the 10th percentile of the distribution of the number of remittance providers is -0.1, which declines to -0.3 and becomes insignificant as the number of remittance providers increases to the 90th percentile of the distribution. In other words, in a less competitive remittance market, a rise in remittance costs may result discourage remittances (potentially boosting informal flows), while in a more competitive remittance market, there is little such effect.

Figure 7.
Figure 7.

Cost-Elasticity of Remittances with Respect to Competition in the Remittance Market

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. RSPs denote remittance service providers.Source: Authors’ calculations.

• Financial sector development

The main function of the financial sector, which is dominated by banks in our sample, is to mitigate market frictions associated with information asymmetries and transaction costs. This includes facilitating the trading of risk, mobilizing savings, and allocating capital to the best uses, monitoring managers, and easing the trading of goods, services and financial contracts (Levine, 1997). As the financial sector develops, it is expected to perform these functions more efficiently, taking advantage of economies of scale that allow to provide financial intermediation at a lower cost. With banks involved in the provision of retail remittance services and facilitating cross-border transactions, financial deepening can lead to a decrease in the cost of remittances, and thus translate into a lower cost elasticity of remittances.

Running the IRFs for a low level of financial depth (measured by the ratio of private sector credit to GDP) and a high level of financial depth, shows that in the latter, the elasticity is not significant at conventional levels (Figure 8). However, the difference in the elasticity is marginal at best, probably because banks are typically not the cheapest channels for remittances. We also use two alternative measures of financial development: the geographical coverage of financial institutions captured by the number of branches per km2 and the number of deposit accounts per 1,000 adults; the results are similar, although the difference between the elasticity at the 10th and 90th percentile is larger (see Annex Figure 3).

Figure 8.
Figure 8.

Cost-Elasticity of Remittances in Countries with Low and High Financial Development

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

• Correspondent banking relationships (CBRs)

CBRs are essential for the smooth operation of the remittance markets as they facilitate cross-border transactions after small remittance payments are aggregated by remittance service providers, including MTOs. In the past decade, many countries, particularly small states, have been confronted with a sharp decline in CBRs. For some global banks, the risks from maintaining business with some jurisdictions and the compliance costs associated with anti-money laundering/combating the financing of terrorism (AML/CFT) rules far outweigh the potential profitability from these activities. While countries that have lost CBRs were able to redirect remittances via alternative channels, generally this has come at a higher cost and a reduced the scope of services, resulting in higher market concentration (IMF, 2017).

Unfortunately, the CBRs time series data published by the Committee on Payments and Market Infrastructures (CPMI) are only available at the sub-region level.12 We use the weighted average number of active correspondents by sub-region13 to test the heterogeneity of the cost elasticity of remittances with respect to CBR development. The results should be interpreted with caution, given the implicit assumption that CBR trends at the country level are correlated with that of the region. Figure 9 shows that remittances appear to react more to costs where the number of correspondent banks is smaller.

Figure 9.
Figure 9.

Cost-Elasticity of Remittances and Correspondent banking relationships

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

Cost-adaptation policies

• Price transparency

Opaque transaction costs create information asymmetries that distort competition and undermine the ability of remittance senders and receivers to choose the most cost-effective option that meet their needs. While the consumer generally is informed of the transfer fee, the awareness of the cost of the foreign exchange is typically unknown (Ratha and Riedberg, 2005). In some cases, a higher exchange rate margin reflects exchange rate risks, but there is mounting evidence that it can also conceal hidden transfer fees.

To measure price transparency, we use the share of remittance providers classified as “transparent” from the World Bank Remittance Prices Worldwide database. A remittance service provider is categorized as transparent if it did provide the exchange rate applied to the transaction, and non-transparent otherwise. At the 10th percentile of the sample (85 percent of transparent remittance service providers), the elasticity or remittances is -0.23 in the first quarter following a shock to transaction cost. This elasticity drops to -.09 for countries where all remittance service providers are deemed transparent (Figure 10). This implies that for countries with low levels of transparency, an increase in transparency level has a larger impact on remittances than in countries where transparency is already large.

Figure 10.
Figure 10.

Cost-Elasticity of Remittances and Price Transparency

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

• Financial Literary

Limited financial literacy constrains the ability of migrants and their families to take full advantage of the range of remitting options available and internalize all information available in the decision to select the most-cost effective option, particularly in an environment where price transparency is a source of concern. Remitters with limited financial literary may also tend to stick to the remittance service they are used to, potentially forgoing the opportunity to switch to a new remittance provider that is less costly or provides a better service. Put differently, if it is the case that individuals are not familiar with certain products, they will not demand them. Additionally, the lack of financial literary can also push some migrants to the informal remittance market. Kosse and Vermeulen (2014) find that more highly educated migrants are less likely to transfer cash via informal intermediaries or to carry cash themselves. Considering the above arguments, boosting financial literacy has the potential to spur price awareness and reduce migrants’ reliance on informal remittance channels. It also facilitates the adoption of innovative remittance services, and stimulates competition, thereby contributing to reduce costs.

Absent a widely available indicator of financial literacy, the gross secondary school enrollment rate of the home country is used as a proxy in the literature, given that studies document a strong correlation between education and financial literary (see Van Rooij, Lusardi, and Alessi, 2011; World Bank, 2017). As expected, the IRF shows that in countries with a highly educated population, remittances appear to be less sensitive to transaction costs than in countries with low levels of education (Figure 11).14

Figure 11.
Figure 11.

Cost-Elasticity of Remittances and Education level

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

As an alternative indicator of financial literacy, we use the number of deposit accounts with commercial banks, since financial literacy goes hand in hand with financial inclusion. The results support the evidence that where financial inclusion is low (and most likely where the lack of financial literacy is widespread), high transaction costs do have a bearing on remittances (see Annex Figure 4).

• Information and communication technologies (ICT)

ICT development has transformed the remittance market in several ways. First, it has enabled Fintech companies to enter the market, which with the resulting increase in competition has helped bring down transaction costs. Second, it fosters digital remittances, which has increased significantly in the wake of the Covid-19 pandemic. Digital remittances do not require a trip to the remittance provider to initiate the transfer or to receive the fund, and increasingly bypass the traditional payment system, resulting in lower transaction costs for migrants. Advancements in mobile money have already demonstrated that financial services can be provided cost-effectively to unbanked and under-banked population (e.g.; M-PESA in Kenya;--see also Andrianaivo and Kpodar (2012) on financial inclusion and mobile phone penetration). Not only does ICT development contribute to cost reductions in the remittance market, but it also has the power to reduce information asymmetries. The information economics literature has illustrated how incomplete information raises transaction costs and prevents financial markets from achieving socially efficient outcomes (Stiglitz, 2017).15 ICT development leads to a higher access to the Internet, which allows remitters to compare costs across various remittance services and pick the most competitive one. Therefore, it can reduce considerably information costs where price transparency is weak.

To capture ICT development, we use the ICT Development Index (IDI) compiled by the International Telecommunication Union (ITU). The IDI comprises 11 indicators divided into 3 groups measuring ICT access,16 ICT use17 and ICT capability18 (see ITU, 2014, for more details). 19The IDI is the simple average of the 3 sub-indices, themselves computed as the simple average of the normalized value of their components.20 Figure 12 shows the IRFs for the high and low regimes of overall ICT development and for the 3 dimensions: access, use, and capability, the results are shown in Annex Figure 5, 6 and 7, respectively.

As expected, higher ICT development is associated with a lower cost-elasticity of remittances. We find the same result for ICT access and ICT capability, but surprisingly the result is not conclusive for the ICT use sub-index. This suggests that the individual dimension of ICT development may not matter taken in isolation, but in aggregate, their complementarity plays an important role.

Figure 12.
Figure 12.

Cost-Elasticity of Remittances and ICT Development

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The ICT development index is the average of the ICT access, use and capability sub-indices.Source: Authors’ calculations.

A key dimension missing from the IDI is ICT affordability, which has been a severe constraint to ICT diffusion in many developing countries. We construct an ICT affordability index21 and re-run the IRF. The result show that the elasticity of remittances to transaction costs is stronger in countries with high ICT costs than in those with low ICT costs (Figure 13). Combining the 4 dimensions of ICT development into an augmented IDI index also confirms the earlier results.22

Figure 13.
Figure 13.

Cost-Elasticity of Remittances and ICT Affordability

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The ICT affordability index is calculated as the average of the normalized value of the fixed-broadband monthly subscription charge, the price of 3-minute mobile local call (off-peak rate), and the price of a 3-minute local call to a fixed-telephone line (off-peak rate). Variables are normalized to have a zero mean and unit variance.Source: Authors’ calculations.

C. Examining the Response of Remittances to Transaction Cost: Evidence from Micro Data from the US-Mexico Corridor

In this section, the objective is to use micro-data to test the robustness of the results found at the macro-level to the extent that the relevant variables are available. The micro-analysis focuses on the US-Mexico corridor for several reasons. First, it is the largest corridor in the world, with a volume of remittances of about US$25bn in 2017, more than 70 percent higher than the second largest corridor: USA-China. Second, it also one of the most competitive remittance markets, although transaction costs have not declined much in the past decade, given that they were already relatively low to start with. The total number of remittance service providers increased from 20 in 2011 to 26 in 2020, reflecting the entrance of new MTOs. While the number of banks providing remittance transfers declined from 7 in 2011 to 4 in 2020, the number of MTOs increased from 13 to 22 over the same period.23 Yet, the cost of a US$200 remittance, which stood at 5.5 percent in 2011, only declined slightly to 4 percent in 2020. Third, it is to our knowledge the only corridor with publicly available micro-data.24,25

The annual survey was carried out by the Bank of Mexico from 2013 through 2017. The survey took place in December of each year and was administrated to Mexican citizens residing in the US who visited Mexico by land or air. It covers demographic characteristics of migrants, the amount and frequency of remittances sent to their families, the fees paid and the means of transfer. The sample size varies over the years from 6,800 to 13,000 individuals. The survey questions also vary somewhat from one year to the next, though it was reasonably straightforward to combine the yearly surveys. It is not a “true panel", as the individuals cannot be tracked over time.

We adopt a simple linear model where we regress the volume of remittances sent in a year per individual on the fees reportedly paid as well as the socio-economic characteristics of the individual. Since the survey does not allow to identify the individuals over time, the identification of the elasticity of remittances to cost relies on the cross-sectional dimension of the dataset.

Table 2 shows the results from the fixed-effect estimations, which provides strong evidence of the detrimental impact of high transaction costs on remittances. The coefficient on transaction costs is negative and highly significant for all specifications, suggesting that migrants who face higher transaction costs tend to remit less, even after controlling for their income level, age, gender, level of education and the number of years lived in the US. The results are similar when the regressions are estimated for each survey year separately (results available upon request). The regressions also include time dummies, fixed-effects for the US state of residence, the state of residence of their families in Mexico and the sector of employment in the US.

Table 2.

Transaction Costs and Remittances in the US-Mexico Corridor

article image

Notes. OLS and fixed effect estimations. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. The dependent variable is the annual remittances reported by the survey respondents. Gender is a binary variable equal to 1 for men, and 2 for women. The level of education ranges from 1 (no education) to 13 (postgraduate)

Regarding the control variables, income is positively correlated with remittances, while men appear to remit more than women. Age is negatively associated with remittances, but the coefficient becomes insignificant once the number of years the individual lived in the US is controlled for. While this is not surprising since there is a strong correlation between the two variables (the correlation coefficient is 0.7), this result implies that as Mexican migrants stay longer in the US, they send less money back home, probably reflecting loosening family connections with the home country. Consistent with Faini (2007), we also find that more skilled migrants exhibit a lower propensity to remit.

In the previous section, we highlighted the role of financial literacy in dampening the sensitivity of remittances to transaction costs. The same hypothesis is tested with the micro data by introducing an interaction term between the education level of the migrant and the transaction costs. The finding in column 1 of Table 3, points to a lower elasticity among skilled migrants as the coefficient on the interaction term is positive and statistically significant. Since the survey also collect information on whether the remittance sender or receiver has a bank account, this dummy variable is used as a proxy of financial literacy and is interreacted with transaction costs. The results show that access to a bank account for both the sender and remitter (column 2 to 4) is associated with a lower cost elasticity for remittances.

Table 3.

Transaction Costs and Remittances in the US-Mexico Corridor: The Role of Financial Literacy

article image
Notes. OLS and fixed effect estimations. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. The dependent variable is the annual remittances reported by the survey respondents. Gender is a binary variable equal to 1 for men, and 2 for women. The level of education ranges from 1 (no education) to 13 (postgraduate)

As remittance providers tend to charge more for smaller amounts, this could raise a concern about a potential endogeneity. Table 4 deals with this issue in three ways. First, the transaction costs are instrumented by the type of service provider or medium of transfer used.26 The idea is to exploit the cost differential between providers such as banks and MTOs, as done for the panel IRF. The result presented in Table 4 (column 1) shows that the elasticity from the IV estimation is not materially different from the previous estimates in Table 2. Second, since migrants that remit more frequently are those that sent small amounts, the frequency of remittances is added to the specification to indirectly control for the size of the remittances. In addition, the share of remittances in the income of the remitter is used as a dependent variable as it is likely to be less correlated with transaction costs than the numerator itself. The result (column 2, Table 4) confirms the previous findings. Third, we kept the specification in column 2, and instrument the cost by the type of service provider as in column 1. Once again, the transaction costs continue to show a statistically significant and negative sign, although the magnitude is lower than previously seen.

Table 4.

Transaction Costs and Remittances in the US-Mexico Corridor: Instrumental Variable Approach

article image
Notes. Fixed effect estimations. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. Gender is a binary variable equal to 1 for men, and 2 for women. The level of education ranges from 1 (no education) to 13 (postgraduate).

IV. Conclusion

This paper uses a novel quarterly data set for 71 countries over a 10-year period, from 2011Q1 to 2020Q4, to investigate the elasticity of remittances to transaction costs. The results confirm the priors. A 10-percentage point decrease in transaction costs leads to a 0.9 percent increase in remittance in the short-run, but has no discernible impact in subsequent quarters, suggesting a short-run effect. This result implies that moving from the 2020 level of transaction costs (6.3 percent) to the SDG target of 3 percent will generate an additional US$32.2 billion in remittances, much larger than the direct cost savings. Therefore, migrants would not only fully pass the cost savings to their families, but also send more than they used to, confirming the powerful nature of cost reductions.

Since remittance costs continue to remain high in many countries, the paper looks at two broad set of policies: cost mitigation and cost adaptation policies that can help reduce the adverse effects of high transaction costs on remittances. While cost-mitigation policies directly tackle the root causes of high remittance costs, cost adaptation policies seek to reduce information asymmetries in the remittance market. Looking at the cost-mitigation policies, the econometric results suggest that when competition in the remittance market is high, the financial system is more developed, and transactions with correspondent banks hold up, the elasticity of remittance to transaction costs is much lower, ceteris paribus. This indicates that high transaction costs have lower impact on remittances send in these cases. Regarding cost-adaptation policies, the results indicate that more transparency on remittance transaction costs, improved financial literary and higher ICT development inhibit the sensitivity of remittances to high transaction costs, all else equal.

The paper also uses micro data from the USA-Mexico corridor with over 37,000 individuals surveyed during 2013-17 to confirm that migrants who face higher transaction costs tend to remit less, after controlling for socio-economic characteristics. In particular, education levels and access to a bank account diminish some of the cost-elasticity of remittances, confirming the findings from the panel results.

While remittances are mostly an individual decision, governments have an important role to play in influencing that decision: they can promote competition among banks and money transfer operators through adapting regulations. Forcing transfer companies to list transparently all their prices and providing information to migrants and their families could help them choose the most cost-effective remittance services, which would ultimately drive the costs down. Improving educational outcomes to facilitate the acquisition of financial literacy should be enhanced as well.

While these various factors were identified to individually modify the elasticity of remittances to costs, there is an interdependence also between the various forces, that typically reinforce each other. Therefore, moving on several fronts simultaneously would have a multiplicative effect. Deepening the financial system and improving financial literacy simultaneously will support each other further and help the remittance flows. Reducing the cost of remitting to the SDG target could significantly boost for remittances, creating an important tool to enhance capital flows to developing countries to finance development without creating excessive government debt.

Annex 1. Sample Composition1

article image

Annex 2. Summary Statistics and Correlation Matrix

1. Summary Statistics

article image

2. Correlation Matrix

article image

Annex Table 1.

Cost-Elasticity of Remittances: Local Projections

article image
Notes. Fixed effect estimations. Time dummies included. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. The shaded line represents the coefficients of the impulse response function shown in Figure 6.

Annex Table 2.

First-Stage Regression

article image
Notes. Fixed effect estimations. Time dummies included. Robust standard errors in brackets. *,**,*** denote significance at 10 percent, 5 percent and 1 percent, respectively. USD exchange rate denotes the units of local currency per USD.
Annex Figure 1.
Annex Figure 1.

Cost-Elasticity of Remittances: Instrumental Variable Local Projections

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

Annex Figure 2.
Annex Figure 2.

Cost-Elasticity of Remittances: Low vs High Transaction Cost

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The IRFs is obtained by interaction the variable on transaction costs with a dummy variable taking 1 if they are above the sample average and zero otherwise.Source: Authors’ calculations.
Annex Figure 3.
Annex Figure 3.

Cost-Elasticity of Remittances with respect to the Geographical Coverage of Financial Institutions

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.

Annex Figure 4.
Annex Figure 4.

Cost-Elasticity of Remittances with respect to Access to Deposit Accounts

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval.Source: Authors’ calculations.
Annex Figure 5.
Annex Figure 5.

Cost-Elasticity of Remittances and ICT Access

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The ICT access sub-index is calculated as the average of the normalized value of the number of fixed telephone subscriptions per 100 inhabitants, number of mobile-cellular telephone subscriptions per 100 inhabitants, international Internet bandwidth per Internet user, proportion of households with a computer, and proportion of households with Internet access. Variables are normalized to have a zero mean and unit variance.Source: Authors’ calculations.

Annex Figure 6.
Annex Figure 6.

Cost-Elasticity of Remittances and ICT Use

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The ICT use sub-index is calculated as the average of the normalized value of the percentage of individuals using the Internet, fixed (wired)-broadband subscriptions per 100 inhabitants, and wireless-broadband subscriptions per 100 inhabitants. Variables are normalized to have a zero mean and unit variance.Source: Authors’ calculations.
Annex Figure 7.
Annex Figure 7.

Cost-Elasticity of Remittances and ICT Capability

Citation: IMF Working Papers 2022, 218; 10.5089/9798400224270.001.A001

Notes: Shaded area represents the 90 percent confidence interval. The ICT capability sub-index is calculated as the average of the normalized value of the adult literacy rate, gross secondary enrolment rate, and gross tertiary enrolment rate. Variables are normalized to have a zero mean and unit variance.Source: Authors’ calculations.

References

  • Alcaraz, C., Chiquiar, D. and Salcedo, A., 2012, “Remittances, Schooling, and Child Labor in Mexico,” Journal of Development Economics, 97(1), pp. 156-165.

    • Search Google Scholar
    • Export Citation

  • Ahmed, J. and Martínez-Zarzoso, I., 2016, “Do Transfer Costs Matter for Foreign Remittances? A Gravity Model Approach.” Economics, 10(1).

    • Search Google Scholar
    • Export Citation

  • Ahmed, J., Mughal, M. and Martínez-Zarzoso, I., 2021, “Sending Money Home: Transaction Cost and Remittances to Developing Countries,” The World Economy, 44(8), pp.2433-59.

    • Search Google Scholar
    • Export Citation

  • Amuedo-Dorantes, C., 2014, “The Good and the Bad in Remittance Flows,” IZA World of Labor.

  • Amuedo-Dorantes, C. and S. Pozo, 2006, “Remittances and Insurance: Evidence from Mexican Migrants,” Journal of Population Economics, 19(2), pp. 227-54.

    • Search Google Scholar
    • Export Citation

  • Auerbach, A. J., and Y. Gorodnichenko, 2012, “Fiscal Multipliers in Recession and Expansion” in Fiscal Policy after the Financial Crisis, Alesina, A. and Giavazzi, F. (eds).

    • Search Google Scholar
    • Export Citation

  • Thorsten, B. and Martinez Peria, M. S., 2011, “What Explains the Cost of Remittances? An Examination Across 119 Country Corridors,” World Bank Economic Review, 25, pp. 105-31

    • Search Google Scholar
    • Export Citation

  • Beck, T., Janfils M., and Kpodar, K., 2022, “What Explains Remittance Fees? Panel Evidence”, IMF Working Paper, forthcoming.

  • Cazachevici, A., Havranek, T. and Horvath, R., 2020, “Remittances and Economic Growth: A Meta-Analysis,” World Development, 134(C).

  • Chami, R., Fullenkamp, C. and Jahjah, S., 2003, “Are Immigrant Remittance Flows a Source of Capital for Development?” IMF Working Paper No.189.

    • Search Google Scholar
    • Export Citation

  • Chiodi, V., Jaimovich, E. and Montes-Rojas, G., 2012, “Migration, Remittances and Capital Accumulation: Evidence from Rural Mexico,” Journal of Development Studies, 48(8), pp. 1139-55.

    • Search Google Scholar
    • Export Citation

  • Choi, H., and Yang, D., 2007, “Are Remittances Insurance? Evidence from Rainfall Shocks in the Philippines.” World Bank Economic Review, 21, pp.219–48.

    • Search Google Scholar
    • Export Citation

  • Cox-Edwards, A. and Rodríguez-Oreggia, E., 2009, “Remittances and Labor Force Participation in Mexico: An Analysis Using Propensity Score Matching,”, World Development, 37(5), pp. 1004-14.

    • Search Google Scholar
    • Export Citation

  • da Silva Filho, T., 2021, “No Easy Solution: A Smorgasbord of Factors Drive Remittance Costs,”, IMF Working Paper WP/21/199.

  • Yang, D., 2011, “Migrant Remittances,” Journal of Economic Perspectives, 25(3), pp. 129-52.

  • Demirguc-Kunt, A., Lopez Cordova, E., Martinez Peria, M. S. and Woodruff, C., 2009. “Remittances and Banking Sector Breadth and Depth: Evidence from Mexico,” Journal of Development Economics, 95(2), pp. 229-41.

    • Search Google Scholar
    • Export Citation

  • Driscoll, J. C. and Kraay, A. C., 1998, “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data,” Review of Economics and Statistics, 80, 549-60.

    • Search Google Scholar
    • Export Citation

  • Ferriani, F. and Oddo, G., 2019, “More Distance, More Remittance? Remitting Behavior, Travel Cost, and the Size of the Informal Channel,” Economic Notes, Banca Monte dei Paschi di Siena SpA, 48(3).

    • Search Google Scholar
    • Export Citation

  • Faini, R., 2007, “Remittances and the Brain Drain: Do More Skilled Migrants Remit More?,”, The World Bank Economic Review, 21(2), pp. 177-191.

    • Search Google Scholar
    • Export Citation

  • Freund, C. and Spatafora, N., 2008, “Remittances: Transaction Costs, Determinants, and Informal Flows”, Journal of Development Economics 86, pp.356-66.

    • Search Google Scholar
    • Export Citation

  • Furceri, D., Loungani, P. and Zdzienicka, A., 2018, “The Effects of Monetary Policy Shocks on Inequality”, Journal of International Money and Finance, 85(C), pp. 168-86.

    • Search Google Scholar
    • Export Citation

  • Gibson, J., McKenzie, D. and Rohorua, H., 2006, “How Cost Elastic Are Remittances? Evidence from Tongan Migrants in New Zealand”, Pacific Economic Bulletin, 21(1), pp. 112-28.

    • Search Google Scholar
    • Export Citation

  • IFAD, 2021, “11 reasons Why Remittances Are Important,” available at https://www.ifad.org/en/web/latest/-/11-reasons-why-remittances-are-important

    • Search Google Scholar
    • Export Citation

  • International Monetary Fund, 2017, “Recent Trends in Correspondent Banking Relationships: Further Considerations,” IMF Policy Paper, International Monetary Fund.

    • Search Google Scholar
    • Export Citation

  • International Telecommunication Union (ITU), 2014, “Measuring the Information Society Report 2014,” available at https://www.itu.int/en/ITU-D/Statistics/Pages/publications/mis2014.aspx

    • Search Google Scholar
    • Export Citation

  • Jordà, Ò., 2005, “Estimation and Inference of Impulse Responses by Local Projections,” American Economic Review, 95(1), pp.161-82.

    • Search Google Scholar
    • Export Citation

  • Olea, J. L. M. and Plagborg-Møller, M., 2021, “Local Projection Inference Is Simpler and More Robust Than You Think” Econometrica, 89(4), pp. 1789-23.

    • Search Google Scholar
    • Export Citation

  • Kakhkharov, J., Akimov, A., and Rohde, N., 2017, “Transaction Costs and Recorded Remittances in the Post-Soviet Economies: Evidence from a New Dataset on Bilateral Flows” Economic Modelling, Vol.60, pp.98–107.

    • Search Google Scholar
    • Export Citation

  • Kpodar, K, Mlachila, M., Quayyum, S. and Vigninou G., 2021, “Defying the Odds: Remittances During the COVID-19 Pandemic,” IMF Working Papers 2021/186.

    • Search Google Scholar
    • Export Citation

  • Kosse, A. and Vermeulen, R., 2014, “Migrants’ Choice of Remittance Channel: Do General Payment Habits Play a Role?,” World Development, 62, pp.213-227.

    • Search Google Scholar
    • Export Citation

  • Mandelman, F. S. and D. Vilán, 2020, “Remittances and COVID-19: A Tale of Two Countries,” FEDS Note, available at https://www.federalreserve.gov/econres/notes/feds-notes/remittances-and-covid-19-a-tale-of-two-countries-20201230.htm

    • Search Google Scholar
    • Export Citation

  • Mbiti, I. and Weil, D., 2011, “Mobile Banking: The Impact of M-Pesa in Kenya,” NBER Working Paper 17129.

  • Andrianaivo, M. and Kpodar, K., 2012, “Mobile Phones, Financial Inclusion, and Growth,” Review of Economics and Institutions, 3(2).

  • Plagborg-Møller, M. and Wolf, C. K., 2021, “Local Projections and VARs Estimate the Same Impulse Responses,” Econometrica, 89(2), pp. 955-80.

    • Search Google Scholar
    • Export Citation

  • Mora-Rivera, J. and van Gameren, E., 2021, “The Impact of Remittances on Food Insecurity: Evidence from Mexico,” World Development, 140(C).

    • Search Google Scholar
    • Export Citation

  • Ratha, D. and Riedberg, J., 2005, “On Reducing Remittance Costs.” Development Prospects Group, World Bank, Washington, DC.

  • Ratha, Dilip. 2021. “Keep remittances flowing to Africa,” available at https://www.brookings.edu/blog/africa-in-focus/2021/03/15/keep-remittances-flowing-to-africa/

    • Search Google Scholar
    • Export Citation

  • Barnichon R. and Brownlees, C., 2019, “Impulse Response Estimation by Smooth Local Projections”, The Review of Economics and Statistics, 101(3), pp. 522-0.

    • Search Google Scholar
    • Export Citation

  • Levine, R., 1997, “Financial Development and Economic Growth: Views and Agenda,” Journal of Economic Literature, 35(2), pp. 688-726.

    • Search Google Scholar
    • Export Citation

  • Stiglitz, J., 2017, “The Revolution of Information Economics: The Past and the Future,” NBER Working Papers 23780.

  • Teulings, C.N. and Zubanov, N., 2014, “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses.” Journal of Applied Econometrics, 29(3), pp.497-514.

    • Search Google Scholar
    • Export Citation

  • van Rooij, M., Lusardi, A. and Alessie, R., 2011, “Financial Literacy and Stock Market Participation,” Journal of Financial Economics, 101(2), pp. 449-72.

    • Search Google Scholar
    • Export Citation

  • World Bank Group, 2017. “Global Financial Inclusion and Consumer Protection Survey, 2017 Report”. World Bank.

1

The SDGs include the objective to “by 2030, reduce to less than 3 per cent the transaction costs of migrant remittances and eliminate remittance corridors with costs higher than 5 per cent” (SDG 10.C).

2

A simple average is adopted as data on the market share of the different remittance service providers are not available.

3

Two countries have quarterly data from the 1970s, and this figure rises quickly to 30 countries in the early 2000s. From 2010, the sample reaches 70 countries before stabilizing above 90 countries from 2017 onwards.

4

Personal transfers include all current transfers in cash or in kind between resident and nonresident individuals, regardless of the source of income of the sender and the relationship between households. Compensation of employees refers to the income of cross-border, seasonal, and other short-term workers who are employed in an economy where they are nonresident, or residents employed by nonresident entities.

5

Figure 2 excludes two outliers: Armenia and Syria with more than a two-fold increase in remittance costs during the same period.

6

The results not shown in the paper are available upon request.

7

We also introduced in the regression an interaction variable between transaction costs and the dummy for the year 2020 and found that the cost-elasticity to remittances was lower during the pandemic year than in the previous ones. This might reflect the limited opportunities for migrants to use informal channels for sending remittances given the border closures.

8

A drop in transfer costs from 6.3 percent to 3 percent represent a 52.4 percent decline, which multiplied by the elasticity of 0.087 results in a 4.5 percent increase in remittances (equivalent to $32.2 billion based on the size of remittances in 2020). On the other hand, the accounting approach which assumes that a $1 dollar saved on transaction costs by migrants results in an additional dollar for their families gives $23.28 billion ((6.3-3)*705.5).

9

This is not an exhaustive list. The focus is on key policies that have been extensively discussed in the literature.

10

We are not taking a view on whether having a low elasticity is good or bad: it simply means less reduction in remittances when transaction cost increase, but it also means lower increase in remittances when transaction costs are reduced.

11

The results are similar using the ratio of the number of MTOs to the number of banks.

13

Average number of active USD, EUR and GBP correspondents, weighted by the volume of USD, EUR and GBP transactions by sub-region.

14

The implicit assumption is that the education level of migrants is correlated with that of their home country. One could argue that those who migrate are typically more educated, and in this case, the elasticity shown in Figure 11 might be smaller in absolute term. This should not have a material impact on the findings if the difference between the education level of migrants and that of the home country does not substantially vary across countries. In the case it does, the country fixed effects will also help control for that. Additionally, not only does the level of financial literacy of the migrant matter, that of the families at the receiving end also does.

15

A key insight of information economics literature is that social returns to information typically differ from private returns, in some cases they are greater, in others lower. In the case of remittances, the social costs of not being able to send remittances may be even higher than the private costs if remittances get invested by recipients in areas such as health and education that produce high social rates of return.

16

The access sub-index includes five indicators: fixed telephone subscriptions per 100 inhabitants, mobile-cellular telephone subscriptions per 100 inhabitants, international Internet bandwidth per Internet user, proportion of households with a computer, and proportion of households with Internet access. All data are provided by the ITU.

17

The use sub-index includes three indicators: percentage of individuals using the Internet, fixed (wired)-broadband subscriptions per 100 inhabitants, and wireless-broadband subscriptions per 100 inhabitants. All data are taken from the ITU database.

18

The capability sub-index includes three indicators as well: adult literacy, gross secondary enrolment, and gross tertiary enrolment. All data are provided by the World Development Indicators (WDI, World Bank).

19

Given that the IDI has been discontinued in 2017 pending the addition of newly adopted indicators, for the purpose of this study, we recalculate the index using the underlying variables.

20

The normalized variables have a zero mean and unit variance. We also used a max-min transformation; the econometric findings are similar. It should be noted that the IDI calculated by the ITU does not normalize the underlying variables. Also, it attributes a 40 percent weight for the access sub-index, 40 percent for the use sub-index and 20 percent or the capability sub-index. In this paper, the underlying variables are normalized to ensure that the scales are fully comparable, and all sub-indices are equally weighted to avoid assigning a subjective weight to a component.

21

The affordability index used in this paper consists of 3 indicators: fixed-broadband monthly subscription charge, price of 3-minute mobile local call (off-peak rate), and price of a 3-minute local call to a fixed-telephone line (off-peak rate). All data are provided by the ITU.

22

Since lower ICT affordability is a negative outcome, the multiplicative inverse of this sub-index is used in combination with the access, use and capability sub-indices to compute the augmented IDI index.

23

The number of remittance operators may not reflect the wide array of remittance services available to migrants. For instance, the “Directo a México” program links the US domestic payment system to Mexico’s allowing to send the remittances from the US to Mexico at competitive rates.

25

There are several studies on remittances to Mexico (e.g Amuedo-Dorantes and Pozo, 2006; Cox-Edwards and Rodríguez-Oreggia, 2009; Demirguc-Kunt, Lopez Cordova, Martinez Peria, and Woodruff, 2011; Alcaraz, Chiquiar and Salcedo, 2012; Chiodi, Jaimovich, and Montes-Rojas, 2012; Mora-Rivera and van Gameren, 2021), but to our knowledge, none of them focuses on the transaction costs.

26

This variable includes 11 categories, with banks and MTOs being the most common.

1

According to World Bank classification.

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Working Papers
How Do Transaction Costs Influence Remittances?
Author:
Mr. Kangni R Kpodar
and
Patrick A. Imam
Volume/Issue:
Volume 2022: Issue 218
Publisher:
International Monetary Fund
ISBN:
9798400224270
ISSN:
1018-5941
Pages:
43
DOI:
https://doi.org/10.5089/9798400224270.001
  • <sup><xref rid="RA001fig01" ref-type="fig">Figure 1.</xref></sup>
    View in gallery
    Figure 1.

    Average fee as a share of a $200 USD remittance (2011Q1-2020Q4)

    (percent)

  • <sup><xref rid="RA001fig02" ref-type="fig">Figure 2.</xref></sup>
    View in gallery
    Figure 2.

    Change in Average fee for a $200 USD remittance (2016Q1-2020Q1)

    (percent)

  • <sup><xref rid="RA001fig03" ref-type="fig">Figure 3.</xref></sup>
    View in gallery
    Figure 3.

    Trends in Average fee for a $200 USD remittance and Interest Rate Spread (2011-20)

    (percent)

  • <sup><xref rid="RA001fig04" ref-type="fig">Figure 4.</xref></sup>
    View in gallery
    Figure 4.

    Average fee for a $200 USD remittance by Type of Provider (2011Q1-2020Q4)

    (percent)

  • <sup><xref rid="RA001fig05" ref-type="fig">Figure 5.</xref></sup>
    View in gallery
    Figure 5.

    Remittances Flows and Average fee for a $200 USD remittance (2011Q1-2020Q4)

  • <sup><xref rid="RA001fig06" ref-type="fig">Figure 6.</xref></sup>
    View in gallery
    Figure 6.

    Cost-Elasticity of Remittances: Local Projections

  • <sup><xref rid="RA001fig07" ref-type="fig">Figure 7.</xref></sup>
    View in gallery
    Figure 7.

    Cost-Elasticity of Remittances with Respect to Competition in the Remittance Market

  • <sup><xref rid="RA001fig08" ref-type="fig">Figure 8.</xref></sup>
    View in gallery
    Figure 8.

    Cost-Elasticity of Remittances in Countries with Low and High Financial Development

  • <sup><xref rid="RA001fig09" ref-type="fig">Figure 9.</xref></sup>
    View in gallery
    Figure 9.

    Cost-Elasticity of Remittances and Correspondent banking relationships

  • <sup><xref rid="RA001fig10" ref-type="fig">Figure 10.</xref></sup>
    View in gallery
    Figure 10.

    Cost-Elasticity of Remittances and Price Transparency

  • <sup><xref rid="RA001fig11" ref-type="fig">Figure 11.</xref></sup>
    View in gallery
    Figure 11.

    Cost-Elasticity of Remittances and Education level

  • <sup><xref rid="RA001fig12" ref-type="fig">Figure 12.</xref></sup>
    View in gallery
    Figure 12.

    Cost-Elasticity of Remittances and ICT Development

  • <sup><xref rid="RA001fig13" ref-type="fig">Figure 13.</xref></sup>
    View in gallery
    Figure 13.

    Cost-Elasticity of Remittances and ICT Affordability

  • <sup><xref rid="RApp01fig01" ref-type="fig">Annex Figure 1.</xref></sup>
    View in gallery
    Annex Figure 1.

    Cost-Elasticity of Remittances: Instrumental Variable Local Projections

  • <sup><xref rid="RApp01fig02" ref-type="fig">Annex Figure 2.</xref></sup>
    View in gallery
    Annex Figure 2.

    Cost-Elasticity of Remittances: Low vs High Transaction Cost

  • <sup><xref rid="RApp01fig03" ref-type="fig">Annex Figure 3.</xref></sup>
    View in gallery
    Annex Figure 3.

    Cost-Elasticity of Remittances with respect to the Geographical Coverage of Financial Institutions

  • <sup><xref rid="RApp01fig04" ref-type="fig">Annex Figure 4.</xref></sup>
    View in gallery
    Annex Figure 4.

    Cost-Elasticity of Remittances with respect to Access to Deposit Accounts

  • <sup><xref rid="RApp01fig05" ref-type="fig">Annex Figure 5.</xref></sup>
    View in gallery
    Annex Figure 5.

    Cost-Elasticity of Remittances and ICT Access

  • <sup><xref rid="RApp01fig06" ref-type="fig">Annex Figure 6.</xref></sup>
    View in gallery
    Annex Figure 6.

    Cost-Elasticity of Remittances and ICT Use

  • <sup><xref rid="RApp01fig07" ref-type="fig">Annex Figure 7.</xref></sup>
    View in gallery
    Annex Figure 7.

    Cost-Elasticity of Remittances and ICT Capability

Figure 1.

Average fee as a share of a $200 USD remittance (2011Q1-2020Q4)

(percent)
Figure 2.

Change in Average fee for a $200 USD remittance (2016Q1-2020Q1)

(percent)
Figure 3.

Trends in Average fee for a $200 USD remittance and Interest Rate Spread (2011-20)

(percent)
Figure 4.

Average fee for a $200 USD remittance by Type of Provider (2011Q1-2020Q4)

(percent)
Figure 5.

Remittances Flows and Average fee for a $200 USD remittance (2011Q1-2020Q4)
Figure 6.

Cost-Elasticity of Remittances: Local Projections
Figure 7.

Cost-Elasticity of Remittances with Respect to Competition in the Remittance Market
Figure 8.

Cost-Elasticity of Remittances in Countries with Low and High Financial Development
Figure 9.

Cost-Elasticity of Remittances and Correspondent banking relationships
Figure 10.

Cost-Elasticity of Remittances and Price Transparency
Figure 11.

Cost-Elasticity of Remittances and Education level
Figure 12.

Cost-Elasticity of Remittances and ICT Development
Figure 13.

Cost-Elasticity of Remittances and ICT Affordability
Annex Figure 1.

Cost-Elasticity of Remittances: Instrumental Variable Local Projections
Annex Figure 2.

Cost-Elasticity of Remittances: Low vs High Transaction Cost
Annex Figure 3.

Cost-Elasticity of Remittances with respect to the Geographical Coverage of Financial Institutions
Annex Figure 4.

Cost-Elasticity of Remittances with respect to Access to Deposit Accounts
Annex Figure 5.

Cost-Elasticity of Remittances and ICT Access
Annex Figure 6.

Cost-Elasticity of Remittances and ICT Use
Annex Figure 7.

Cost-Elasticity of Remittances and ICT Capability