Remittances from Tongan migrant workers: Channels, costs, and potential gains from switching

Maeda, Hiroshi; Edwards, Ryan Barclay; Suryadarma, Daniel Working Paper Remittances from Tongan migrant workers: Channels, costs, and potential gains from switching ADBI Working Paper, No. 1427 Provided in Cooperation with: Asian Development Bank Institute (ADBI), Tokyo Suggested Citation: Maeda, Hiroshi; Edwards, Ryan Barclay; Suryadarma, Daniel (2024) : Remittances from Tongan migrant workers: Channels, costs, and potential gains from switching, ADBI Working Paper, No. 1427, Asian Development Bank Institute (ADBI), Tokyo, https://doi.org/10.56506/VZFR3397 This Version is available at: https://hdl.handle.net/10419/296819 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/ ADBI Working Paper Series REMITTANCES FROM TONGAN MIGRANT WORKERS: CHANNELS, COSTS, AND POTENTIAL GAINS FROM SWITCHING Hiroshi Maeda, Ryan Edwards, and Daniel Suryadarma No. 1427 January 2024 Asian Development Bank Institute The Working Paper series is a continuation of the formerly named Discussion Paper series; the numbering of the papers continued without interruption or change. ADBI’s working papers reflect initial ideas on a topic and are posted online for discussion. Some working papers may develop into other forms of publication. Suggested citation: Maeda, H., R. Edwards, and D. Suryadarma. 2024. Remittances from Tongan Migrant Workers: Channels, Costs, and Potential Gains from Switching. ADBI Working Paper 1427. Tokyo: Asian Development Bank Institute. Available: https://doi.org/10.56506/VZFR3397 Please contact the authors for information about this paper. Email: [email protected], [email protected], [email protected] Hiroshi Maeda was a master’s student at the Crawford School of Public Policy, Australian National University. Ryan Edwards is a fellow at the Crawford School of Public Policy, Australian National University. Daniel Suryadarma is a senior research fellow at the Asian Development Bank Institute, Tokyo, Japan. The views expressed in this paper are the views of the author and do not necessarily reflect the views or policies of ADBI, ADB, its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms. Discussion papers are subject to formal revision and correction before they are finalized and considered published. We are grateful to Dil Rahut for providing feedback. Asian Development Bank Institute Kasumigaseki Building, 8th Floor 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: +81-3-3593-5500 Fax: +81-3-3593-5571 URL: www.adbi.org E-mail: [email protected] © 2024 Asian Development Bank Institute ADBI Working Paper 1427 H. Maeda et al. Abstract The background of high remittance costs and countermeasures against the issue has drawn the attention of policymakers in the Pacific island countries in maximizing the benefits of migration and remittances on their economic development. We use a newly collected household survey about temporary migration in the region and implement a data audit on remittance costs to examine the current state of the remittance market and household decision-making around remittance channels in Tonga. We find that the household choice of remittance channels leads to a higher remittance cost because a large gap in remittance costs exists between high-cost and low-cost remittance service providers. Tongan workers choose high-cost providers mainly because of their ease of use. Within this context, Tongan households and the economy can gain more than 2% of remittances by switching to the lowest-cost providers. We also find that exposure to new information thanks to migration experiences, frequent remittances, and residential islands are the key household characteristics to voluntarily shift from cash transfers to online transfers or mobile money. The findings indicate that micro-level intervention on the household choice of remittance channels and providers can potentially contribute to reducing remittance costs in the region, but we first need to identify the barrier to switching to low-cost providers to design effective policy interventions. Keywords: remittance costs, migration, Tonga JEL Classification: F24, O15 ADBI Working Paper 1427 H. Maeda et al. Contents 1. INTRODUCTION .......................................................................................................... 1 2. BACKGROUND ............................................................................................................ 3 2.1 Why is Reducing Remittance Costs Important for Developing Countries and the PICs? ................................................................................................... 3 2.2 Remittance Market in Tonga ............................................................................ 5 3. DATA ............................................................................................................................ 5 4. RESULTS ..................................................................................................................... 9 4.1 The Choice of RSPs Increases Remittance Costs ........................................... 9 4.2 Household Characteristics and the Choice of Remittance Channels ............. 13 4.3 Simulation: How Much Would Be the Gain Made by Switching RSPs? ......... 19 5. DISCUSSION ............................................................................................................. 21 6. CONCLUSION ........................................................................................................... 22 REFERENCES ...................................................................................................................... 24 APPENDIXES 1 Issues in the PLMS Data Set ......................................................................... 28 2 Supplemental Figures and Tables .................................................................. 32 ADBI Working Paper 1427 H. Maeda et al. 1 1. INTRODUCTION Remittances play a crucial role in the economic development of the low- and middleincome countries including the Pacific island countries (PICs). Remittance inflows to low- and middle-income countries recorded $605 billion in 2021, uplifting the livelihoods of people through various channels, such as poverty reduction, the improvement of food security, and human capital investment (World Bank 2023c). However, high remittance costs remain a barrier. According to the Remittance Prices Worldwide, international migrants pay, on average, 6.25% of costs when they send $200 to their home countries (World Bank 2023b). The situation is even worse in the PICs, with the average remittance costs in the region reaching 8 to 10% (World Bank 2023a; Raithatha et al. 2021: 19). Yet, there is not much knowledge about the background of the high remittance costs in spite of a strong interest from policymakers in the region (PACER Plus 2023; NRBT 2017: 26). In this context, this paper attempts to unpack the current state of remittance market and the household behavior around the choice of remittance channels in Tonga. We choose Tonga as a case country because of the significance of remittances in the economy, the maturity of the remittance market, and a gap between policy interests and existing evidence. Tonga recorded the highest remittances-to-GDP ratio in the PICs at 43.9% in 2021 (Doan et al. 2023). Existing evidence shows that international remittances contribute to improving household welfare through poverty alleviation, wealth creation, increased household income and consumption, and improving subjective well-being (Gibson and McKenzie 2014; World Bank 2017b; Brown, Connell, and Jimenez-Soto 2014). The remittance market in Tonga is mature, consisting of low-and high-cost remittance service providers (RSPs) including commercial banks, international and localized money transfer operators (MTOs), and mobile money.1 The data from Remittance Prices Worldwide indicates that the choice of RSPs plays a crucial role in determining remittance costs in the country. The average remittance cost is higher than 8% in the Australia– and New Zealand–Tonga corridors, but there are remittance channels that cost less than 5%. Tongan policymakers are already keenly aware of high remittance costs in the country. The National Reserve Bank of Tonga expresses the need for a deeper understanding of why Western Union remains in a dominant position even though the remittance cost is higher than other services such as mobile money (NRBT 2017: 26). The Ministry of Finance shows an interest in measuring high remittance costs and identifying how they can reduce remittance costs (PACER Plus 2023: 24). However, there has been little research that has delved into the background of the high remittance cost in Tonga to aid effective policy reforms. Thus, it is important to provide empirical evidence on high remittance costs in Tonga to meet such strong needs from policymakers. We use two types of remittance cost data. First, we use the Pacific Labor Mobility Survey (PLMS) initiated by the World Bank and the Australian National University. The second type of data is audit data that we manually collected from Send Money Pacific and Saver Pacific. These are platforms that compare remittance costs by different RSPs and transfer methods, providing a wide range of information, including fixed fees, exchange rate margins, total cost in percentage, and the speed of transfers. 1 Money transfer operators (MTOs) are the operators such as Western Union and Moneygram but do not include commercial banks and mobile money. When we include either commercial banks or mobile money, we call them remittance service providers (RSPs). ADBI Working Paper 1427 H. Maeda et al. 2 The main results are fourfold. First, the audit data show that there are large differences in remittance costs between high-cost and low-cost RSPs. Ria provides the lowest-cost services at an average cost of 1.76% in the Australia–Tonga corridor, and Ave Pa’anga Pau provides the lowest-cost services at an average cost of 3.33% in the New Zealand–Tonga corridor. These figures are more than 20 percentage points lower than the highest-cost RSPs in the two corridors. Second, Tongan workers and households bear a higher cost of remittance by mainly using Western Union and Moneygram. Audit data show that these channels do not have cost advantage except the online transfers through Moneygram in the Australia–Tonga corridor. The PLMS reveals that ease of use is the most important factor in choosing RSPs and the users of the two popular RSPs are not cost-sensitive compared to the users of low-cost RSPs. Third, the estimated results of multinomial logit regression suggest three key factors to voluntarily shift remittance channels from cash to online transfers. A higher chance of obtaining information about remittance costs promotes the usage of online transfers because the participation in temporary migration programs and living experiences in Australia and New Zealand are associated with a higher uptake of online transfers through banks or MTOs and a lower uptake of cash transfers through MTOs. Frequent remittances also have a negative association with the use of cash transfers through MTOs while having a positive association with the use of online transfers through MTOs. Living in Tongatapu (the main island) positively relates to the uptake of online transfer services and mobile money, probably because of a higher mobile phone ownership rate and the household internet access rate in Tongatapu compared to other islands (Tonga Statistics Department 2022a: 10, 18). Living in Tongatapu also has a negative association with the use of cash transfers. Fourth, the simulation study reveals that switching the RSPs from two popular providers to the lowest-cost ones can bring a substantial benefit to Tongan households and the Tongan economy. We calculate the extent of gain that Tongan households and the Tongan economy would accrue if households who currently receive remittances from temporary migrants through Moneygram and Western Union switch to Ria in the Australia–Tonga corridor and to Ave Pa’anga Pau in the New Zealand–Tonga corridor. The results show that, for six months, Tongan households would gain 1,103,809 pa’anga, gaining 2.4% in the Australia–Tonga corridor. In the New Zealand–Tonga corridor, they would gain 424,596 pa’anga, or 2.3% of the total remittance sent by RSE workers. In total, the Tongan economy would gain 2.3% higher total remittances sent by the participants of the temporary migration programs, amounting to 1,528,405 pa’anga. This research contributes to the literature about remittance costs and the policymaking arena by providing the first quantitative evidence about the choice of remittance channels and updating the gain by switching RSPs in the PICs. The research on remittance costs in the PICs has used descriptive statistics or qualitative methods (Raithatha et al. 2021; NRBT 2017; Dayrit et al. 2016). After the report by the World Bank , there has been no research calculating the gain made by switching RSPs in the PICs, although the popularity of RSPs has shifted over time (World Bank 2017b:38). This research unpacks the key factors that affect the households’ choice of remittance channels with multinominal logit regression, and, using the latest data in Tonga, recalculates the gain made by switching to the lowest-cost RSP. ADBI Working Paper 1427 H. Maeda et al. 3 The rest of this paper proceeds as follows. The next section explains the important role of remittances in economic development, remittance costs in the PICs, and the issues around remittances in Tonga. The third section explains two types of data used in this study. The fourth section presents the results in three subsections: The remittance costs and the choice of RSPs, the household traits affecting the choice of remittance channels, and the potential gain by switching RSPs. The fifth section illustrates the potential barriers to switching to low-cost RSPs in Tonga by reviewing the existing evidence. The last section concludes the paper with a summary and policy implications. 2. BACKGROUND 2.1 Why is Reducing Remittance Costs Important for Developing Countries and the PICs? Migration and remittances improve household welfare and promote economic development in developing countries. The stock of international migrants has increased from 152 million to 280 million in the last three decades (UNDESA 2020). One driver of such active migration is a large income gap between developed and developing countries (World Bank 2023c: 2–4). People from developing countries migrate to secure better and more stable economic opportunities and send remittances back to families who remain in their home countries. Consequently, remittance inflows to low- and middle-income countries have grown from less than $50 billion in 1990 to $605 billion in 2021, which has sparked discussion on the development impacts of remittances. Existing evidence shows that remittances mainly contribute to poverty reduction through various channels, including increased income and consumption, the improvement of food security, human capital development, and narrowing the gender gap (World Bank 2023c: 127–143; Adams Jr and Cuecuecha 2010, 2013; Bouoiyour and Miftah 2016; Gibson and McKenzie 2014; Mansuri 2006; Mobarak, Sharif, and Shrestha 2020; Edwards 2023). An important source of friction is the high remittance costs. The Sustainable Development Goals aim to reduce remittance costs to less than 3% and eliminate the remittance corridor with average costs higher than 5% (United Nations n.d.). Remittance Prices Worldwide is a website about remittance costs managed by the World Bank, providing remittance cost data on 367 corridors as well as reports about remittance costs across the globe. According to the website, as of the first quarter of 2023, the global average remittance cost is 6.25% with only 39% of major remittance corridors meeting the global target of 5% (World Bank 2023b). Thus, it is imperative to make efforts to reduce remittance costs, and thereby remittances will further uplift the living standard of people across the globe. In this vein, Tonga and the other PICs attract special attention because of the long history of migration, the importance of remittances in their economic development, and the high remittance cost. The characteristics of small island developing states, such as geographical constraints and the small size of domestic markets pose difficulties for the PICs in promoting economic development and creating employment opportunities (World Bank 2017a). Within this context, migration mainly to Australia, New Zealand, and the United States (US) has, for many decades, provided opportunities to earn a stable income for the people from the PICs (Brown et al. 2006: 49–52; Doan et al. 2023: 2–3). ADBI Working Paper 1427 H. Maeda et al. 4 In recent years, the temporary migration programs by the governments of Australia and New Zealand have opened migration opportunities to low- and middle-skilled workers, becoming a crucial migration pathway for the people from the PICs (Doan et al. 2023:3–7). New Zealand launched Recognized Seasonal Employment (RSE) in 2007 while Australia established Seasonal Worker Program (SWP) in 2012 and the Pacific Labour Scheme (PLS) in 2019. In the near future, migration from the PICs to Australia and New Zealand is expected to grow because of the scale-up of temporary migration programs by both governments and the introduction of the Pacific Engagement Visa by the Australian government (Sharman 2022; Bedford 2023; Minister’s Media Centre 2023). The importance of migration and remittances in the economy of the PICs is salient. Bertram and Watters (1985) refer to some of the PICs as the MIRAB economy, emphasizing the importance of Migration, Remittances, Aid, and Bureaucracy in their economy. Remittances are the key source of foreign exchange in Samoa and Tonga as the two PICs have recorded two of the highest remittance-to-GDP ratios in the world ranging from 15% to 50%.2 Research also shows that remittances contribute to poverty alleviation and wealth accumulation in Fiji and Tonga and increase the expenditure on various goods, including food, housing, education, and community use in the PICs (Jimenez-Soto and Brown 2012; Brown et al. 2014; Brown and Jimenez 2008; Connell and Brown 2005). In Samoa and Tonga, remittances might be associated with financial development and economic growth (Jayaraman, Choong, and Kumar 2011). The temporary migration programs also improve the living standard of the PICs by increasing household income and consumption, upgrading the quality of dwellings, developing human capital, and shifting gender norms (Bailey 2015; Edwards 2023; Gibson and McKenzie 2014; World Bank 2017b). Having said that, High remittance costs dampen the development impacts of remittances in the region. According to Remittance Prices Worldwide, the average remittance cost in the four PICs (Fiji, Samoa, Tonga, and Vanuatu) was 8.68% in the fourth quarter of 2022 (Figure 1). When we include other PICs by using the data from Send Money Pacific,3 the average cost further increases to 10.4% as of February 2021 (Raithatha, Farooq, and Sharma 2021:19). However, this does not mean that migrants and households from the PICs can choose only high-cost money transfer options. Smart Remitter Target (SmaRT),4 the average remittance costs of the three lowest-cost and accessible RSPs for sending $200, shows that Fijian and Tongan migrants can send money at a cost of less than 5% to their home countries if they choose the lowcost providers (Figure 1). 2 The data are obtained from World Development Indicators. Find the data from: https://databank.worldbank.org/source/world-development-indicators. The series name is “Personal remittances, received (% GDP).” 3 Send Money Pacific is a platform about remittance costs established through a joint initiative by the Australia and New Zealand government. The program aims to promote a better engagement of migrants with remittance service providers and to enable migrants to choose the remittance service providers best-suited to their needs. The data section describes the information provided on the website in detail. url: https://sendmoneypacific.org/. 4 Smart Remitter Target (SmaRT) is one of the indicators to measure remittance costs in remittance corridors. This indicator accounts not only for costs, but also other user perspectives, such as accessibility and the speed of transaction. This is achieved by dropping RSPs that do not satisfy four criteria. To obtain more information about SmaRT, go to the methodology paper from the link. url: https://remittanceprices.worldbank.org/sites/default/files/smart_methodology.pdf. ADBI Working Paper 1427 H. Maeda et al. 11 Figure 4: Remittance Costs in the New Zealand–Tonga Corridor (%) Source: The graph is created by the author from the audit data. This result indicates that most Tongan households pay much higher remittance costs due to their use of high-cost RSPs. In the Australia–Tonga corridor, both household and worker surveys show that Moneygram is the most popular RSP, providing a lowcost online transfer service (Figure 2), but 44.39% of households using Moneygram to receive money in cash (Table A5 in Appendix). In the other popular RSP, Western Union, costs are 5.6 to 6.5 percentage points higher than Ria. Having said that, this observation confirms a gradual and favorable shift in the choice of RSPs. Western Union has lost its share among Tongan migrants since nearly 98% of SWP workers from the PICs including Tonga sent remittances through Western Union from 2015 to 2017 (World Bank 2017b: 38). Instead, there is a growth in new and low-cost RSPs, such as Ave Pa’anga Pau and KlickEx, of more than 10% if their shares are combined. ADBI Working Paper 1427 H. Maeda et al. 12 In the New Zealand–Tonga corridor, such a shift is more outstanding. Some 30.7% of households and 32.4% of workers reported they used the lowest-cost RSP, Ave Pa’anga Pau (Figure 2). However, the household survey shows that 54.7% of households receive money through the high-cost RSPs, Moneygram, and Western Union, meaning that they pay around 5.8 to 9.7% in remittance costs. However, the share of the two MTOs decreases to 27.8% in the worker survey. Given the high share of the high-cost RSPs, the next question is “What makes Tongan households choose high-cost rather than low-cost RSPs?” The worker survey indicates that they choose a high-cost RSP because it is the easiest to use compared to other services and migrants are less reactive to the cost advantage of remittance services. The survey asked binary questions about the reason why they chose remittance channels (online transfers, over-the-counter transfers, mobile money, through friends, and others). Migrants answered yes or no on multiple factors: cost, speed, ease of use, awareness, word of mouth, availability in home countries, availability in host countries, safety, and bank account ownership.8 Figure 5 shows that migrants choose remittance channels because of three factors: ease of use (72.6%), cost (43.5%), and speed (33.7%). The users of Western Union and cash transfers with Moneygram have similar characteristics. Figure 6 shows that 79.9% of them choose remittance channels because of the ease of use, followed by cost (38.0%) and speed (34.2%). It is thus reasonable that the Tongan households do not use Ria because it only has one agent in Tongatapu and it is new to Tongan households. On the other hand, the users of Ave Pa’anga Pau pay more attention to cost (60.9%) although ease of use is still the most popular reason for using it, at 67% (Figure 7). A gap in the percentage of cost is suggestive that those who keep using Moneygram and Western Union might not be incentivized by cost advantages, but stick to the favorable MTOs because they prioritize the ease of use in choosing RSPs. Figure 5: The Reason Why Tongan Migrants Choose Remittance Channels (%) Source: The graph is created by the author from the PLMS data. 8 The questions and labeling are listed in the Appendix. ADBI Working Paper 1427 H. Maeda et al. 13 Figure 6: The Reason why Tongan Migrants Who Use Western Union and Moneygram Choose Remittance Channels (%) Source: The graph is created by the author from the PLMS data. Figure 7: The Reason Why Tongan Migrants Who Use Ave Pa’anga Pau Choose Remittance Channels (%) Source: The graph is created by the author from the PLMS data. 4.2 Household Characteristics and the Choice of Remittance Channels To understand how household characteristics affect the choice of remittance channels, we estimate a multinomial logit regression model, following the existing literature about the choice between the formal and informal channels (Karafolas and Konteos 2010; Siegel and Lücke 2013; Kosse and Vermeulen 2014; Amuedo-Dorantes and Pozo 2005 Amuedo-Dorantes et al. 2005; MacIsaac 2023). This paper has two differences from the modeling in previous literature. First, the previous papers have controlled ADBI Working Paper 1427 H. Maeda et al. 14 the three stages of the choice of remittance channels—remitter in the destination countries, MTOs, and households in the destination countries—by applying the first-to-last mile framework (Hernandez-Coss 2005). However, our research examines only the last mile of decision-making by using information about the households in Tonga because the face-to-face household survey and the worker survey are not linked well.9 Second, this paper calculates the average marginal effects rather than the relative risk ratio to interpret the effects of explanatory variables because we want to independently understand the factors associated with remittance channels including the baseline outcome. The outcome variable is a polychotomous variable with four remittance channels: 1. online transfer through bank, 2. online transfer through MTOs, 3. over-the-counter transfer through MTOs, and 4. mobile money. The question in the PLMS contains six categories, but we drop two remittance channels, “Through friends” and “Others,” because only a very small share of the respondents uses these channels. The baseline is the over-the-counter transfer through MTOs. Following Amuedo-Dorantes and Pozo (2005), we assume that households decide on the MTO to maximize the utility: 𝑈() =-𝛽) *𝑥() +-𝜀() (2) where 𝑈() represents the utility of households 𝑖 when they choose remittance channel 𝑗, and 𝑥() is the vector of explanatory variables. The probability that the household 𝑖 uses the 𝑗th remittance channel is given by: 𝑃 ) (𝑌(= 𝑗) = exp:𝛽) *𝑥(); 1 + ∑exp:𝛽+ *𝑥(); , + . /,+12 (3) where 𝛽) is the coefficient estimated by maximum likelihood. We attempted to mainly control four factors when modelling household decisionmaking on the remittance channels. The first factor is the chance of finding or searching for information about remittance costs. Qualitative research suggests that word of mouth plays a vital role in switching to online financial services. Group discussions in seven African countries suggest that recommendations from friends and family, both via in-person and SNS, encourage people to change transfer methods from cash to online (FSD Africa 2018: 30–31). Word of mouth also helps to build trust in digital financial services like mobile money, which leads to regular usage of online services (Cohen 2014: 13). Quantitative research also supports the importance of social networks in choosing remittance channels. Amuedo-Dorantes and Pozo (2005) show that Mexican migrants who have Mexican friends in the host city in the US are more likely to choose formal channels rather than informal channels in the US–Mexico corridor. We use two variables to account for this channel. The first proxy is a categorical variable that classifies the participation status of the temporary migration programs in Australia and New Zealand. The variable controls the current chance of being exposed to new information and experiences in temporary migration. Temporary migrants can get new information about remittance costs through migrant communities and they might voluntarily search for information about those costs. The variable equals 0 if 9 In Africa, interviews with recipients of remittances reveal that the recipients of remittances play a major role in deciding the remittance channels (FSD Africa 2018: 29). ADBI Working Paper 1427 H. Maeda et al. 15 households do not have temporary migrants, 1 if households have temporary migrants participating in the program for the first time, and 2 if households have repeated migrants who have participated in the program multiple times. In the regression, we use the households without temporary migrants as the baseline. The second proxy is a binary variable for living experience in Australia and New Zealand in the past. It takes the value of 1 if any household member has lived more than one month in Australia or New Zealand in the past, and 0 otherwise. In a similar idea, it accounts for the exposure to information about remittance costs in the past and potential information flows, thanks to the connection they built in Australia and New Zealand. The second factor is education level. Even without information from other people, a better-educated individual will be more likely to have and be able to use information about remittance costs because of better financial and digital literacy. Previous research shows that migrants who completed secondary education are more likely to choose formal channels compared to those who do not graduate from secondary school because of a better comprehension of the risk of informal channels and the remittance markets in both destination and origin countries (Amuedo-Dorantes and Pozo 2005; Kosse and Vermeulen 2014; Siegel and Lücke 2013). To account for the education level of households, we use the share of individuals who completed secondary school in the household members aged 18 or over. The third factor is access to remittance channels. In Africa, the convenience of cash transfer and habitual behavior relates to the choices of remittance channels (FSD Africa 2018: 37–41). When the agents of high-cost MTOs are located close to the recipient’s home, households can stick to using cash transfers even after they are aware of online services with a lower remittance cost. Hernandez-Coss (2005) finds that improving access to the formal channels expanded the take-up of formal channels in the US–Mexico corridor. Kosse and Vermeulen (2014) mention that living in urban areas and the density of ATMs affect the migrants’ choice of remittance channels. In our study, we try to account for this channel with the dummy variable for residential islands, which takes the value of 1 if the household resides in Tongatapu, and 0 otherwise. The data about the travel costs to the agent of MTOs are not available, but residential islands can substitute information about the travel costs and the geographical differences in financial institutions since the number of access points of MTOs and banks differs across the four residential islands (NRBT 2017: 40–41). Another key factor is the amount and frequency of remittances because the choice of remittance channels can depend on diverse needs and the role of remittances in household finance. Previous research includes the value of the remittances as an explanatory variable, but the effect depends on the context. Kosse and Vermeulen (2014) argue that the amount of the remittance is an important factor because a higher amount of remittance decreases the probability of remitting through informal intermediaries, ATM withdrawal, and carrying cash by hand among the migrants staying in the Netherlands. Amuedo-Dorantes and Pozo (2005) find that a higher amount of remittances is associated with the use of unspecified remittance channels rather than bank or MTOs for Mexican migrants in the US. Siegel and Lücke (2013) find that there is no association between the amount of the remittance and the choice of remittance channels in Moldova. In our study, we use a monthly remittance received as an explanatory variable. ADBI Working Paper 1427 H. Maeda et al. 16 Different usages of remittances affect their frequency. In Tonga, households use remittances mainly for three purposes: daily needs, including church donations (Bedford, Bedford, and Nunns et al. 2020: 68–70; Maeda and Edwards 2023), education and the improvement of their dwelling (Bedford, Bedford, and Nunns 2020; World Bank 2017b: 47–49), and special events such as funerals and Christmas from international remittances (Fifita 2021; Connell and Brown 2005: 30–37). Migrants from households who use remittances for daily expenditures need to remit a small amount more frequently than other migrants. In contrast, migrants who only need to remit for emergencies or the festive expenditures send a large amount of money occasionally rather than on a regular basis. The PLMS asks about the frequency of remittances with five categories: 1. Monthly or more frequent, 2. Every two months, 3. Every three months, 4. Every four to six months, and 5. Only on special occasions. For analytical purposes, we modified this categorical variable into three categories. The first category is “Monthly or more frequent,” which accounts for households who use remittances for daily needs. The second category is “Every two to six months,” which accounts for households who spend remittances on larger, but less frequent spending, like housing or education. The third category is “Only on special occasions,” which accounts for households who receive remittances only for emergencies and festive occasions. In the regression, we use the third category as the baseline. We include the other two explanatory variables in our model. Monthly saving per head captures the use of bank transfers because the recipient households do not necessarily deposit remittances in their bank account if they send and receive remittances by cash. The number of adults relates to the cost of receiving remittances by cash. When people receive remittances at the MTO’s agent, they have to queue for a long time. A greater number of adults reduces the costs of queuing at the agent because it affects the time allocation of household members to a lesser extent by distributing care work and chores to those household members who are not formally employed. Table 3 presents the summary statistics. Households using online transfers or mobile money receive more remittances than those using cash transfers. Households receiving money through online bank transfers possess a greater amount of savings and have a richer migration experience than the other groups. Mobile money users receive remittances most frequently, on average, followed by households using cash transfers, but the number of adults in the households is the least among the four groups. Households receiving money by cash transfers are more likely to live in the outer islands and less likely to have rich migration experiences. The estimation results are shown in Figure 8. They reveal that three key household factors exist in the choice of remittance channels. The exposure to information about accessibility to MTOs, remittance costs, and frequency of remittances would urge households to use online transfers rather than cash transfers. ADBI Working Paper 1427 H. Maeda et al. 17 Table 3: Summary Statistics of Explanatory Variables Online/ Bank Online/ MTOs Cash Mobile Money Living in Tongatapu (Yes = 1) .862 .873 .748 .926 (.346) (.334) (.434) (.264) Temporary migration status No migrants (Yes = 1) .198 .339 .510 .407 (.400) (.474) (.500) (.496) First time (Yes = 1) .190 .248 .209 .204 (.394) (.433) (.407) (.407) Repeated (Yes =1) .612 .413 .281 .389 ‘.198) (.339) (.510) (.407) Past migration (Yes = 1) .605 .652 .551 .551 (.276) (.268) (.354) (.333) Frequency of remittances Monthly (Yes = 1) .698 .640 .581 .500 (.461) (.481) (.494) (.505) 2–6 months (Yes = 1) .138 .214 .140 .278 (.346) (.411) (.348) (.452) On request (Yes = 1) .164 .146 .278 .222 (.372) (.354) (.449) (.420) Monthly remittances (Pa’anga) 876.121 900.642 566.577 740.340 (1,031.385) (1,252.912) (1,021.680) (1,099.312) Monthly saving per head (Pa’anga) 174.828 71.600 79.196 73.229 (844.35) (228.183) (278.038) (243.497) Number of adults 3.259 3.295 3.339 2.685 (1.818) (1.677) (1.739) (1.412) The share of adults who completed secondary education .322 .344 .317 .303 (.311) (.293) (.299) (.305) MTO = money transfer operator. Figure 8: Average Marginal Effects ADBI Working Paper 1427 H. Maeda et al. 18 The result of the residential islands gives a mixed picture of the choice around remittance channels. We expected that living in Tongatapu would be associated with a higher probability of using over-the-counter transfers through MTOs compared to other islands, because a better access to financial access points of MTOs can induce households to stick to using over-the-counter transfers (NRBT 2017: 40–41; FSD Africa 2018: 37–41). However, the result is the opposite, showing that households living in Tongatapu have a lower probability of using over-the-counter transfers through MTOs by 17.8 percentage points, while having a higher probability of using online transfers through MTOs by 11.0 percentage points. Living in Tongatapu also leads to a higher take-up of mobile money by 5.2 percentage points compared to other islands. This is probably because of a higher rate of mobile phone ownership and mobile data usage in Tongatapu than in the other three islands (Tonga Statistics Department 2022a: 10; Tonga Statistics Department 2022b: 137–138). The current participation in the temporary migration programs makes cash transfers a less attractive option for remittances but leads to a higher probability of using online transfers through banks. Households with a temporary migrant who participated in the temporary migration programs for the first time have an 8.9 percentage point lower probability of using the over-the-counter transfers compared to those households without any participants of the temporary migration programs while keeping other factors constant. When households have repeated participants in the temporary migration programs, they are 18.2 percentage points less likely to use the over-the- counter transfers. On the other hand, households with a participant of the temporary migration programs for the first time have a 5.2 percentage point higher probability of using online transfer through banks. Those with repeated migrants have a higher probability of using online transfers through banks by 15.5 percentage points. Such changes occur probably because households become aware of a relatively higher cost of cash transfers as the audit data show the price advantage of online bank transfers over cash transfers through MTOs. Past migration experience provides opportunities to switch to online transfers. Households with living experiences in Australia and New Zealand for more than one month have a lower probability of using cash transfers by 15.6 percentage points, whereas they have a higher probability of receiving remittances with online transfers through MTOs by 20.6 percentage points. For the remitting characteristics, frequent remittance drives a shift from cash transfers to online transfers within MTOs. Households who send remittances every month or more frequently have a lower probability of using over-the-counter transfers through MTOs by 7.0 percentage points compared to those households who receive remittances on request at the significance level of 10%. They also have a higher probability of using online transfers through MTOs by 8.8 percentage points. Households who send remittances every two to six months show a larger effect, with a 23.4 percentage point lower probability of using over-the-counter transfers through MTOs and a 21.8 percentage point higher probability of using online transfers through MTOs. Unlike the frequency, we find that the amount of remittances is not associated with the choice of remittance channels. The coefficients of a monthly remittance received per head are statistically insignificant at a 10% level or the magnitude of the coefficient is small. In contrast to the previous literature, we find that education level is not an important factor. The coefficients for the education level of households are statistically insignificant at the 10% level. As for the other two covariates, households with a greater number of adults have a lower probability of using mobile money by 1.6 percentage points, but this does not relate to a probability of using cash transfers ADBI Working Paper 1427 H. Maeda et al. 19 through MTOs. The coefficients of the monthly saving per head are statistically insignificant at a 10% level or the magnitude of the coefficient is too small. 4.3 Simulation: How Much Would Be the Gain Made by Switching RSPs? When policymakers try to intervene in the choice of remittance channels and RSPs, one important question is how much the intervention would benefit the affected citizens. We calculate the potential gain by switching from the two popular MTOs, Moneygram and Western Union, to the lowest cost RSPs, Ria in the Australia–Tonga corridor and Ave Pa’anga Pau in the New Zealand–Tonga corridor. Data for households with temporary migrants are used to calculate the gain because we can identify the remittance corridor that the household belongs to. Figure 9 shows three steps of our simulation. First, we derive the gain per transaction of 200 AUD/NZD by calculating the difference between the cost of the lowest-cost RSP and the two popular MTOs in both local currency and the percentage of remittance costs. We use data for both cash and online transfers for Moneygram and Western Union. Second, we derive the average remittances sent in six months by Tongan temporary migrants in Australia and New Zealand from the worker survey, respectively. On average, Tongan temporary migrants in Australia send A$8,116.5 (12,695.7 Pa’anga), and RSE workers in New Zealand send NZ$9,265.3 (13,410.2 Pa’anga) in six months. We obtained the multiplier, 40.58 for Australia and 46.33 for New Zealand, by dividing the average remittances by A$/NZ$200. These multipliers are then used to calculate the per-household gain in six months made by switching providers. Third, we calculated the gain of the Tongan economy by multiplying the gain per household by the number of temporary migrants in Australia and New Zealand on 30 April 2023. There were 3,692 and 1,386 participants of the temporary migration programs in Australia and New Zealand, respectively (Department of Home Affairs 2023; Ministry of Business, Innovation and Employment 2023). Figure 9: Simulation Study ADBI Working Paper 1427 H. Maeda et al. 20 Table 4 shows the simulation results based on data from SMP. Households who use online transfers through Moneygram would gain 0.88% if they switched to Ria. If households switched from cash transfers with Moneygram to Ria, they would gain 3.64%. Those households who use Western Union could gain a larger amount, saving 5.62% for online transfers and 6.47% for cash transfers. If all households who use Moneygram and Western Union switched to Ria, the Tongan economy would be able to receive 1,103,809 Pa’anga more in a six-month period, amounting to 2.4% of the remittances sent by Tongan temporary migrants in Australia. Table 4: The Gain Made by Switching Based on Send Money Pacific Australia–Tonga New Zealand–Tonga WU (Online) WU (Cash) MG (Online) MG (Cash) WU (Online) WU (Cash) MG (Online) MG (Cash) Total Remittance costs (%) 7.38 8.23 2.64 5.4 7.39 7.53 9.73 5.84 Gain by switching (%) 5.62 6.47 0.88 3.64 4.07 4.20 6.40 2.51 Gain in 6 months (TOP and %) 1,103,809 TOP (2.4%) 424,596 TOP (2.3%) 1,528,405 TOP (2.3%) MG = Moneygram, WU = Western Union. In the New Zealand–Tonga corridor, compared to the case in Australia, households who send and receive money through Moneygram would be able to save a greater amount of money by switching to Ave Pa’anga Pau. Households who use online transfers could save 6.4% of remittances when they send 200 NZD, while those who use cash transfers would gain 2.51%. Those households who use Western Union could save 4.07% for online transfers and 4.20% for cash transfers. If all these households switched to Ave Pa’anga Pau, the Tongan economy would be able to save 424,596 Pa’anga, amounting to 2.3% of the remittances sent by RSE workers. In total, the gain of the Tongan economy would equal 1,528,405 Pa’anga which is 2.3% of the estimated total remittances sent by temporary migrants in the two destination countries. For the robustness check, we did the same simulation analysis with data from SP. Table 5 shows that the results are slightly different, but still similar. In the Australia–Tonga corridor, the total gain by switching becomes 0.1 percentage points larger than the main result. This is because the gain of the main channel, Moneygram, becomes larger compared to the SMP results. The gain for online transfers is 0.42 percentage points higher at 1.3% and the gain for cash transfers is 0.32 percentage points higher at 3.96%. For households who send and receive money with Western Union, households who use online transfers would gain 5.73% while households who use cash transfers would save 5.83%. In the New Zealand–Tonga corridor, all estimated gain is lower compared to the results from SMP data. Those who use online and cash transfers of Moneygram would save 5.77% and 1.95%, respectively. Western Union users could save 4.07% if they use online transfers and 4.20% if they use cash transfers. In total, the gain by switching is 0.1 percentage points higher than the SMP results at 2.4%, amounting to 1,572,246 pa’anga. ADBI Working Paper 1427 H. Maeda et al. 27 Tonga Statistics Department 2022b. Tonga 2021 Census of Population and Housing Volume 1: Basic Tables, Nuku’alofa, https://tongastats.gov.to/census-2/ population-census-3/census-report-and-factsheet/ (accessed 2 August 2023). UNDESA. 2020. International Migrant Stock. https://www.un.org/development/ desa/pd/content/international-migrant-stock (accessed 21 July 2023). United Nations. (n.d.) Goal 10 | Reduce Inequality Within and Among Countries. https://sdgs.un.org/goals/goal10 (accessed 11 July 2023). World Bank. 2017a. Pacific Possible: Long-term Economic Opportunities and Challenges for Pacific Island Countries. Washington, DC: World Bank. doi:10.1596/28135. ———. 2017b. Maximizing the Development Impacts from Temporary Migration. Washington, DC: World Bank. doi:10.1596/29622. ———. 2023a. Remittance Prices Worldwide. https://remittanceprices.worldbank.org/ data-download (accessed 2 August 2023). ———. 2023b. Remittance Price Worldwide Quarterly Issue 45 March 2023. Washington, DC: World Bank Group. https://remittanceprices.worldbank.org/ sites/default/files/Remittance Prices Worldwide_main_report_and_annex_ q123_final.pdf (accessed 11 July 2023). ———. 2023c. World Development Report 2023: Migrants, Refugees, and Societies. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/ 10986/39696 (accessed 20 May 2023). ADBI Working Paper 1427 H. Maeda et al. 28 APPENDIX 1: ISSUES IN THE PLMS DATA SET The combined data set of the face-to-face survey and phone-based survey is not yet ready to be used, at least in researching remittance channels and MTOs, due to the following two issues. First, the information about the main bank or MTO stored in the combined data set is not consistent with the information stored in the initial data set only for the face-to-face survey. In the initial data set, as Tables A2 and A3 present, Moneygram, Western Union, and Ave Pa’anga Pau are the three main MTOs. However, Table A15 in the Appendix shows that ANZ is the most popular remittance service provider while less than 15% of households use Moneygram or Western Union. Also, the table below shows that the two data sets report a different main MTO or bank for the household with the same pairing ID. w_id Initial Data Combined Data 1101205 Western Union ANZ 1101190 Western Union ANZ 1101191 Bank of South Pacific ANZ 1101018 Moneygram ANZ 1101014 Moneygram ANZ Second, the information about the main remittance channel stored in the combined data is not clean. The data must be a categorical variable with five categories, but a simple tabulation shows that it contains many unknown numbers. 10.04. Which Is the Most Usual Channel Through Which Your Household Receives the Money? Freq. % ONLINE TRANSFERS 93 10.15 OVER-THE-COUNTER TRANSFERS 116 12.66 MOBILE WALLET 5 0.55 CASH THROUGH A THIRD PARTY 1 0.11 OTHER (specify) 1 0.11 200 2 0.22 250 1 0.11 300 4 0.44 400 4 0.44 450 2 0.22 500 10 1.09 600 8 0.87 620 1 0.11 700 3 0.33 800 6 0.66 900 5 0.55 1,000 37 4.04 1,200 7 0.76 1,259 1 0.11 1,280 1 0.11 continued on next page ADBI Working Paper 1427 H. Maeda et al. 29 Appendix 1 table continued 10.04. Which Is the Most Usual Channel Through Which Your Household Receives the Money? Freq. % 1,300 3 0.33 1,390 1 0.11 1,400 3 0.33 1,450 1 0.11 1,500 14 1.53 1,600 1 0.11 1,800 10 1.09 1,950 1 0.11 2,000 38 4.15 2,100 2 0.22 2,400 7 0.76 2,500 15 1.64 2,600 3 0.33 2,700 1 0.11 2,800 4 0.44 3,000 72 7.86 3,300 2 0.22 3,400 1 0.11 3,500 15 1.64 3,600 8 0.87 3,700 2 0.22 3,800 1 0.11 4,000 66 7.21 4,100 1 0.11 4,200 1 0.11 4,400 1 0.11 4,500 10 1.09 4,600 1 0.11 4,800 6 0.66 5,000 53 5.79 5,400 3 0.33 5,500 7 0.76 5,600 2 0.22 5,800 3 0.33 6,000 58 6.33 6,500 6 0.66 6,800 1 0.11 6,900 1 0.11 7,000 23 2.51 7,200 8 0.87 7,500 1 0.11 8,000 16 1.75 8,400 3 0.33 8,600 1 0.11 9,000 4 0.44 continued on next page ADBI Working Paper 1427 H. Maeda et al. 30 Appendix 1 table continued 10.04. Which Is the Most Usual Channel Through Which Your Household Receives the Money? Freq. % 9,600 4 0.44 9,650 1 0.11 10,000 30 3.28 10,600 1 0.11 11,000 5 0.55 12,000 21 2.29 12,400 1 0.11 13,000 1 0.11 14,000 1 0.11 14,200 1 0.11 14,300 1 0.11 14,400 1 0.11 15,000 17 1.86 16,000 1 0.11 16,800 1 0.11 17,900 1 0.11 18,000 4 0.44 19,000 2 0.22 20,000 9 0.98 21,600 2 0.22 22,000 1 0.11 24,000 3 0.33 25,000 5 0.55 30,000 5 0.55 35,000 3 0.33 40,000 5 0.55 Total 916 100.00 Questions about the Reason Why Temporary Migrants Choose Remittance Channels The worker survey asks, “Why did you use the main remittance channel?” I modified the label of the variables below when creating the graphs and tables. Questions in the Survey Labeling in the Graph a. Cheapest Cost b. Fastest Speed c. Easiest to use Ease of Use d. Only channels I know of Awareness e. Many other work mates use this channel Word of Mouth f. Only channel available at home Available_home g. Only channel available in your workplace area Available_host h. Do not have an account on bank Bank Account ADBI Working Paper 1427 H. Maeda et al. 31 The Average Remittances in 6 Months Sent by Temporary Migrants We calculated the average remittances sent by temporary migrants in Australia and New Zealand in the following procedures. The worker survey asked the amount and frequency of remittances: “How often do you send money to your country?” and “Normally, how much did you send to your country each time?” Depending on the frequency of remittances, we multiply the amount of remittances per each time by a constant number to get the estimate of average remittances in six months. For instance, the average remittance in six months is 500 ×24 =12,000-𝐴𝑈𝐷 , if households send a weekly remittance of 500 AUD. When calculating the average, the migrants who belong to the top 5 percentile are dropped to remove an excessive amount of remittances such that they send 2,000 AUD weekly. This happens because the PLMS allows the workers to report the amount of remittances they sent last time if the amount of remittances varies over time. Table A6 presents the average remittances in six months. The threshold is 24,000 AUD which equals a weekly remittance of 1,000 AUD. Frequency A Constant Number (Multiplier) Weekly 24 Every 2 weeks 12 Once a month 6 Once every 2–3 months 3 Once every 4 months or more 1 ADBI Working Paper 1427 H. Maeda et al. 32 APPENDIX 2: SUPPLEMENTAL FIGURES AND TABLES Table A1: The Main Remittance Channels Reported by Households with Temporary Migrants in Australia or New Zealand Channel Freq. % Online/Bank 85 16.67 Online/MTOs 193 37.84 Counter/MTOs 198 38.82 Mobile wallet 31 6.08 Other 3 0.59 Total 510 100.00 Table A2: The Main MTOs Reported by Households with Temporary Migrants in Australia MTOs Freq. % Bank of South Pacific BSP 8 2.22 ANZ 2 0.55 TDB 28 7.76 MBF 1 0.28 Digicel 15 4.16 IMEX Money Transfer 9 2.49 KlickEX 2 0.55 Melie Mei Langi 2 0.55 MoneyGram 196 54.29 Nikua Money Transfer 2 0.55 Pacific Ezy 2 0.55 Western Union 72 19.94 Other 22 6.09 Total 361 100.00 Table A3: The Main MTOs Reported by Households with Temporary Migrants in New Zealand MTOs Freq. % Bank of South Pacific BSP 1 0.67 TDB 46 30.67 Currency online 1 0.67 Digicel 14 9.33 IMEX Money Transfer 1 0.67 Melie Mei Langi 1 0.67 MoneyGram 50 33.33 Nikua Money Transfer 1 0.67 Pacific Ezy 1 0.67 Western Union 32 21.33 Other 2 1.33 Total 150 100.00 ADBI Working Paper 1427 H. Maeda et al. 33 Table A4: The Main MTOs Reported by Temporary Migrants in Australia Main Freq. % ANZ 11 2.03 Westpac 5 0.92 Western Union 75 13.81 Moneygram 285 52.49 KlickEx 5 0.92 Ria money transfer 11 2.03 Wantok money 2 0.37 TDB 54 9.94 Island Flexi 36 6.63 Other (specify) 59 10.87 Total 543 100.00 Table A5: The Main MTOs Reported by Temporary Migrants in New Zealand Main Freq. % ANZ 7 3.43 Kiwi bank 2 0.98 BNZ 3 1.47 Westpac 14 6.86 Western Union 31 15.20 Moneygram 25 12.25 KlickEx 9 4.41 KlickEx pacific 1 0.49 Wantok money 1 0.49 TDB 66 32.35 Island Flexi 2 0.98 Other (specify) 43 21.08 Total 204 100.00 Table A6: The Channels Used by Moneygram Users in the Australia–Tonga Corridor Channel Freq. % Online/MTOs 109 55.61 Counter/MTOs 87 44.39 Total 196 100.00 ADBI Working Paper 1427 H. Maeda et al. 34 Table A7: The List of Other (specify) in the Worker Survey Other (Specify) Talafi paanga pe fanga kii talafi ofi ange kanautolu mobile money Freedom Pacific koloa fakatau Lupe Hume koloa fakatau facebook Koloa Fakatau koloa fakatonga fb page Talafi paanga Mosimani mobile money MOSIMANI mobile money pacific eo transfer no money Noah mobile money Mosimani mobile money Koloa Fakatau - Ma’ufanga Koloa Fakatau Malakai Huni - Noah Palu Transfer world remit Imex Mosimani Mosimani AMLS money transfer Mosimani louena transfer Mosimani world money transfer Rowena Mosimani Building BSP Ria n Noah MOSIMANI Mosimani Kautaha Noa Rowena money transfer Mosimasi Money Transfer Rowena Mosimani Money Transfer personal account of a co workers Koloa Fakatau manatu ofa Noah Money Transfer mobile money digi Society Federation Money Transfer mosimani transfer Rowena Money Transfer Palu Money Transfer Noah Money Transfer Vavau Palu Money Transfer Siasi Tonga Houeiki Rowena money transfer Tonga Post private Mosimani NOAH Palu Money Transfer nasita dvd Palu Noah Melie Mei Langi Mosimani Money Transfer Meliemeilangi Mosimani Money Transfer Talafi Paaga Siasi STT Mosimani Money Transfer Mana Chinese Shop post office Chinese Shop Money Transfer pangike fakalakalaka dont know regional australian bank koloa fakatonga Mosimani Building Koloa Fakatau pangike fakalakalaka Koloa Fakatau mobile money Koloa Fakatau online Koloa Fakatau doesn’t have a fee for the transfer Koloa Fakatau online Koloa Fakatau Koloa Fakatau Koloa Fakatau Noah Koloa Fakatau LOUENA FINANCE Koloa Fakatau koloa fakatau Mosimani mosimani Koloa Fakatau mobile money Mobile Money ADBI Working Paper 1427 H. Maeda et al. 35 Table A8: The Average Remittance Costs Across Different MTOs in the Australia–Tonga Corridor (Send Money Pacific) MTOs Mean Standard Deviation ANZ 3.471 .300 Ave Pa’anga Pau 2.849 .574 KlickEX 5.086 1.168 Moneygram (Cash) 5.402 .223 Moneygram (Online) 2.639 .227 NAB 23.444 .383 Ria 1.76 .322 Western Union (Cash) 8.231 .379 Western Union (Online) 7.381 .389 Table A9: The Average Remittance Costs Across Different MTOs in the Australia–Tonga Corridor (Saver Pacific) MTOs Mean Standard Deviation ANZ 3.626 1.018 Ave Pa’anga Pau 2.628 .145 Moneygram (Cash) 5.169 .204 Moneygram (Online) 2.509 .207 OFX 15.727 1.167 Ria 1.209 .345 Wantok 5.549 .13 Western Union (Cash) 7.034 .12 Western Union (Online) 6.934 .12 Table A10: The Average Remittance Costs Across Different MTOs in the New Zealand–Tonga Corridor (Send Money Pacific) MTOs Mean Standard Deviation ANZ 4.523 .474 ASB 21.676 .494 Ave Pa’anga Pau 3.329 .591 KlickEX 5.411 .742 Moneygram (Cash) 5.843 .295 Moneygram (Online) 9.731 .285 Wantok 5.676 .464 Western Union (Cash) 7.533 .414 Western Union (Online) 7.394 .413 ADBI Working Paper 1427 H. Maeda et al. 36 Table A11: The Average Remittance Costs Across Different MTOs in the New Zealand–Tonga Corridor (Saver Pacific) MTOs Mean Standard Deviation ANZ 5.736 .083 Ave Pa’anga Pau 3.529 .414 Kiwi Bank 15.482 .086 Moneygram (Cash) 5.478 .061 Moneygram (Online) 9.299 .059 Wantok 6.024 .426 Western Union (Cash) 7.467 .12 Western Union (Online) 7.328 .119 iMEX 3.849 .072 Table A12: The Estimated Results of Multinominal Logit Regression (1) (2) (4) Variables Online Bank Online MTOs Mobile Money Living in Tongatapu (Yes = 1) 0.517 0.753*** 1.768*** (0.335) (0.231) (0.573) Temporary migration status (No migrants is the baseline) First-time 0.827** 0.312 0.103 (0.339) (0.212) (0.415) Repeated 1.722*** 0.490** 0.465 (0.297) (0.207) (0.367) Past migration (Yes = 1) 0.264 1.004*** –0.241 (0.376) (0.247) (0.473) Frequency of remittances (On request is the baseline) Monthly 0.288 0.463** –0.434 (0.306) (0.212) (0.377) 2–6 months 0.508 1.252*** 0.939** (0.398) (0.257) (0.437) Monthly remittances (Pa’anga) –0.000 0.000 0.000 (0.000) (0.000) (0.000) Monthly saving per head (Pa’anga) 0.000** –0.000 0.000 (0.000) (0.000) (0.000) Number of adults –0.044 –0.021 –0.316*** (0.068) (0.046) (0.106) The share of adults who completed the secondary education 0.251 0.239 –0.094 (0.380) (0.258) (0.505) Constant –3.024*** –2.390*** –2.725*** (0.600) (0.344) (0.766) Observations 941 941 941 Pseudo-R-squared 0.0679 0.0679 0.0679 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.