Does the cost of remitting matter? Evidence from low-income countries on the effect of remittances on income distribution

ABSTRACT Part C of the United Nation’s (UN) 10th Sustainable Development Goal to reduce inequality addresses the high transaction costs faced by migrants when sending money home. The target is to reduce transaction fees to a maximum of 3%. To test the strength of these targets in reducing income inequality, we conducted instrumental variable (IV) estimations on 32 low-income countries for the period 2008–2021. The results show that remittances – typically made by low-paid workers–help reduce income inequality in the home country and that high transaction fees probably reduce the average remitted amount, rather than discouraging migrants from sending money home. Both remitting behaviors contribute to widening the income gap. Based on our estimations, the ‘affordable level’ is up to 1.93%. Fees greater than this would result in fewer remittances given the budget constraints faced the remitters – often employed in unskilled jobs in high-income countries. Hence, our results do not support the SDG 10.c target of 3%, and suggest that the UN should thus revise the target and set fees to be no higher than 2%.


Introduction
The economic intuition behind reducing income inequality through the flow of migrant remittances was discussed at the UN 2030 Agenda for Eradicating Poverty and Preserving the Environment, more commonly known as the 17 Sustainable Development Goals (SDGs).SDG 10.c, covering the reduction of inequalities, sets out a target to lower the fees associated with migrant remittances to an average of 3% of the transaction amount and to close high-cost transfer corridors completely.Including the question of cost of remitting in the SDGs is of substantial interest because, as Gautum notes: to reduce inequalities, political, economic and social policies should be universal, paying attention to the needs of disadvantaged and marginalized populations.The provision of financial services has been considered one of the most useful mechanisms for reducing poverty in developing countries (2019,92).
The belief is that transaction costs may be especially burdensome for the poor, who lack collateral and/or credit history.Hence, decreasing these costs will help to reduce the gap between the haves and the have-nots (Galor and Moav 2004).Further, through the dynamics of 'trickle-down economics', implementing policies that make remittances affordable has the potential to impact not only the families directly involved in the transfer but also the well-being of entire communities.This is because reducing costs means that more disposable income goes to poor families, which lowers the income gap within communities in the home country of the migrant (MacNaughton 2017).
Thus, the main goal of our research is to explore the role that the cost of remitting plays in the distributional effects of remittances.As argued by some researchers, such as MacNaughton (2017), and by international organizations like the United Nations (SDG indicator 17.3.2, launched in 2015), remittances are a financial flow vital for accomplishing macroeconomic goals (for instance fair income distribution).Focusing on the cost of remitting allows us to test if this instrument can mitigate the gap between the poor and the rich.
It has been calculated that a 10% increase in remittances can reduce poverty levels by 1% (Rocher and Pelletier 2008).For example, in sub-Saharan Africa, remittances have reduced poverty by increasing the recipient households' disposable income (World Bank 2015) and reduced inequality by equalizing income across the region, albeit to a minimal extent (Akobeng 2016).This reduction in inequality is a motivation for many migrants to remit.Further, as Gautam (2019) states, the additional income received by a household can help to offset exclusionary practices, such as only providing access to formal financial services to the wealthy few.The result is greater economic empowerment.
However, according to data from Remittances Prices Worldwide (Second Quarter, 2020), bank fees attached to remittances can be substantial, ranging from US$50 and US$60 (per transaction of an amount equivalent to US$200) sent by Zambian migrants working in South Africa.Furthermore, fees charged by money transfer operators (i.e.Money Gram and Western Union) range between US$28 and US$32.Migrants, particularly Africans, are, in effect, paying a remittance 'super tax' that drains much-needed funds from families, communities, and entire countries (Morreti-Langford 2014, 540), holding back a nation's development (Watkins and Quattri 2014).Reducing this super tax would mean that more of the nearly US$2 billion per year spent on fees would go to the families of migrant workers, giving them additional opportunities to consume, save, and invest (Overseas Development Institute 2014).Lower transaction costs may also help to curb blackmarket and other informal forms of transfer, shifting more remittances into the formal economy (Kakhkharov, Akimov, and Rohde 2017) or incentivizing more migrants to send money home.So, while making remittances more affordable may result in decreased revenue per transfer, overall revenues could rise due to a rise in the volume of remittances.In fact, international remittances are already growing annually.However, at the premium fees paid today to remit, these increases are only serving to exacerbate the problem, as income continues to be concentrated among the few at the expense of the many.It is therefore becoming increasingly urgent to address this issue.
That said, globally, inequality has declined over the last several decades, but only as the result of higherthan-average reductions in certain countries.In many countries, inequality has actually increased.Interestingly, this includes China and India -the two top remittance destinations (MacNaughton 2017).Hence, one of the main motivations for this study is to go beyond the specific phenomenon of migrant transfers and their related costs and to explore the feasibility of achieving SDG Goal 10. c. 1 The inconclusive findings in past studies, combined with new channels in the distributional impact of remittances, support a call for renewed study of the issue.To the best of our knowledge, this paper is the first attempt to address the distributional effects of remittances in the recipient countries with a special interest in the cost of remitting.Our theory is based on the premise that this relationship is strongly dependent on the transaction costs faced by migrants when sending money home.We test for amounts equal to the equivalent of US$200.We hypothesize that two kinds of remitting behaviors may arise if the cost of remitting is high: (i) remitters are dissuaded from remitting (i.e. the number of remittances would drop); and (ii) remitters continue to remit but the remitted amount is less.We use stylized facts on a sample of 32 low-income countries to test which behavior is more probable.In this study we recognize the existence of other means to send money home.However, in our estimates, we only use recorded remittances.
Our analysis is based on the remarkable shortcomings in the literature, flaws rooted in a lack of consideration given to the impact of cost of remitting as a conditional factor in making remittances.We comprise a quantitative study using two econometric techniques and a literature review.In the literature review, we refer to the set of theories on which we build our hypothesis and discuss why remittances are expected to benefit relatively poor households and, therefore, positively impact income equality.We also outline evidence that shows remitters are highly sensitive to the cost of remitting, particularly with small amounts.The models capture total fees spent on remittances and the relationship between formal remittances and income inequality.Additionally, we control for other potential determinants of income inequality.The preliminary analysis shows that an increase in the cost of remitting leads to a reduction in the average amount remitted which is more than the reduction in the average number of remittances.The regression analysis reveals that when remitters are dissuaded from remitting, this probably diverts important resources from low-income families and exacerbates income distribution issue vis-a-vis remitters who continue to remit smaller amounts.Both effects support our hypothesis that high transfer costs deter small remittance amounts.This reduction in remittances is most likely borne by recipients in the bottom share of the income distribution and thereby amplifies the income gap in the recipient country.The first contribution of this paper is its novel approach.The second contribution concerns the specificity of the study sample (most of remitters from low-income countries are from poor families).
The remainder of this paper is organized into four sections.Section 2 presents the theoretical framework and the literature review and the intuition behind our hypothesis.Section 3 describes the econometric model, data, and empirical findings.The results are discussed in Section 4. Section 5 summarizes our main findings and provides recommendations to encourage competition among money transfer services and eliminate the common duopoly of national post offices and money transfer providers.

Theoretical framework and literature review
We argue that the New Economics of Labour Migration (NELM) framework (Stark and Bloom 1985) and the neoclassical and middle-range theories of migration may justify a theoretical link between remittances and income inequality (c.f.Fransen and Mazzucato 2014).The former theory recognizes that migration decisions are based on household strategies to diversify sources of income and to ease access to credit.Migration acts as an informal social protection strategy that helps households to cope with weak or absent credit (Alpaslan et al. 2021).The latter two theories presume that migration is motivated by economic gains that would be concretized mainly through remittances.The relative deprivation hypothesis under NELM argues that relatively poorer households are more likely to receive remittances because they are more in need of them than better-off households.Migrants that come from relatively less privileged socioeconomic backgrounds tend to remit relatively more (especially during times of shock and crisis) than those that come from relatively affluent families (MacIsaac 2021).However, empirical evidence on the relative deprivation hypothesis is mixed.Fransen and Mazzucato (2014) argue that relatively wealthier households are more likely to receive remittances because they can more easily afford the migration costs of household members.This creates an informal contract whereby the migrating family member becomes obliged to 'pay back' his/her family.Lucas and Stark (1985) are among the first who identified this altruistic behavior.Remittances in this setting can be considered altruistic transfers to help family members back home.Yet another theoretical argument for remitting is strategic dissuasion (Fokkema, Cela, and Ambrosetti 2013), which posits that migrants remit to discourage their family members from joining them in their destination country.Migrants do this to protect their wages.In either case, poorer countries are helped to bridge the gap in income levels.
Another strand of theory explains the framework between changes in the cost of remitting and the remittance behavior.Ajzen (1991) developed the planned behavior model, which holds that there are three categories of consideration that guide human behavior.
The first is related to the likely expectations from the behavior.The second is related to the normative reactions of others.The third is related to the existence of elements that may impede or facilitate the performance of the behavior.The first two elements have been discussed by proponents of altruistic theory.The third factor points out that the behavior could be influenced by the presence of factors that play the role of facilitator or impeder.This might explain why remitter behavior may change when cost of remitting changes.In this regard, we refer to the wave of literature that has recently looked at the costs of remitting as an influential factor impacting the behavior of remitters and the growth rate of remittances (see Ahmed, Mazhar, and Martinez-Zarzoso 2020;Ferriani and Oddo 2019).Several studies have reported that when the cost of remitting increases, migrants either refrain from sending money or turn to informal channels, as seen in the case of Pakistani emigrants (Ahmed and Martinez-Zarzoso 2016). 2 Because banks tend to focus on high-value remittances, poor immigrants may feel uneasy about using those services, preferring more informal channels to remit money (Ahmed, Mazhar, and Martinez-Zarzoso 2020).Yet informal channels are not as secure; unrecorded remittances are at a high risk of being misused to launder money or to fund armed conflict (Nyamongo et al. 2012).Tellingly, migrants in Italy switch from informal channels to formal channels to send remittances when costs are low (Ferriani and Oddo 2019).The same behavior is seen in immigrants to Russia (Kakhkharov, Akimov, and Rohde 2017).However, we need to bear in mind that in accordance with altruism theory, it is also possible that remitters continue to remit but the disposal remitted amount is less.Mourao (2016) claims that Portuguese remitters tend to remit less when they experience increased economic challenges (i.e. higher costs of remitting) in the sending country.
Some studies have suggested that the stereotype of remitters to low-income countries may explain this behavior.First, remitters to low-income countries are temporal immigrants, tending to frequently remit small amounts, usually for altruistic purposes (Kratou, Pillai, and Sharif 2023).They are more sensitive to variations in the cost of remitting, in contrast to remitters of larger amounts, who appear less sensitive to transfer costs as they tend to remit infrequently (Ahmed, Mazhar, and Martinez-Zarzoso 2020).Ahmed, Mazhar, and Martinez-Zarzoso's (2020) study emphasizes that remittances in small amounts sent to cover Pakistani household needs are highly sensitive to transaction fees.Another strand of the literature argues that the capacity for low-earning migrants to send money home strongly depends on the budget constraint and the economic conditions in the country where they work (such as the level of income inequality, taxation considerations, and the cost of living).Migrants with low levels of qualifications and decreased purchasing power tend to remit less when they experience increased economic challenges in the destination country (Mourao 2016), and being less able to help their family can make them feel powerless and uneasy (Morreti-Langford 2014).
We also refer to sustainable development theory to explain why SDGs are interested in the cost of remitting by fixing a target to lower the fees associated with migrant remittances to an average of 3% of the transaction amount and to close high-cost transfer corridors completely.Pohoață, Diaconaşu, and Crupenschi (2020, 6) claim that according to this theory, 'everyone should benefit from growth; increased growth occurs without pollution, flatting Kuznets' curve as much as possible'.The belief is that transaction costs faced by migrants when sending money home may be especially burdensome for the poor, who lack collateral and/or credit access and credit history.Hence, relaxing these constraints has a more proportionate benefit in helping to close the gap between the haves and the have-nots (Galor and Moav 2004).
With regard to these theories that framework our study, we first hypothesize that remittance inflows reduce income inequality in low-income recipient countries.Second, while previous studies have not explored the factors that might prevent remittances to poor households in the home country, we claim that cost of remitting is highly influential.More specifically, we conjecture that low transfer costs are positively correlated with recorded remittances to poor families, which, in turn, contributes to closing the income gap in recipient countries.In the next section, we present the econometric model and the data.

The econometric model
We used two econometric techniques in our analysis.The impact of formal remittances on income inequality and the typical direction of the relationship was first estimated using ordinary least squares (OLS).Second, to systematically address the possibility of problems with endogeneity, we implemented instrumental variables.The standard errors of the different estimations show robustness in terms of autocorrelation and heteroscedasticity.The main relationships modeled included the distributional effect of recorded remittances on income inequity, as formulated in Equation ( 1), and the influence of cost of remitting on that correlation, as formulated in Equation ( 2): where Gini denotes the Gini coefficient of the remittances to the recipient country i, and Rem denotes the annual volume of international remittances received by country i per capita and as a percentage of its Gross Domestic Product (GDP) over period t in current US dollars.To correct for the size effect, the amount of the volume of remittances and remittances per capita have been adjusted for Purchasing Power Parity (PPP).
Cost denotes the cost of remitting the equivalent amount of US$200 to each country i of the sample over period t.We follow Brambor, Clark, and Golder (2006) methodology on interaction models.Thus, Rem*Cost is the interaction term between remittances Rem and costs Cost.It reflects the association between remittances and the Gini coefficient at average or median cost.
X is a vector of the explanatory variables, i.e. the economic/demographic/financial/political influences, for country i at Time t. a i and g t reflect the country and Time fixed effects and Ɛ is the error term.
Following Ajefu and Ogebe (2021), the potential endogeneity was handled with the instrumental variable technique in a second stage of analysis.Two appropriate instrumental variables were selected to mitigate measurement errors and to remove the problem of reverse causality.The validity of the external instruments was assessed with the Hansen (1982) test of overidentification.The null hypothesis is that there is no correlation between the instrument and the error term.A more detailed discussion of the two instruments is provided in sub-section 4.2.

Data
Our analysis on the distributional effect of recorded remittances on income inequality uses data spanning 32 low and low-middle income countries (covering six regions).Two main reasons justify this choice.For reliable inferences, we first notice that remittances to low-income countries grew much faster compared to the other groups of countries (Dash 2023).Further, using a more homogenous sample of countries then other studies enabled us to draw more insightful conclusions.Second, we needed a sample with available data; however, these data must also date back over a sufficient enough length of time to adequately measure longitudinal trends.We used panel data analysis over a 14-years period, from 2008 to 2021.Table A1 in the Appendix includes the list of countries by region.The instrumental variable (IV) technique was used to tackle the issue of data measurement.
We followed cross-country studies on income inequality (such as Tung and Thang 2023) that typically favor the Gini coefficient as the measure of inequity over other indexes, such as the Theil and Atkinson indexes, due to the widespread availability of Gini-based data and tools.Further, 'it is the most popular measure of socioeconomic inequality, especially in income and wealth distribution' (Sitthiyot and Holasut 2020, 1).The Gini coefficient series collected by the World Bank (WDI 2019) provides statistics measured against a range of more than 70 world development indicators (WDI) distributed from zero to 100%.The series pertaining to our analysis were personal remittances (personal transfers) and employee compensation measured in terms of amount.Personal transfers cover all current transfers in cash or in-kind made or received by resident households to or from non-resident households.Employee compensation reports the income of border, seasonal, and other short-term workers employed in an economy where they are not a resident and of residents employed by non-resident entities.Data are from the World Bank, based on International Monetary Fund (IMF) balance of payment data.
The limitation of this data source is that it does not record accurately remittance flows through informal channels: 'Because of difficulties in obtaining data on informal remittance transactions, the remittance transactions undertaken through informal channels are sometimes not well covered in current balance of payment data' (World Bank Development Indicators, Metadata, 2022, https://databank.worldbank.org/source/world-development-indicators).Many studies on remittances have noted this shortcoming, i.e. that data on informal remittance flows are scarce and too patchy to be able to construct time series data for an analysis with any degree of reliability (Ahmed and Martinez-Zarzoso 2016).Despite this limitation, this source is considered highly reliable as the IMF uses one specific definition, making the data easily comparable across space and time.Likewise, we were only able in this study to cover remittances flowing mostly through formal channels in our analysis.
It is true that remittances can come into a home country from many migrants' host countries.However, we followed the methodology of Freund and Spatafora (2008) to determine the main host country for each country i. by looking to the Organization for Economic Co-operation and Development (OECD) migration database and revealing the country j, which contains the largest share of country i's migrant workers.For example, Tunisia's migrant workers have different OCED destinations, with France being the main destination (i.e.contains the largest share of Tunisia's migrant workers).We used therefore the cost of remitting an amount equal to US$200 from France to Tunisia.
Data on the cost of remitting were collected from Remittance Prices Worldwide (RPW) data source.The data were calculated as percentage fees for remitting an amount equal to US$200.The average fees were provided for four quarters.Then we calculated the average of the four quarters.This database only collects data from firms (banks, money transfers operators and post offices) that have a transparent record and exclude informal financial channels', such as the Hundi network.This informal value transfer system is based on the performance and of a huge network of money brokers recognized as relatively inexpensive (less than 2% of the remitted amount) (Raza 2008).Furthermore, the senders do not need to provide identification (Suleri and Savage 2006).Covering the Indian sub-continent, Middle East, North Africa and Horn of Africa, these brokers have a geographical coverage, including a large number of developing countries, but they operate outside or parallel to traditional banking and financial systems and neither the amount nor the transfer fees are tracked.However, as previously mentioned, the purpose of this study is to emphasize on recorded remittances and recorded data of remittances fees to test our hypothesis.
In terms of control variables, the literature on the determinants of income inequality shows that changes in a nation's income inequality can be driven by many factors, including the level of economic development, inflation, government expenditures and trade openness, among other factors.Kuznet's relationship is one way of describing these changes.Details on each of the control variables follow.
Following previous empirical evidence (such as Kratou and Goaied 2018), we controlled for GDP per capita (the proxy for economic development) and its square value as the key explanatory variables to test the Kuznets (1955) hypothesis.The latter recognizes that during the early phase of development, inequality increases as a result of the rural and urban income disparity, which tends to narrow when economic development spreads.Macroeconomic instability is another influence on income inequality, and is thought to have a detrimental impact on the poor (i.e. a positive coefficient) because only the rich can invest in capital, land, foreign investment and other such assets (Bulíř 2001).
To capture this effect, we used the GDP deflator from WDI as a measure of inflation.
We controlled for public spending to test whether pro-poor public policies are effective in ameliorating income inequality.Ismaulina, Abd Majid, and Nasir (2022) showed that when government allocates sufficient spending for public facilities and finance projects that create job opportunities, income inequality is probably reduced.Following a range of studies (Chiara and Gerussi 2013), we tested the distributive effect of trade openness in our sample.There is a controversy in the literature on this relationship.On one side, some empirical studies (such as Kratou and Goaied 2016) align with the Heckscher-Ohlin-Samuelson hypothesis in recognizing that trade openness can stimulate relative demand and increase the income of unskilled workers and thus reduce the gap between unskilled and the skilled wages.Other empirical studies (such as Sharma and Abekah 2017) contradict the trade theory and argue that the new wave of trade openness requires complementarities between manufacturing, services and skills.
We also controlled for statistics on the number of unskilled migrants originating from each developing country (of our sample) to OECD countries.This variable is a proxy for the number of remitters.This aligns with the literature covered by Faini (2007) and Niimi et al. (2010).Data on the 'brain drain' was sourced from Brücker, Capuano, and Marfouk (2013).Notably, these data are not entirely accurate due to the difficulties in gauging migration by level of education or skills at the macro level.Moreover, the data also do not capture migrants staying in non-OECD countries, particularly the Persian Gulf, although they are accurate enough to provide a reasonable proxy.Other control variables include age dependency and level of democracy.Following Ebeke and Le Goff (2010), we controlled for age as the proportion of people under 15 or over 64 against people of working age.A high ratio suggests a wider income gap.This is because some individual economic agents are the sole caregivers of a household and this makes the dependency ratio high.We also tested for the influence of political regimes on income inequity, using the level of democracy from the Polity IV index.This variable was counted on a scale of 0 (fully autocratic) to 10 (fully democratic).This indicator is broadly used in political science and economics (e.g.Bjørnskov 2010).We used it to check the assumption that democratic regimes facilitate the distribution of wealth more fairly through pro-poor and inequality-averse policies.
We also confirmed robustness checks using data on income percentiles (i.e. the income shares of the bottom 50% of income earners to the middle 40%) from the World Inequality Database (WID.world2019) as an alternative variable for income inequality.The interpretation of this coefficient is opposite to that of the Gini coefficient (i.e. higher ratios indicate less inequity).A full description of the variables used in this study is provided in Table A2 of the Appendix.Descriptive statistics are shown in Table A3.
The preliminary analysis, illustrated in Figure 1, indicates that formal remittances are negatively correlated with the Gini coefficient in recipient countries, aligning with previous findings, such as Taylor et al. (2005) and Zhu and Luo (2008).This might corroborate the assumption that countries that are highly dependent on remittance inflows experience a low level of income inequality.
Figure 2 is a scatter plot of the cost of remitting (counted as the percentage fees of sending an amount equal to US$200) versus the Gini coefficient.The positive relationship shows that a cluster of countries with high levels of income inequality appear to be among the  most expensive remittance corridors (i.e. of about 20%), such as Zambia (ZMB), Rwanda (RWA), Kenya (KEN), Malawi (MWI) and Lesotho (LSO).The percentage fees of remitting vary between 15% and 20% of a remitted amount equivalent to US$200 (i.e.US$390 to US$40).This, notably, demonstrates that the remittances fees are excessive in these low-income countries.Migrants from Africa are, in effect, paying a remittance 'super tax' that drains much-needed funds from families, communities, and countries (Morreti-Langford 2014, 540).In the case of Zambia, a low-income country, the percentage fees (of remitting the equivalent US$200 from South Africa) are extremely expensive, despite it varying greatly depending on the firm.Money transfer operators charge between US$28 and US$32 (i.e.Money Gram and Western Union) and between US$50 and US$60 charged by the banking system (Remittances Prices Worldwide, these data correspond to the second quarter of 2020).At the same time, the Gini coefficients vary for countries with overall lower cost of remitting (i.e. less than 5%).This may indicate a threshold effect between the cost of remitting and income inequality, consistent with SDG 10.c.
Figure 3(a,b) shows stylized facts related to the effect of changes in remittances fees on remittance behavior.A 1% change (increase) in the cost of remitting leads to a 5.7% (slope, Figure 3(a)) change (drop) in the average remitted amount.A 1% change (increase) in the cost of remitting leads to a 4.9% (slope, Figure 3(b)) change (drop) in the average number of remittances.Despite the slight gap between the two slopes, we can deduce that an increase in the cost of remitting reduces the disposal amount to be remitted more than the reduction in the number of remittances.In the case of low-income countries, remitters continue to remit, but the amount to be remitted is less when he faces an increase in the cost of remitting.'Remittances continue to provide a critical lifeline for the poor and vulnerable, especially during crisis' (World Bank Press Release, 12 May 2021, 1).Altogether, these interesting stylized facts drive vital predictions that support the theoretical framework.However, the nature of the nexus between remittance, cost of remitting and income inequality will be more convincing if we control for the other determinants of income inequality using empirical tests.

Distributional effect of recorded remittances: pooled OLS
The estimated results of the pooled ordinary least squares (OLS) for the three remittance variables are presented in Table 1.These include: (a) remittances to GDP ratios; (b) volume of remittances in current US dollars (adjusted for PPP); and (c) remittances per capita (adjusted for PPP).The interpretation of the coefficients is consistent: the remittances and most of the control variables are statistically significant and persistent across the specifications irrespective of the measure of remittances which is used.As expected, the coefficient related to remittances is negative and highly significant for the different specifications.A 1% growth in remittances leads to a reduction in the Gini coefficient of approximately 2%.This result highlights the vitality of these flows for reducing income inequality in lowincome recipient countries and is consistent with previous findings (Azizi 2021;World Bank 2006).Furthermore, we support NELM theory and reveal consistency with the stylized facts (see Figure 1).We also corroborate with recent studies testing the same hypothesis on different data, approaches and samples (for instance: Gurbanov, Mammadrzayev, and Isgandar (2021) on the former Soviet Union; Kóczán and Loyola (2021); on Mexico (Kratou and Khlass 2022); on the Middle East and North Africa (MENA) region and (Tung and Thang 2023); on emerging countries).All these scholars have advocated altruism theory in claiming that this reduction in inequality is why migrants send money home to support poor families (Dasgupta and Kanbur 2011).The results related to control variables show that, as expected, the age dependency ratio is positive and statistically significant in all of the three regressions.High inactive population levels lead to more dependence on the household and a smaller share of income per capita.In contrast to our expectations, an increase in price level lowers income inequality.We refer to Alavijeh et al.'s (2017, 684) argument that 'a contractionary monetary policy is more effective at reducing inequality'.The results also show that when government spending is allocated in a productive manner, it contributes to reducing the community's income inequality (Ismaulina, Abd Majid, and Nasir 2022).In contrast to the Heckscher-Ohlin-Samuelson prediction, trade openness jeopardizes income inequality in the case of low-income countries, a finding which could be driven by a high demand for skill-based technology.Our results reveal that democracy is positive and highly significant.This suggests that democracy might have a detrimental effect on income inequality in low-income countries, supporting Bjørnskov's (2010) conclusion that democracy is not essentially a pro-poor political arrangement.

How the cost of remitting affects the distribution of recorded remittances
In the next regression, we test our second hypothesis and verify whether a change in the cost of remitting changes remittance behavior and, in turn, income distribution.Despite the robust results of the OLS, which point to the egalitarian impact of remittances, it is still possible that remittances are endogenous with respect to income inequality.One source is reverse causality, i.e. income inequality may attract remittances, whereby migrants send more money home when the income gap within households is high.In addition, remittances might increase the income gap if transfers benefit mostly affluent families.Other sources of endogeneity are also possible, such as measurement errors and omitted variables.The first refers to the fact that the volume of unrecorded remittances to low-income countries is not negligible.The second is related to uncontrolled exogenous shocks in the model (i.e.climate shocks, price shocks …).
Tackling the potential problem of endogeneity is crucial to avoid biased results.The instrumental variable method is the most convenient technique for overcoming this issue (Ajefu and Ogebe 2021).Here, the choice of external instrumental variables must be both correlated with the endogenous explanatory variable and act only on the dependent variable (i.e. the Gini coefficient) through its effects on the endogenous variable.It is also vital that the instrument shows time variability.In our case, the instrument should be correlated with the remittances and not be related to income inequality.Note: ***, ** and * denote significance at, respectively, the 1, 5 and 10% levels.Robust-t-statistics are in parenthesis.
We identified two potential external instruments meeting these criteria.The first is the sum of the recorded remittance inflows across the entire sample minus the remittances of the country (Chami et al. 2008).This instrument not only satisfies the validation criteria but also shows variability over time.For the choice of the second instrument, we refer to the literature on the determinants of remittances, such as the work of Swamy (1981), Lianos (1997), Abdel-Rahman (2006), and, more recently, Bidawi et al. (2022).All these scholars agree on the positive correlation between remittances and the host country's GDP per capita.Remittances are procyclical with the host country's economic growth (Hathroubi and Aloui 2016).In our case, we assume that remittance flows are responsive to the economic conditions of the migrants' destinations (Bidawi et al. 2022).A highincome per capita level would probably affect the behavior of remitting (Ajefu and Ogebe 2021).Migrants living in wealthier destination countries have a higher probability of sending remittances and send larger amounts of money to their households left behind (Bredtmann, Flores, and Otten 2019).
Table 2 presents the first-stage results for the migrant remittances instrumental variable, where the two external instruments are positive for predicting remittances and are relevant to be used in the second-stage (presented in Table 3).
In this regression, given by Equation ( 2), we include all the constitutive variables of the multiplication between remittances and the cost of remitting.The results (see Table 3) show, firstly, that remittance coefficients remain negative when we control for the cost of remitting.Second, the interaction variable between the cost of remitting and remittances shows a positive coefficient and is significant at the 1% and 5% levels.As hypothesized, two kinds of behavior may arise if the cost of remitting is high: (i) remitters may be dissuaded from remitting (i.e. the number of remittances would drop); and (ii) remitters may continue to remit but the remitted amount is less.When a 1% increase in the cost of remitting reduces the number of remittances to the recipient countries, this leads to a widening in income inequality by 2.1% (see Table 3, column 1).However, when a 1% increase in the cost of remitting reduces the dollar  Note: ***, ** and * denote significance at, respectively, the 1, 5 and 10% levels.Robust-t-statistics are in parentheses.The interaction variable between remittances and the cost of remitting is instrumented by the interaction between each instrument of remittances and the cost of remitting, respectively.
amount of remittances sent home, this leads to a rise in income inequality of 0.41% (see Table 3, column 2).An increase in the cost of remitting widens income inequality.The effect is more pronounced when the frequency and the number of remitting drops than the effect associated with a reduction in the remitted amount.
From our results, we can conclude that the effect of the cost of remitting on remittance behavior determines the ultimate effect of income inequality.That is, the results from low-income countries show that when remitters are dissuaded from remitting, this probably diverts important resources from low-income families and exacerbates the income distribution issue compared to a remitter who continues to remit smaller amounts.
Both effects support our hypothesis that high transfer costs deter small remittance amounts (deemed for our purposes to be US$200), and the money lost will most likely be borne by those in the bottom share of the income distribution, which in turn increases the income gap.Our analysis shows that this disincentive disappears at a transaction fee of about 1.93% of the remitted amount.These data are counted from the first derivative of the sum of the linear measure of remittances b 1 and the interaction term b 3 (see Table 3, column 1).This threshold is below the target of 3% and does not support SDG 10.c.Remitters from low income countries are not even able to afford a cost of remitting exceeding 2%.
In terms of the validity of the instrument used, we undertook the Kleibergen and Paap (2006) tests for weak instruments to verify the identification model.The Hansen-J statistics of over-identification suggest that the instruments are exogenous to remittances in all specifications.Therefore, use of these instruments in our model is warranted.
The results of the additional robustness checks are shown in Table 4.We excluded countries with a high level of dependence on remittances from the sample to detect reverse causation.The possible deterioration in income distribution in these economies could be due to a high ratio of remittances to GDP.Then we ran the estimation on the other countries of the sample to test Equation (2).The results are presented in column 1 of Table 4.The coefficient of the remittances to GDP ratio and of the constitutive term between remittances and the cost of remitting remain negative and positive, respectively, and highly significant.These findings do not alter the results from the entire sample.Our results are not affected by the particularities and specificities of countries heavily dependent on remittances.As a last robustness check, we refer to research that recommends using alternative measures of income inequality (i.e.Glassman and Branch 2016, 19).We further calculated the 'ratio of 40' -defined as the income share of the bottom 50% compared to that of the middle 40% -to capture the disparity between aggregate household income shares received by each quantile (Zhang, Simmel, and Nepomnyaschy 2022).Coefficient signs should be interpreted as opposite to the Gini coefficient.If, for instance, the bottom 50% experience a rise in their income compared to the middle 40%, the income dispersion between the two groups would be closed.As expected, the resulting values of remittances and of the interaction between remittances and the cost of remitting were positive and negative, respectively, and both coefficients were statistically significant at the 1% level.Notably, the size of the coefficients was less pronounced with this indicator compared to the Gini regressions.These results corroborate with those found with the Gini coefficient.The first derivative of the sum of the linear measure of Note: ***, ** and * denote significance at, respectively, the 1, 5 and 10% levels.Robust-t-statistics are in parenthesis.The interaction variable between remittances and the cost of remitting is instrumented by the interaction between the instrument of remittances and the cost of remitting.
remittances b 1 and the interaction term b 3 (see Table 4, column 2) show that a transaction fee of about 2.33% nullifies the marginal effect of remittances on income inequality.We find support for our hypothesis from all analyses and estimations.Our final conclusion is that recorded remittance flows improve the income levels of the poorest segments of the population in low-income recipient countries.But the reverse is true when we control for the variable of interaction between remittances and the cost of remitting.
The novelty in this paper is, firstly, the approach.We investigated how changes in the cost of remitting affect remitting behavior, which in turn affects in different ways income inequality.Second, the specificity of the remitter to low-income countries, which are low-paid and most likely to belong to the lowest deciles of the income distribution, make them highly sensitive to variations in the cost of remitting.One practical implication would be that efforts to reduce the cost of remitting to an affordable level (2% of the cost of remitting) enables the remitter's family to receive a substantial remittance and help to reduce income disparity in the recipient economies.In this sense, this study provides a new policy implication for the literature on development and income inequality.

Conclusion
In this paper, we approached the effect of recorded migrant remittances on income inequality from the premise that this relationship is not monotonic and depends heavily on the cost of remitting.To the best of our knowledge, this is the first research to test the basis of SDG 10.c by studying the changes in remittance rates associated with different pricing levels.We referred to theories such as the NELM framework, the neoclassical and middle-range theories of migration, planned behavior theory and sustainable development theory to build our hypothesis.
To tackle endogeneity issues, we applied the instrument variable technique and found that an increase in recorded remittances has a positive effect on income equality in low-income countries.This result, and other research in the field (e.g.World Bank (2006); Akobeng (2016)) hint at the possibility that the sender population of remittances is generally made up of recent immigrants with low skill levels and low incomes who frequently send remittances to support their families left behind.
An important finding from our stylized facts shows that an increase in the cost of remitting reduces the disposal amount to be remitted more than the reduction in the number of remittances.In the case of low-income countries, remitters continue to remit, but the amount remitted is less when he faces an increase in the cost of remitting.This result aligns perfectly with altruism theory.Low-paid remitters are highly sensitive to the cost of remitting when the amounts are small.When transfer costs are high, immigrants are left feeling powerless and uneasy about the process of sending much-needed money home (Morreti-Langford 2014).The results from our regressions show that when a 1% increase in the cost of remitting reduces the number of remittances to the recipient countries, this reduction leads, in turn, to a widening of income inequality by 2.1%.A corresponding reduction in recorded remittances should undercut habitual financial support for the immigrant's family.In turn, this widens the income gap among households in the home countries.However, when a 1% increase in the cost of remitting reduces the dollar amount of remittances sent home, this leads to an income inequality rise of only 0.41%.The effect on income inequality is more pronounced when the remitter is dissuaded from remitting.This result corroborates with the NELM framework and the neoclassical and middle-range theories of migration in revealing the importance of remittances as a lifeline and an economic gain for the recipient families.Based on our estimations, the affordable level is up to 1.93%; fees greater than this result in fewer remittances by low-paid workers.From this result we deduce that the SDG 10.c focus on lowering fees is important, but should actually be more ambitious and aim at fees not exceeding 2%.Remitters to low-income countries are not even able to afford the targeted remitting cost by the UN.
Several policy implications can be drawn from these findings.First, high cost of remitting is a drain on the resources of poor migrants and their families back home, for which remittances are a unique source of income.A commitment to reducing transaction costs means more of that income will be at the disposal of these vulnerable families.Second, an increased volume of remittances will improve financial access for the poor in developing countries (World Bank 2006).Third, a shift from informal to formal channels would give rise to more accurate measurements of remittances, which, in turn, would permit a better estimate of the funds available for investment in the economy (Kakhkharov, Akimov, and Rohde 2017).In addition, such a shift might improve the quality of empirical studies on remittances.Lastly, the shift from informal to formal channels is a profound issue in the modern-day international security landscape (Kakhkharov, Akimov, and Rohde 2017).The increase of armed conflicts and terrorist activities requires careful monitoring of international financial flows, which is easier to do when funds flow through transparent/ official channels.
Overall, our results suggest that, for recorded remittances to provide a development stimulus, the cost of remitting must be reduced by removing the common duopoly of post offices and money transfer providers over the market to encourage competition.Further, the targets set out in SDG 10.c remain unaffordable for remitters from low-income countries.Actually, the target should be revised by the UN and we suggest that the fees should not exceed 2%.This would need greater transparency in the pricing of different money transfer operators and of the banking system as well.
One limitation of this paper is that we only estimated the impact of recorded (not necessarily the total) remittances on income inequality.The scenario that low-paid migrants may send money home through informal channels (for instance, the huge network of money brokers operating in Hundi) when the cost of remittances is high remains possible.Future studies may look at this assumption.However, this strongly depends on the availability of micro data.

Notes
1. SDG target 10.c specifically advocates for reducing to less than 3% the transaction costs of migrant remittances and eliminating remittance corridors with costs higher than 5%.2. Costs are estimated at 3 to 5% for informal channels compared to up to 17% through formal channels (Morreti-Langford 2014).

Figure 2 .
Figure 2. Gini coefficient per average cost of remitting.

Figure 1 .
Figure 1.Gini coefficient per average number of remittances.

Figure 3 .
Figure 3. (a) Change in the cost vs Change in the amount remitted.(b) Change in the cost vs Change in the number of remittances.

Table 2 .
First-stage instrument variable estimates for remittances.

Table 3 .
The role of cost of remitting in the distributional effect of recorded remittances (dependent variable: Gini coefficient).