Measuring remittances through surveys: Methodological and conceptual issues for survey designers and data analysts

DEMOGRAPHIC RESEARCH

VOLUME 31, ARTICLE 41, PAGES 12431274

PUBLISHED 21 NOVEMBER 2014

http://www.demographic-research.org/Volumes/Vol31/41/

DOI: 10.4054/DemRes.2014.31.41

Research Article

Measuring remittances through surveys:

Methodological and conceptual issues for survey

designers and data analysts

Richard P.C. Brown

Sonja Fransen

Jørgen Carling

Melissa Siegel

This open-access work is published under the terms of the Creative Commons

Attribution NonCommercial License 2.0 Germany, which permits use,

reproduction & distribution in any medium for non-commercial purposes,

provided the original author(s) and source are given credit.

See http:// creativecommons.org/licenses/by-nc/2.0/de/

Table of Contents

Introduction

1244

Objectives of remittances surveys

1245

2.1

Mapping remittances

1246

2.2

Investigating channels and costs of remittances

1246

2.3

Analysing determinants and motives

1247

2.4

Assessing the effects and the impact of remittances

1248

Meanings and forms of remittances

1250

3.1

Does it matter whether value is transferred in the form of money?

1251

3.2

Does it matter whether the money is sent by a migrant?

1251

3.3

Does it matter whether the money is sent between relatives?

1251

3.4

Does it matter whether the money is transferred to the sender’s

country of origin?

1252

3.5

Does it matter whether the money becomes the recipient’s

property?

1252

3.6

Does it matter whether the transfer is unrequited?

1253

3.7

Is it important to also capture possible reverse transfers?

1253

3.8

Does it matter whether money is transferred from one place to

another?

1254

3.9

Does it matter whether remittances are international or internal?

1254

Research design and sampling

1256

4.1

Individuals and households

1256

4.2

Senders and recipients

1259

Framing of questionnaire items

1259

5.1

Remittance frequency

1260

5.2

Remittance amounts

1260

5.3

Remittance senders and recipients

1261

5.4

Transfer mechanisms

1261

5.5

Contextualizing information

1262

Reliability and sensitivity

1264

6.1

Knowledge, encoding, and recall

1264

6.2

Sensitivity and bias

1265

Conclusions

1267

Acknowledgements

1268

References

1269

Demographic Research: Volume 31, Article 41

Research Article

http://www.demographic-research.org 1243

Measuring remittances through surveys: Methodological and

conceptual issues for survey designers and data analysts

Richard P.C. Brown

Jørgen Carling

Sonja Fransen

Melissa Siegel

Abstract

BACKGROUND

The accurate measurement of migrant remittances is central to understanding both the

dynamics and the consequences of migration. Empirical studies rely on survey data

about remittances, which are difficult to collect in a precise and reliable form.

OBJECTIVE

We explore the complexities and the challenges associated with the design of surveys

on migrants‘ remittances, drawing upon our experiences over a number of studies in

which we compiled and analysed data on migrants‘ remittances across a wide range of

countries and contexts. Our goal is not to prescribe specific solutions, but rather to raise

awareness of key methodological choices and their implications, not only among those

engaged in survey design, but also among analysts who rely on data collected by others.

METHODS

We specify a number of definitional, methodological, and conceptual issues the

researcher must address before embarking on the task of survey and questionnaire

design. These include the need to take into account the various forms that remittances

may take, the different channels through which they are transferred, the parties and

relationships which could be involved, and the differences in the perspectives of

senders and recipients, and of individuals and households. We also examine the

implications of survey issues related to the sensitivity and the reliability of responses.

University of Queensland, Australia. E-Mail: r.brow[email protected]q.edu.au.

Peace Research Institute Oslo, Norway. E-Mail: jorg[email protected].

Maastricht Graduate School of Governance, The Netherlands.

E-Mail: sonja.fransen@maastrichtuniversity.nl.

Maastricht Graduate School of Governance, The Netherlands.

E-Mail: melissa.siegel@maastrichtuniversity.nl.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1244 http://www.demographic-research.org

CONCLUSIONS

Our results suggest that both survey designers and analysts who are relying on survey

data collected by others should (1) be aware of the specific methodological choices

involved in collecting survey data about remittances; and (2) set priorities that reflect

(a) the substantive objective of the study, (b) the particular context at hand, and (c) the

value of generating results comparable to those of surveys in other contexts.

1. Introduction

Our purpose in this paper is to examine remittance measurement in surveys, and the

methodological and conceptual challenges associated with it. Many researchers in the

field have addressed particular aspects of remittance survey methodology (Plaza,

Navarrete, and Ratha 2011; Millis, Orozco, and Raheem 2008; de Brauw and Carletto

2008). What is lacking is a comprehensive overview of the conceptual and

methodological issues that need to be addressed at an early stage in all survey research

on remittances. This is what we seek to provide in this paper.

There are certainly pitfalls researchers should be aware of, and there are

approaches that can be dismissed as faulty. However, there exists an array of

methodological choices which are neither clearly right nor wrong. Analysts must

therefore consider which approach is most appropriate in a given case based on the

specific empirical context, the objective of the research, and the resource limitations

that the survey is subject to. In order to make wise choices, researchers must be aware

of the options and of the implications of each choice. We promote an awareness of

these issues in this paper. While we point to implications for subsequent analysis, we

stop short of discussing the methodology of analysis itself. Instead, we seek to stress

what we consider to be the fundamental sources of problems and biases which

frequently undermine the analysis of remittances, irrespective of the strength and the

robustness of the particular methods and techniques employed by the analyst.

The issues we address are relevant not only to those engaged in the design and

implementation of surveys, but also to analysts who rely on survey data collected by

others. All too often, researchers work with survey data while having little or no

knowledge of how the questionnaire was designed; how well the aspects relevant to

their analysis were covered in the survey; and how the questions were framed, phrased,

and presented by the interviewers. For example, as we discuss later, many surveys ask

only about financial transfers, while ignoring in-kind transfers or purchases and other

indirect payments made by the migrant on behalf of the recipient. As these types of

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1245

transfers may constitute a substantial share of total remittances, the results of such

surveys could lead researchers to develop biases and draw misleading conclusions.

Quantitative information about remittances comes primarily from two sources:

aggregated records from financial institutions, which form the basis of balance-of-

payment figures; and sample surveys conducted among people who might send or

receive remittances. Aggregated records have the obvious advantages that they are not

samples and that the figures are ‒ at least in principle ‒ comparable between countries.

There are three reasons why sample surveys are needed as a complement to these

official figures.

First, the remittance records of financial institutions capture only a portion of the

actual remittance flows. This is primarily because remittances are often sent through

informal channels, and because formal transfers are not always recorded as remittances.

In addition, the diversity and the complexity of migration-related financial transfers

raises many questions about which transfers should be regarded as remittances. The

answers to these questions may be expected to vary depending on the empirical context

and the research objective. These questions are thoroughly addressed in later sections.

Second, surveys can provide information about individuals and/or about

transactions, whereas official statistics focus on aggregate amounts. In other words, the

units of analysis are fundamentally different. While balance-of-payments statistics can

be used to calculate per capita remittance inflows, only survey data can tell us how

those inflows are distributed across the receiving population.

Third, surveys allow for the collection of complementary information that makes it

possible to analyse determinants or impacts of remittance flows. Cases in point are

household-level data on income, and information about the kinship relationships

between senders and recipients.

In the next section, we discuss the possible objectives of measuring remittances in

surveys, which should provide a basis for methodological choices. We then critically

examine the unitary notion of ―remittances,‖ and consider the range of transfers which

could inadvertently be included or excluded in surveys. Next, we discuss research

design and sampling, as well as specific questionnaire items. In the final section before

the conclusion, we address issues related to reliability and validity.

2. Objectives of remittances surveys

We consider here four principle objectives underlying survey-based research on

remittances, while acknowledging that these objectives are not necessarily mutually

exclusive.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1246 http://www.demographic-research.org

2.1 Mapping remittances

The most basic objective is to gather data about remittance flows, either as important

information in its own right, or as a foundation for conducting analyses of their

determinants or impacts. Among the basic questions that might be addressed are the

following:

 How often are remittances sent/received, if at all?

 To and by whom are remittances sent?

 What are the amounts of the remittances sent/received?

 What forms do these remittances take?

 What transfer channels are used?

Researchers cannot be certain that a remitter or a recipient who is asked about

―remittances‖ will consider all of the relevant transfers. Moreover, an individual

remitter or recipient could send or receive remittances in a variety of forms and through

various channels. The objective of mapping remittances requires us to pay attention to

the range of transfer types which might be relevant (cf. section 3), and to the factors

which could lead interviewees to be unwilling or unable to provide accurate answers

(cf. section 6).

2.2 Investigating channels and costs of remittances

Transfer mechanisms differ on a number of parameters, including cost, speed, and

accessibility. When making monetary transfers, a migrant can choose to use formal

bank channels, other formal non-bank money transfer organisations, or informal

channels. These informal channels include hand-carried cash transfers and the use of

informal couriers, such as hawala and its variants (El Qorchi, Maimbo, and Wilson

2003).

Understanding the choice of remittance channel is relevant to several policy

objectives. First, since transfer costs are highly variable and can exceed 20% of the

transfer amount, lowering average costs could have a large direct impact on the

remittances received. Second, shifting transfers from informal to formal channels is

considered by some to be relevant in the fight against money-laundering and terrorist

financing. Third, greater use of the banking system can have indirect benefits for

development, stimulating the availability of additional credit and the demand for

financial services which underpin longer-term financial development.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1247

Questionnaires aimed at investigating the forms, channels, and costs of remitting

must take into account the possibility that respondents have limited information. It is,

for example, highly unlikely that recipients will have any knowledge of the transfer

costs. Moreover, the full cost of transfers is usually not transparent to the sender either,

as it is often obscured by a combination of fees and exchange rate margins. Surveys can

contribute to information on preferred forms and channels of remittances, and, where

financial institutions are involved, market knowledge, usage patterns, and preferences.

Issues to consider include the following:

 Which transfer forms and mechanisms are available and known to the

respondent?

 Which transfer forms and mechanisms are used?

 What is the basis for choosing between various forms and mechanisms?

2.3 Analysing determinants and motives

Understanding the determinants of remitting is essential for assessing which changes in

conditions, events, or interventions could stimulate (or deter) the sending of

remittances, and for crafting policies aimed at fostering remittance flows. There is a

range of factors which could explain individual patterns of remitting, as Hagen-Zanker

and Siegel (2007) and Carling (2008a) have observed. Much of the economics literature

has sought to link the effects of specific determinants to the underlying motives for

remitting, most notably altruism, insurance, investment, and repayments (Lucas and

Stark 1985; Rapoport and Docquier 2006; Brown, Connell, and Jimenez-Soto 2014).

Other studies have focused on the separate determination of migrants‘ capacity and

desire to remit (Carling and Hoelscher 2013).

In analysing the determinants of and the motives for remitting, we cannot rely on

direct questions to the respondent; instead, we need to assess the motives indirectly by

testing alternative hypotheses about the responsiveness of remitting behaviour to

changes in the migrants‘ and/or recipients‘ income or welfare, and/or to particular

events as predicted by the different motivations. Issues analysts should consider in

designing the questionnaire include the following:

 What is the history of past transactions between senders and recipients?

 What are the transnational family structures of the respondents? In other

words, who are the potential senders and recipients of both intra-familial

remittances and remittances beyond the migrants‘ own families?

 What are the financial needs and resources available to the sending and the

receiving households?

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1248 http://www.demographic-research.org

 Who makes decisions about remitting?

 How are the sender and the recipient implicated in other remittance

transfers, which could have indirect effects on the transfers being studied?

 Are there particular events that affect remittance patterns during the recall

period (illnesses, weddings, funerals, natural disasters, unemployment,

etc.)?

Strictly speaking, properly modelling the determinants of remittances requires us

to have information about each dyad of potential senders and receivers (Carling 2008a).

Even in the textbook case of a single migrant remitting to the household of origin, there

could be parallel flows to elderly parents or siblings who live in other households—

flows that are shaped in part by other determinants. In order to make good use of survey

resources, analysts must adopt informed strategies for reducing the complexity while

retaining the information that is most valuable to the research. For instance, when

seeking to construct dyadic information without a matched sample, it is pertinent to ask

what information can realistically be obtained from senders about (potential) recipients,

and vice versa (cf. section 4.2 and 6.1).

The determinants of and motives for remitting are intricately connected to the

broader relationships between the senders and the recipients (Carling 2008b). Although

money is (primarily) sent in one direction, these transfers are typically associated with

various forms of reciprocity, both material and non-material. A recent series of

ethnographical studies on remittances has looked explicitly at the elusive elements of

reciprocity (e.g., Åkesson 2011; Singh and Cabraal 2013; Thai 2014). This

ethnographic literature has provided us with fresh insights into the dynamics of

remitting behaviour which have yet to be integrated into survey-based approaches

(Carling 2014).

2.4 Assessing the effects and the impact of remittances

The impact of remittances on economic development processes can be examined using

information on the recipient households‘ income, expenditures, wealth, savings,

entrepreneurial activities, and investments; including their investments in health,

education, agriculture, and other business activities (e.g., Brown et al. 2006).

Increasingly, we are seeing scholars investigate the impact of remittances on the general

well-being of recipient households. For example, Sen (1999), Moser (1998), and others

have argued that a holistic approach to well-being would take factors in addition to

income into account (Loschmann and Siegel 2014; Brown and Jimenez 2011). Such

approaches are often multi-dimensional investigations of factors like health, education,

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1249

and living standards; or they examine and interpret well-being in a more subjective or

relative manner.

Nonetheless, the usual starting point for these kinds of studies is to

look at the effects of remittances on income. Research on the effects of remittances on

income needs to address two basic questions:

 What is the size and composition of household income from all sources?

 What would the size and composition of this income be in the

counterfactual scenario without migration and remittances?

The latter question represents a considerable methodological challenge: the analyst

needs to compare the observed state of the world with migration and remittances to

some hypothetical state of the world without migration and remittances. To this end, the

questionnaire design must allow for the estimation of both observed and counterfactual

income. A full discussion of the various analytical approaches is beyond the scope of

this paper,

but it is useful to point to the different effects that must be taken into

account: (i) the direct effect of remittance inflows as an addition to other observed

sources of household income; (ii) the direct effect of the lost income due to the out-

migration of one or more household members; (iii) the direct effect of migration on per

capita income, given that the number of household members will be lower than in the

counterfactual, no-migration scenario; and (iv) the indirect effects of both the out-

migration of household members and the inflow of remittances on the household‘s

allocation of resources, and thus income from other sources (cf. Brown and Leeves

2011).

Second, a counterfactual scenario without migration and remittances for any

households in the research area would have general equilibrium effects on, for example,

wages, production, expenditures, and prices in the domestic economy. It should not be

assumed that only the incomes of migrant- and/or remittance-receiving households will

be affected; there are likely to be indirect effects on other households through, for

example, the labour market and the price effects of migration and spending from

remittances.

Before embarking on a study of the impact of remittances, it is worth considering

whether and how other effects of migration on welfare should be taken into account and

The effects of remittances on development appear to have an even broader scope if we consider ―social

remittances,‖ or the ideas, practices, identities, and social capital that flow from receiving to sending country

communities (Levitt 1998). However, although this concept makes use of the term ―remittance‖ as a

metaphor, it essentially describes the broader interactions of migration, transnationalism, and social change.

Our focus here remains on remittances in the conventional sense of financial transfers.

Methodological approaches to addressing this challenge are discussed by Adams (2011) and McKenzie and

Sasin (2007). We are assuming here that the analyst is restricted to non-experimental survey data, and must

therefore rely on estimations of hypothetical, counterfactual income from non-migrant households.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1250 http://www.demographic-research.org

reflected in the questionnaire. As we noted previously, we favour a multi-dimensional

approach to measuring well-being because we believe it is more effective than a purely

monetary income approach when assessing the development and the well-being of

individuals and households. For instance, school absenteeism might be higher among

children who have to compensate for parental absence by working inside and outside of

the home. There could also be social and psychological costs associated with migration

that may be revealed using appropriate questionnaire items.

Finally, a focus on the effects of migration raises critical questions about the unit

of analysis: almost no studies have accounted for the impact on the migrants

themselves. If the absent migrants are considered a part of the household or of the

country of origin, a complete analysis would consider the costs, both material and

psychological, incurred by the migrant as well (cf. Clemens and Pritchett 2008). It is

important to note here that the definition of ―household‖ may be unclear, as this concept

varies depending on the country context. For example, in some countries a typical

household may consist of a nuclear family living under one roof, or a multi-generational

family sharing a home, or a family with multiple wives (Randall, Coast, and Leone

2011). It is therefore important that researchers always use a definition that makes sense

in the particular country context, and that they are explicit about the definition used.

3. Meanings and forms of remittances

The term ―remittance‖ is most often defined roughly as follows: ―funds transferred by

migrants abroad to their families at home‖ (International Organization for Migration

2005:24). But when we start to look at the details of a survey ‒ i.e., who is being

interviewed, where the survey is being conducted, and what the goal of the survey is ‒ it

becomes clear that ―remittance‖ can refer to a very wide range of transfers. Depending

on the research design and questionnaire items, specific forms of transfers could be

inadvertently included or excluded, which could have serious implications for the data

and the analysis. Ethnographic studies of remittances can often yield valuable insights

into the ways in which actual remittances practices may deviate from the expectations

of remittance analysts (see Carling 2014 for an overview).

In designing the survey instrument, the analyst needs to be aware of all of the

possible forms, channels, senders, and recipients of remittances. But given the research

objectives and resource constraints, it will not always be possible or even necessary to

capture every form of remittance. The following nine questions serve as a checklist

intended to help the researcher make better decisions about the use of scarce survey

resources. Some of the answers have implications for design and sampling; others are

related to the questionnaire. We address each of these areas in subsequent sections.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1251

3.1 Does it matter whether value is transferred in the form of money?

In some remittance corridors, in-kind remittances represent a substantial proportion of

the value that is transferred (Unger and Siegel 2006; Young, Osman, and Dale 2007;

Brown et al. 2006). Material gifts may be preferred to cash for a number of reasons. For

example, such gifts may have greater emotional value than cash, or the items given may

be unavailable or unreasonably expensive in the recipient‘s country. In the latter case, it

may be preferable for the migrant to send goods for sale instead of remitting money.

Value can also be transferred indirectly to recipients when migrants make direct

payments to third parties for services that benefit their relatives. For example, migrants

may cover medical bills, school fees, or airfares. Such transfers often are not captured

by typical survey questions about remittances. If the objective of the research is to

quantify the determinants or the effects of remittances, in-kind transfers and payments

for services should be considered. On the other hand, if the goal of the research is to

gain a better understanding of the use and the cost of alternative monetary transfer

channels, or of the implications for financial development in the recipient country, only

questions about monetary transfers from migrants to recipients would be relevant.

3.2 Does it matter whether the money is sent by a migrant?

Several types of transfers that do not come from migrants might nevertheless be

relevant in studies of remittances, such as pensions sent to return migrants and money

sent by descendants of migrants. The sending of remittances is a common indicator in

studies of second-generation transnationalism (Kasinitz et al. 2002; Rumbaut 2002), but

the very concept of ―remittances from the second generation‖ is a contradiction in terms

if we adhere strictly to the dominant definitions of remittances.

3.3 Does it matter whether the money is sent between relatives?

There is little doubt that intra-familial transfers account for the bulk of global

remittances, but non-family transfers can be significant in certain contexts. Migrants

could send remittances to more distant relatives or friends, or to community-level

development initiatives such as hometown associations. Remittances may also take the

form of donations to religious or community organisations. Even if institutionalised

transfers are deemed inconsequential, it could make sense to define money sent

between unrelated friends as remittances. Whether the analyst chooses to include these

types of transfers will depend on the extent to which the study seeks to investigate the

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1252 http://www.demographic-research.org

wider, community-level effects of remittances, as well as the effects on the migrants‘

own households; such as in Brown, Leeves, and Prayaga (2014).

3.4 Does it matter whether the money is transferred to the sender’s country of

origin?

In transnational family networks, remittance recipients might not be located in the

sender‘s country of origin. For instance, the refugee diaspora typically encompasses

communities in distant, high-income countries, as well as communities in the region.

Substantial remittances may flow from the former to the latter communities: e.g., from

Somalis in Europe to Somalis in Kenya, or from Afghans in the United States to

Afghans in Pakistan. Again, whether the analyst chooses to include such transfers will

depend on whether the objective of the study is to investigate the implications for the

migrant-sending community, or for the remitting migrant.

3.5 Does it matter whether the money becomes the recipient’s property?

A remittance is generally thought of as a transfer from one person to another, possibly

within a household group who pool their resources. In many cases, however, the

migrant retains exclusive ownership of money that is transferred. When migrants send

money for the construction of a house, for instance, the funds could simply be managed

by relatives on the migrant‘s behalf. Similarly, the money could be intended for onward

transfers to secondary recipients (Erdal 2012).

If the migrant sends money for the purpose of direct asset accumulation in the

home country, it will not necessarily be transferred via anyone other than the person in

that country. If one of the goals of a given research project is to examine the

implications of remittances for saving and investment in the migrants‘ country, it would

be important for these kinds of remittances to be covered.

This also raises the important issue of whether it is necessary to capture the full

extent of the migrants‘ accumulated savings and investment in their home countries. A

migrant saver/investor has a number of options with respect to the timing and the

location of asset accumulation. One option is to remit to the home country for the

purpose of personal asset accumulation. Another option is to accumulate assets in the

host country with a view to transferring these as a lump-sum capital transfer at some

time in the future; usually when the migrant returns to the home country. As Amuedo-

Dorantes and Pozo (2013) have argued, the choice of the country in which the

accumulated savings are held ‒ and thus of the timing of financial transfers to the home

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1253

country ‒ is essentially an issue of portfolio choice, which will be influenced by the

migrant‘s perception of relative rates of return in the host and home country economies.

Moreover, some studies have found that the migrant‘s intentions to return home will

affect the amounts and timing of such transfers (Ahlburg and Brown 1998). For these

reasons, it is important to combine survey questions about current/recent remittances

with additional questions about the migrants‘ savings and accumulation of assets in

both the host and home countries, as well as their intentions for returning to the home

country (see section 5.5). Thus, it is also important that surveys of migrant households

in the home country include questions not only about remittances received from current

migrants, but about assets transferred by returning/returned migrants.

3.6 Does it matter whether the transfer is unrequited?

According to common perception and the International Monetary Fund‘s definition of

the term, remittances are transfers without any quid pro quo. However, the broader

reciprocal relationships between senders and recipients can make it difficult to ascertain

whether specific services are provided in return for remittances, especially as these are

often informal and implicit. For instance, if a migrant has left a child behind with a

caregiver, remittances could be sent to cover the additional expenses and labour

associated with caring for the extra child. In a study of the well-being of the recipients,

it matters whether the remittances are truly ―income‖ or are ―reimbursements‖ for

expenses that the household would not have incurred if the migrant had not left.

Modelling the impact requires data which enable the researcher to estimate a

counterfactual situation (cf. section 2.4).

3.7 Is it important to also capture possible reverse transfers?

The bulk of remittances flow from migrants to their country of origin country. It is

common, however, for money to also move in the opposite direction. For instance,

people who migrate for work might receive financial support from their families in

times of crisis. In some cases, such as those of student migrants, there may be

substantial reverse remittances. A questionnaire that properly captures two-way flows

will be longer ‒ and thus more costly to use ‒ than a survey of one-way remittances. In

deciding whether to capture this information, analysts must consider the importance of

two-way flows to the research at hand, as well as the likely prevalence and volume of

reverse remittances in the context.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1254 http://www.demographic-research.org

3.8 Does it matter whether money is transferred from one place to another?

A migrant who holds capital assets in the country of origin could avoid transfer costs by

channelling returns on such assets directly to relatives, thereby reducing the need for

international transfers. Homeowners can provide relatives with free housing or allow

them to keep any rent from tenants. If such internal transfers are in lieu of international

remittances, and the focus of the research is on the determinants and the effects of

remittances, these transfers should indeed be treated as another component of

remittances. It is increasingly common for migrants to arrange for recipient households

to make withdrawals from an account held by the migrant using an ATM debit card left

with the recipient. This arrangement would still represent an international transfer, but

because a withdrawal might not be considered by the respondent as ―sending money,‖ it

may be necessary to include explicit questions about such transactions in the

questionnaire (Brown et al. 2006).

3.9 Does it matter whether remittances are international or internal?

In much of the developing world, remittances between individuals within the same

country ‒ e.g., transfers to family members in rural areas from migrants working in

urban areas ‒ are much more widespread than international remittances. Depending on

the research objectives, it may be relevant to ask about remittances in general

(regardless of the origin), to pose separate questions about domestic and international

remittances, or to ask about international remittances only. Remittance recipients are

often expected to share their income from abroad within their communities. Through

internal transfers to non-migrant households, the gains from migration can be

distributed more widely in the community. If the focus of the research is on the welfare

gains from migration and remittances, including questions relating to internal transfers

made and received would be essential (Brown et al. 2006).

Inspired by these nine points, Table 1 illustrates how survey questions on

remittances differ in their coverage of various transfer types. How questions are

formulated will depend on the context in question and the purpose of the research.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1255

Table 1: Examples of survey questions on remittances and their coverage of

various transfer types

Sending-side surveys

Receiving-side surveys

Type of transfer

Do you send money

regularly to somebody

who is living in another

country?

Do you ever give

economic assistance

to family members in

your country of origin?

In the past 12 months,

did this household

receive help in the

form of money or

goods from another

individual living in

another country?

Does this household

receive money from

the United States?

Pensions sent to return

migrants from their former

country of employment.

N/A

Yes

Person-to-person remittances

sent between non-related

friends.

Yes

Person-to-person transfers

made in return for a service.

Possibly

Yes

Money deposited by migrants

in their own bank accounts in

the country of origin.

Rent for a migrant’s house,

collected and kept by the

migrant’s relatives.

Yes

Possibly

Goods shipped by migrants to

relatives in the country of

origin.

Possibly

Yes

Sources and notes: 1) Migrations between Africa and Europe (MAFE), coordinated by the Institut National d’Études Démographiques, France. Adapt

ed from a

retrospective question. 2) Levekår blant innvandrere i Norge 2005/2006, coordinated by Statistics Norway. Translated from Norwegian original. 3) Gallup World

Poll. Slightly adapted; the original referred to internal or international remittances. 4) Mexican Migration Project, coordin

ated by Princeton University and

Universidad de Guadalajara. ‘N/A’ denotes ‘Not applicable’. ‘Possibly’ indicates that it is open to interpretation or case-by-case variation whether or not the transfer

type is covered by the survey question.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1256 http://www.demographic-research.org

4. Research design and sampling

Two sets of issues are central to the design of surveys on remittances: the relations

between household-level and individual-level data, and the relations between sending-

side and receiving-side information. In the remainder of this section we address each of

these issues.

4.1 Individuals and households

Much research on remittances implicitly assumes that money is transferred from ―the

migrant‖ to ―the household.‖ It is not coincidental that the survey questions reproduced

in Table 1 address remittance senders as ―you‖ and recipients as ―this household.‖

These formulations reflect the approach of what has been labelled the ―New Economics

of Labour Migration‖ (NELM), which treats remittance senders primarily as absent

members of the household of origin (Lucas and Stark 1985; Taylor 1999). The legacy

of the NELM raises several issues regarding the role of households in data collection on

remittances. These issues are in addition to the general challenges that arise when

conducting household-based surveys in different cultural contexts (Randall, Coast, and

Leone 2011).

First, it might be misleading to see migrants as ―members‖ of specific households

in the country of origin; the migration patterns and household structures in the context

at hand determine whether this is a valid assumption. Whether it is necessary to assign

household membership to remittance senders depends on the purpose of the research: it

may be a methodological necessity for estimating a counterfactual non-migration

scenario, but not for conducting research with other objectives.

Second, even if most migrants can be linked to a household of origin, it is

important to also consider remittances to households without migrants, as noted in the

previous section (see point 2). Such transfers could be particularly positive because they

are not a corollary of the absence of a family member (Amuedo-Dorantes, Georges, and

Pozo 2010; Brown et al. 2006; Brown, Leeves, and Prayaga 2014). Intra-familial

transfers often cross household boundaries, and in some contexts non-familial

remittances may be considerable.

Third, income might not be pooled within the household. This point has been

increasingly recognised in studies of household economies over the past decade, but the

methodological implications of such arrangements remain poorly understood (Fisher,

Reimer, and Carr 2010). But if the household is the unit of analysis, the dynamics of

intra-household distribution are not necessarily critical to data collection. It could be

decided legitimately to limit the study, for instance, to the impact of remittances on

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1257

inequality between households. What is lost, then, are the distributional effects within

the household ‒ which could be significant if the income balance between spouses or

generations is shifted.

Fourth, information about income might not be pooled within the household. This

is potentially critical to data collection. The receiving side survey questions in Table 1

are typical in that they are based on the assumption that respondents have information

about remittances to their household. While such questions are typical, many

ethnographic studies have indicated that there is often a considerable degree of secrecy

surrounding remittances. In a more general analysis of information-sharing within

households, Fisher Reimer, and Carr (2010) concluded that a single respondent may not

be able to account adequately for all of the sources of household income. This is

especially likely to be the case, they noted, when households are more complex or

resourceful, as is often the case with remittance-receiving households.

These four observations challenge the prominence of the household in data

collection among remittance recipients. But the contrasting, individual framing of

remittance senders is also problematic, as a sender‘s identity can be ambiguous. There

are three factors that could identify the sender, and which can point to different

individuals: 1) original ownership of the money or goods, 2) the act of making the

transfer, and 3) the relationship with the recipient(s). In the case of a migrant couple,

the husband may be the sole income earner and the one making the transfer, but he may

send the money to his parents-in-law on his wife‘s behalf. If the husband or the wife is

sampled in a survey of individuals, it is not obvious how they should respond, nor how

their answers should be interpreted alongside other individual-level variables.

When we scrutinise the roles of households and individuals in the sending of

remittances, two overarching methodological implications emerge. First, data collection

strategies must be assessed with respect to the specific family and household patterns in

the given context. In a cross-country survey, there may be trade-offs between designs

that are feasible across the board, and designs that are adapted to each setting (Fransen,

Kuschminder, and Siegel 2012). Second, the purpose of the research should inform the

analyst‘s choice of survey design. Table 2 presents the three main designs, each of

which has advantages and disadvantages.

Since the exact relationship between individuals and households is both variable

and unknown, there is a case to be made for including individual- as well as household-

level questions. For instance, the four questions reproduced in Table 1 could be

reformulated to a pair of questions about ―you personally‖ and ―others in your

household.‖ If this dual approach is adopted in individual surveys and in standard

household surveys, the two designs become similar. The remaining difference reflects

the sampling: in a household survey the individual interviewees are usually not a

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1258 http://www.demographic-research.org

relevant sample of persons; in an individual survey, the households of interviewees are

not necessarily representative.

Table 2: Characteristics of the main designs for surveys on remittances

Individual survey

Standard household

survey

Extended household

survey

Basic structure

Individuals are sampled

and interviewed

Households are sampled,

and a well-informed

household member is

interviewed

Households are sampled

and several household

members are interviewed

individually

Implicit assumptions

related to

remittances

Remittances can be

studied through the

transactions of individuals

Remittances are primarily

household-level

transactions; information

is pooled within the

household

Remittances must be

understood in the context

of households; household

members have different,

relevant information

Advantages

Possibility of different

sampling methods, not

only based on residence;

possibility for targeted

samples with a higher

proportion of remittance

senders/recipients;

relatively low cost and

complexity

Cost-effectiveness of

covering many

households with few

interviews;

representativeness at the

household level

Possibility for both

individual-level and

household-level analysis;

possibility for examining

intra-household

distribution;

independence of

information-sharing within

households

Disadvantages

Limited possibility for

examining household-

level processes; sample

not representative at the

household level

Possibly unrealistic

assumptions about

information-sharing within

the household; necessity

of residence-based

sampling; limited value of

the sample of individuals

Possible ethical and

logistical challenges of

individual interviews; high

costs; necessity of

residence-based

sampling

Costs (assuming

face-to-face

interviews)

Relatively low, because

interviewees are sampled

directly

Medium; partly dependent

on procedures for

selecting the interviewee

within the household (and

therefore the necessity of

multiple visits)

Very high, also compared

to standard household

surveys; multiple visits to

the same household are

more likely; a much larger

number of interviews are

required for a given

household level N

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1259

4.2 Senders and recipients

There are three basic designs with respect to the geographical perspective of data

collection. Sending-side surveys are conducted in the migrants‘ destinations among

potential remitters. Receiving-side surveys are conducted in migrants‘ areas of origin

among potential recipients. Matched-sample surveys are conducted at both the receiving

and the sending end among members of the same households or families. Due to the

difficulties and costs involved in tracking migrants‘ home country households, such

studies are rare (Mazzucato 2008; Chort, Gubert, and Senne 2012; Osili 2007).

Sending- and receiving-side surveys typically need to collect information about the

other side to facilitate analysis. For instance, a receiving-side survey on remittance

effects must collect fairly detailed information about migrants in order to estimate their

counterfactual, non-migration income. Similarly, a sending-side survey on determinants

must collect information on the potential recipients and their characteristics. The

advantage of matched-sample surveys is that they collect first-hand (and therefore more

reliable) information on each part of the sender–receiver dyad. Moreover, matched-

sample surveys can yield additional information, such as differences in perceptions

between migrants and their non-migrant counterparts.

Most migration and remittances surveys are single cross-sectional surveys,

providing a snapshot of one point in time. Questions about past behaviour and changes

over time are subject to recall errors. Moreover, inferring causality in the relationship

between variables over time from observed differences between individuals or

households at the same point in time is methodologically questionable. If resources

permit, it makes sense to design the survey to include second (or subsequent) waves of

interviews of the same households. Panel data greatly enhance the analytical capacities

of the dataset and the rigour of the statistical analysis.

5. Framing of questionnaire items

In this section, we discuss the most common remittance-related questionnaire items and

the challenges and potential pitfalls that accompany each. We show that the content of

remittance questionnaire items depends heavily on the context in which the survey is

conducted.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1260 http://www.demographic-research.org

5.1 Remittance frequency

Basic remittance questionnaire items ask respondents whether the household (or the

individual) sends or receives remittances at all, and, if so, how often. A common

weakness of such questions is that there is often no time reference included. For

example, some questions are posed as follows: ―Do you send money or goods home?‖

Yet it is essential to add a time frame, such as ―in the past 12 months,‖ ―since your

arrival in this country,‖ or ―currently.‖ The analyst should also consider, however,

which time frames are realistic in light of the respondents‘ capacity to recall. Since

there are often seasonal variations in remittance sending, the 12 months preceding the

survey is a commonly adopted time frame.

Subsequent questions can measure the frequency of the transfers within the chosen

time frame. A common misconception here is that transfers are regular, and that

response categories such as ―once a month‖ or ―once a year‖ are appropriate. The

timing of remittances often depends on the financial situation of the migrant, or on the

occurrence of irregular events such as a job loss, wedding, or funeral. One possible

solution is to ask how many times the respondent sent/received remittances during the

reference period.

5.2 Remittance amounts

Questions about the volume of remittances sent or received can request the amount per

transfer or the total amount within a given time frame. Here too, choices should reflect

the purpose of the research and the characteristics of the research setting. Aggregate

amounts are what matters for analysing the effect of remittances on income and

expenditures. However, for studies of transfer mechanisms or of the psychology of

financial decision-making, the size of each transfer makes a difference. How difficult it

is for respondents to answer various questions about amounts will depend on the

characteristics of the target population and their finances.

Questionnaire items about remittance amounts must specify the currency in which

the response is expected, but would ideally allow for deviations. Depending on the

transfer mechanism they use, respondents in the same setting could quantify their

remittances in different currencies. Allowing respondents to choose the currency might

strain the researcher‘s resources at the time of data analysis, but will yield more reliable

results than forcing respondents to convert currencies on the spot.

In-kind remittances can be quantified in a number of ways, including by asking

respondents for the purchase prices of the goods sent by the migrant or for the local

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1261

prices of the goods received, or by assigning values to the listed goods after data

collection (Brown et al. 2006).

5.3 Remittance senders and recipients

Surveys of remittance senders will often need to contain questions about the recipients,

and vice versa. As we noted in section 4.2, information from both ends of the transfer is

required to analyse the determinants of and the motives for remitting behaviour.

However, the kinds of information that can be reliably obtained through a one-sided

survey are limited. Remittance recipients often know little about a migrant‘s

employment situation abroad, and the information they do have could reflect the

migrant‘s desire to appear successful, rather than his or her actual situation. Among the

types of information that remittance recipients can realistically provide are the

migrant‘s year of departure, year of birth, and current country of residence. Questions

about the migrant‘s city of residence, employment, or household may be more difficult

for a remittance recipient to answer, depending on the context.

Potential intra-household differences are important to consider. For example, one

household could include both senders and recipients of remittances. Consider a

household in Kenya, from which the eldest son went to Europe to study. The migrant‘s

father still lives in Kenya and sends him a monthly allowance; while at the same time,

his mother, who is living in the same household as his father, receives money from her

sister in the US. Such cases are complex, but are definitely no exception. As we

discussed in section 4.1, the analyst must weigh the value of individual-level data

against the higher data collection costs and the possible implications for sample size.

5.4 Transfer mechanisms

The transfer mechanism that is chosen by the remitter depends on the transaction costs,

the speed, and the security of the specific channel, among other factors (Carling et al.

2007). Some countries have a poorly developed infrastructure which limits the

possibilities for sending remittances. When asking questions such as: ―How did you

receive the money that was sent to you in the past 12 months?‖ or ―How did you send

money to your family in the past 12 months?‖ the respondent needs to be presented

with an exhaustive list of possibilities that are suited to the context. It is imperative for

the researcher to have detailed knowledge of the financial infrastructure and the

remittance market in question. It is preferable to analyse the remittance market at the

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1262 http://www.demographic-research.org

local level, given the substantial within-country variation in terms of financial

infrastructure.

When migrants send remittances to different recipients, different transfer

mechanisms may be chosen for each recipient. Moreover, the mode of transfer to a

specific recipient could differ depending on the amount or the season. If transfer

mechanisms are a key concern in the study, the questionnaire should accommodate such

variability. Open questions about why the migrant chooses a specific transfer

mechanism could provide additional valuable information.

5.5 Contextualizing information

Analyses of information about remittances usually require contextual data about the

households and individuals concerned, as well as contextual data about the social norms

shaping relationships in the community under study. Many of these issues fall outside

the scope of this paper, but the following areas merit attention.

Transnational family formations underpin the bulk of remittance transfers. In order

to understand the determinants of remittances, it is essential to include information not

only about actual transfers, but about potential senders and recipients. For instance, the

remittance patterns of migrants who are settled abroad with their closest family

members could depend primarily on whether or not their parents in the country of origin

are still alive. If this information is missing, economic variables alone could have

limited explanatory power. The remittance behaviour of an individual migrant can also

depend on the number of other migrants from the same home country household. For

instance, in cases in which remittances are driven primarily by altruism, and thus by

some perception that the receiving household needs support, the amount each migrant

actually remits may be reduced if there are several remitters from the same household.

It is therefore important to collect information about other migrants from the same

household, not only when surveying in the home country, but also among the migrants

in the host country. And, as we have already noted (see section 3.5), contextual

information from the households about returned migrants and the future migration

intentions of current migrants, returned migrants, and non-migrants must also be

collected when assessing the levels and the timing of remittances and other migrant

transfers, such as the accumulated assets brought home by migrants upon their eventual

return (as in Ahlburg and Brown (1998) and Brown and Connell (2004)), or remittances

For instance, it is also likely that differences in socio-cultural norms and customs could influence the

migrant‘s choice of transfer mechanism and recipient. For example, a migrant from a patriarchal society may

be reluctant to send all of his remittances directly to a female household member. (We thank an anonymous

referee for raising this point.)

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1263

sent specifically to fund the migration of other household members (see, for example,

Brown and Poirine (2005)).

The remitter’s migration history can be central to explaining remittance patterns. If

the survey in question is not a full-fledged migration survey, the analyst may have to

find a way to capture key aspects of the migration history using a few well-chosen

questions. The main challenge is how to deal with the possibility of multiple migrations

and multiple destinations. As a given migrant may have had repeated periods of

residence at the destination, analysts will have to decide whether they are primarily

interested in the beginning of the current period, the beginning of the first period, or the

aggregate duration of stay. Or, as a given migrant may have previously stayed in other

countries, researchers will have to decide whether they are most interested in the year of

arrival at the current destination, or the year of (first) departure from the country of

origin. The choices a researcher ultimately makes should be based on the needs of the

analysis and the prevalence of different migration patterns in the context at hand. (See

Carling (2012) for further details on collecting migration history information.)

Household income is central to analyses of both determinants and impacts of

remittances. The researcher faces the key decision of whether to use the household‘s

stated income or expenditures, or whether to use both methods; each of these methods

has its advantages and disadvantages (see Deaton (1997) for a full discussion). If the

income method is used, the questionnaire needs to allow for each conceivable source of

income, and should be designed to minimise the possibility of double counting. For

instance, if there is a separate section on remittances, then the section on income should

be carefully designed to exclude remittances. Similarly, if the household is engaged in

household production, agricultural and/or other commercial businesses, there must be a

section allowing for the estimation of net income (value added) from these activities (cf.

Brown et al. 2006).

Household wealth can be a valuable complement to data on income. One

commonly used method for capturing this information is to construct a comprehensive

list of assets, the ownership or non-ownership of which is believed likely to vary across

households depending on their relative wealth (Filmer and Pritchett 2001). For each

household a wealth index is accordingly derived as a proxy measure of the value of the

household‘s assets. This variable can be used either as an explanatory or a control

variable, or as the variable of interest in assessing the effects of remittances, as in

Brown et al. (2006). In surveys that avoid asking about amounts of money (for either

remittances or income), asset lists can be the basis for the socio-economic profiling of

households.

Inter-household transfers could be central to understanding the household finances

of remittance-receiving households. They can receive or make transfers to other

households or organisations such as community groups and charities within the country,

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1264 http://www.demographic-research.org

either in cash or in kind. Given that remittance-receiving households are often under

substantial social pressure to assist other ‒ especially non-remittance-receiving ‒

households, it is essential to capture such intra-household transfers as part of the

analysis of the effects of remittances on poverty and distribution across the wider

community (cf. Brown, Leeves, and Prayaga 2014).

6. Reliability and sensitivity

Survey research is always subject to resource constraints that limit the size of samples,

the methods of interviewing and the length of questionnaires. Integral to the

management of scarce resources is an understanding of what is realistic to ask in a

given context. In this section, we therefore address different aspects of the psychology

of survey responses as they relate to information about remittances.

6.1 Knowledge, encoding, and recall

Answering a survey question presupposes that respondents have had access to the

information, have encoded it as knowledge, and are able to retrieve it at the time of the

interview (Foddy 1993; Groves et al. 2004). Each of these steps can be related

specifically to research on remittances.

Access to information depends primarily on intra-household information sharing.

Both on the receiving and the sending side, an individual respondent may not know

about the remittances other household members are sending or receiving (cf sections

4.1 and 5.3). This problem increases with the level of detail requested. For instance,

respondents might know whether money is being remitted and have an idea of the

frequency, but not know the amounts.

Encoding is a pertinent issue in relation to information on transfer mechanisms:

respondents might not have internalised knowledge about how the money was sent or

received. Researchers typically work with taxonomies of transfer mechanisms ‒ e.g.,

formal vs. informal ‒ which could be alien to respondents. If money transfer services,

such as those of Western Union are offered through banks, it could be particularly

difficult for respondents to account for the transfer mechanism used; i.e., whether the

transfer was made through a bank or a money transfer service. In such circumstances,

more detailed follow-up questions could be needed to determine the transfer type.

Recall errors represent a limitation in all survey research, but vary greatly with the

type of information requested (Foddy 1993; Groves et al. 2004). The longer the time

span of the question, and the less salient the event the respondent is being asked about,

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1265

the harder it is for him or her to remember. As a result, the analyst faces a choice in

designing a questionnaire: i.e., whether to ask detailed questions about the recent past,

or more general questions about a longer time period. Since there is often a seasonal

aspect to remittance-sending, there are good reasons for asking about behaviour during

the past 12 months. But when such a long period is chosen, it may be unrealistic to

request details of the amounts remitted.

Difficult questions do not necessarily lead to high non-response rates, however.

The United States Census Bureau (2008) included a set of questions on remittances in

the August 2008 round of the Current Population Survey, including the following:

 In the last 12 months, did anyone in this household give or send money to

relatives or friends living outside the U.S.? Please include all gifts of

money. Do not include loans.

 In the last 12 months, about how much in total dollars was given or sent?

(Your best guess is fine.)

Among remittance senders ‒ the 5% of the sample who answered the first question

affirmatively ‒ a full 87% gave a response to the question about the total amount

(Fantom 2009). This question was manageable for people who had only remitted a

couple of times during the year, or who did so with a regular monthly amount. For

people from households with several remittance senders or irregular remittance

patterns, the question was difficult to answer, even with a ―best guess.‖ Thus, the

challenge for researchers lies not so much in dealing with the 13% non-response rate as

in assessing the reliability of answers that were given.

6.2 Sensitivity and bias

Non-response and reliability in survey data are closely related to the sensitivity of

questions. Remitting can be a sensitive issue in variable and context-specific ways.

Understanding and addressing this challenge requires the researcher to pay attention to

the social environment of the respondents as well as to their perceptions of the survey.

Several mechanisms could be at work and create biases.

First, at the micro level, the issue of who sends (or does not send) remittances to

whom has moral dimensions, and can be a delicate aspect of interpersonal transnational

relationships (Carling 2008b). The experiences of Amery and Anderson (1995) in rural

Lebanon illustrate this point. They interviewed household heads and initially enquired

about the level of remittances received from each migrant son. However, respondents

were reluctant to answer ―because they perceived it as ‗ranking their children‘ (to an

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1266 http://www.demographic-research.org

outsider, the investigator) on the basis of money received from each‖ (Amery and

Anderson 1995:57). The total amount received was, however, willingly reported.

Second, norms within respondents‘ communities could influence their answers.

Migrants from low-income countries sometimes have a strong sense of duty to remit,

and may be ashamed of their inability to fulfil remittance expectations (Akuei 2005;

Hernandez 2002; Lindley 2009). The general point that socially desirable behaviour

may be over-reported in surveys also applies to remitting. In a widely cited example of

over-reported donations, Parry and Crossley (1950) found that only half of survey

respondents who reported giving money to a specific local charity had actually done so.

Third, policies or debates in the respondents‘ social environment could affect their

perceptions of remittances, and thus influence their responses. For instance, media

coverage linking remittances to tax evasion, benefits fraud, or the financing of militant

groups can make respondents reticent, even if their own transfers come from regular

income and are sent to family members (Carling, Erdal, and Horst 2012; Horst et al.

2014). If the environment is perceived as being hostile to remittances, respondents

could reply negatively to an initial question about remitting in order to avoid the

subject.

Fourth, if respondents perceive that the information they provide could be passed

on to the government or other agencies, they could have incentives to give specific

answers. For instance, senders could under-report the use of illegal transfer mechanisms

for fear of prosecution, and recipients could under-report remittance income for fear of

taxation or other government interference, regardless of whether such concerns are

based in reality. In developing countries where there are public or private needs-based

assistance programmes, respondents could fear that reporting remittance income would

undermine their eligibility for such benefits.

Survey specialists have developed a number of approaches to asking sensitive

questions in ways that maximise response rates and the reliability of answers

(Bradburn, Sudman, and Wansin 2004; Foddy 1993; Groves et al. 2004; Tourangeau

and Smith 1996; Tourangeau and Yan 2007). But before using these tactics, the analyst

should have a thorough understanding of whether specific questions are sensitive; and,

if so, how. For instance, in settings with strong social pressure to remit, questions could

be loaded to excuse non-remitting. If, however, hostile media attention is the main

concern, different forms of loading might be appropriate. The value of adapting

questions to the context in this way needs to be weighed against the reduced

comparability the findings with those of studies conducted elsewhere.

Sensitivity assessments should also inform other aspects of survey administration.

For example, how can respondents ‒ and, where applicable, their local communities ‒

best be prepared for the survey? How can interviewers best prevent misunderstandings

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1267

about the nature and purpose of the survey? Is there a need for interviewers to record

their own perceptions of question sensitivity?

The most difficult decisions for researchers are about including or excluding

questions altogether. For instance, information about remittance amounts is undeniably

valuable, but asking a question about these amounts could be counterproductive if

doing so results in high non-response rates and unreliable data. If a question is

perceived as being overly threatening, posing it would be a poor use of scarce interview

time. Moreover, asking a threatening question could undermine the respondent-

interviewer relationship, which could in turn affect the responses to subsequent

questions.

7. Conclusions

When designing surveys about remittances or when using survey data collected by

others in a subsequent analysis, researchers are likely to encounter a series of dilemmas.

Making good decisions in such cases requires analysts to be aware of the options and

their implications. In this paper, we have sought to provide an overview of these issues.

The first overarching dilemma is the complexity dilemma. Remittance-related

transfers can be dauntingly complex. How much of that complexity should a

questionnaire aim to capture? If the available resources are finite, every additional

question has an impact on the sample size, and must be properly justified. For example,

when is it pertinent to collect data on reverse remittances or to elicit information from

each household member? To make wise choices, the researcher must be knowledgeable

about the context at hand ‒ e.g., about the prevalence and importance of reverse

remittances ‒ and remain focused on the purpose of the research.

However, when the researcher decides to keep the questions basic, the number of

potential pitfalls actually increases. While simple questions can be good, they can also

be based on simplifying assumptions that contradict reality. For instance, it would be

ill-advised to phrase questions based on the assumption that the recipient household

receives money from only one person. It may, however, be acceptable to simplify the

survey by asking only about aggregate remittances or only about the person who remits

the most.

The second overarching dilemma is the variability dilemma. The best

questionnaire in a given context will reflect the specifics of that setting and the relevant

remittance transactions. But when the questions are heavily adapted to local conditions,

the survey results can be difficult to compare with the findings of studies conducted

elsewhere. Again, the analyst must make a choice based on the purpose of the research,

as well as possible secondary uses of the data.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1268 http://www.demographic-research.org

Both the complexity and the variability dilemmas can be more appropriately

addressed with solid contextual knowledge. While we have sought to show that such

knowledge matters greatly for the proper design and interpretation of surveys, we can

offer no blanket recommendations for how to obtain it. In some cases, preparatory

qualitative research can support the design of surveys. In other cases, the solution might

lie in (possibly interdisciplinary) collaboration with researchers who have the relevant

context-specific expertise. Thus, even if the researcher has no ambition to conduct a

mixed-method project that includes qualitative data collection, he or she will often need

to draw upon insights from qualitative research to design the surveys appropriately.

(For comprehensive discussions on the integration of surveys and qualitative data in

demography, see Coast, Hampshire, and Randall 2007; Randall and Koppenhaver 2004;

and Schatz 2012.)

Conducting a survey is no guarantee that the researcher will reap the potential

benefits of collecting individual-level information; it simply opens up opportunities.

Whether the survey results are of high quality will depend on whether the researcher

made well-informed choices about design, sampling, and questionnaire items. By

choosing wisely, analysts can produce new and better data that will help us understand

why remittances are sent, how transfer mechanisms are chosen, and how the transfers

affect poverty, inequality, and development. Moreover, for those involved in the

analysis of survey data collected by others, being more aware of the potential

limitations, gaps, and biases which can arise from improperly designed surveys could

encourage a more balanced focus on the full range of methodological issues facing the

analyst, and a move away from the current preoccupation with potential biases and the

relative robustness of alternative statistical methods and techniques.

8. Acknowledgements

The authors gratefully acknowledge the insightful comments of two anonymous

referees on a previous draft of the paper.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1269

References

Adams, R.H. (2011). Evaluating the Economic Impact of International Remittances on

Developing Countries using Household Surveys: A Literature Review. Journal

of Development Studies 47(6): 809–828. doi:10.1080/00220388.2011.563299.

Ahlburg, D.A. and Brown, R.P.C. (1998). ‗Migrants‘ intentions to return home and

capital transfers: a study of Tongans and Samoans in Australia‘. Journal of

Development Studies 35(2): 125–151. doi:10.1080/00220389808422567.

Åkesson, L. (2011). Remittances and relationships: exchange in Cape Verdean

transnational families. Ethnos 76(3): 326–347. doi:10.1080/00141844.2011.

577229.

Akuei, S.R. (2005). Remittances as unforeseen burdens: The livelihoods and social

obligations of Sudanese refugees. Geneva: Global Commission on International

Migration.

Amery, H.A. and Anderson, W.P. (1995). International migration and remittances to a

Lebanese village. Canadian Geographer 39(1): 46–58. doi:10.1111/j.1541-0064.

1995.tb00399.x.

Amuedo-Dorantes, C., Georges, A., and Pozo, S. (2010). Migration, Remittances, and

Children's Schooling in Haiti. Annals of the American Academy of Political and

Social Science 630(1): 224–244. doi:10.1177/0002716210368112.

Amuedo-Dorantes, C. and Pozo, S. (2013). ‗Remittances and portfolio values: an

inquiry using immigrants from Africa, Europe, and the Americas. World

Development 41: 83–95. doi:10.1016/j.worlddev.2012.05.036.

Bradburn, N., Sudman, S., and Wansin, B. (2004). Asking Questions. San Francisco:

Jossey Bass.

Brown, R.P.C. and Connell, J. (2004). The Migration of Doctors and Nurses from

South Pacific Island Nations. Social Science and Medicine 58(11): 2193–2210.

doi:10.1016/j.socscimed.2003.08.020.

Brown, R.P.C. and Leeves, G. (2011). Comparative effects of migrants‘ remittances on

composition of recipient household income in two small, island economies.

Applied Economics 43(27): 3965–3976. doi:10.1080/00036841003742611.

Brown, R.P.C., Connell, J., and Jimenez, E.V. (2014). Migrants‘ Remittances, Poverty

and Social Protection in the South Pacific: Fiji and Tonga. Population, Space

and Place. 20(5): 434–454. doi:10.1002/psp.1765.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1270 http://www.demographic-research.org

Brown, R.P.C., Connell, J., Jimenez, E.V., and Leeves, G. (2006). Cents and

Sensibility: the economic benefits of remittances. In: World Bank (ed.). Pacific

Islands at Home and Away: Expanding Job Opportunities for Pacific Islanders

through Labor Mobility. Washington: The World Bank: 42-97.

Brown, R.P.C. and Jimenez, E.V. (2011). Subjectively-assessed welfare and

international remittances: evidence from Tonga. Journal of Development Studies

47(6): 829–845. doi:10.1080/00220388.2010.501376.

Brown, R.P.C., Leeves, G., and Prayaga, P. (2014). Sharing Norm Pressures and

Community Remittances: Evidence from a Natural Disaster in the Pacific

Islands. Journal of Development Studies 50(3): 383–398. doi:10.1080/00220388.

2013.858127.

Brown, R.P.C. and Poirine, B. (2005). ‗A model of migrants' remittances with human

capital investment and intrafimilial transfers‘. International Migration Review

39(2): 407–438. doi:10.1111/j.1747-7379.2005.tb00272.x.

Carling, J. (2008a). The determinants of migrant remittances. Oxford Review of

Economic Policy 24(3): 582–599. doi:10.1093/oxrep/grn022.

Carling, J. (2008b). The human dynamics of migrant transnationalism. Ethnic and

Racial Studies 31(8): 1452–1477. doi:10.1080/01419870701719097.

Carling, J. (2012). Collecting, analysing and presenting migration histories. In: Vargas-

Silva, C. (ed.) Handbook of Research Methods in Migration. Cheltenham:

Edward Elgar: 137–162. doi:10.4337/9781781005231.00014.

Carling, J. (2014). Scripting remittances: making sense of money transfers in

transnational relationships. International Migration Review 48(s1): 218–262.

doi:10.1111/imre.12143.

Carling, J. and Hoelscher, K. (2013). The Capacity and Desire to Remit: Comparing

Local and Transnational Influences. Journal of Ethnic and Migration Studies

39(6): 939–958. doi:10.1080/1369183X.2013.765657.

Carling, J., Erdal, M.B., Horst, C., and Wallacher, H. (2007). Legal, Rapid and

Reasonably Priced? A Survey of Remittance Services in Norway. Oslo:

International Peace Research Institute (PRIO Report 2007/3).

Carling, J., Erdal, M.B., and Horst, C. (2012). How does conflict in migrants‘ country

of origin affect remittance-sending? Financial priorities and transnational

obligations among Somalis and Pakistanis in Norway. International Migration

Review 46(2): 283–309. doi:10.1111/j.1747-7379.2012.00888.x.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1271

Chort, I., Gubert, F., and Senne, J.-N. (2012). Migrant networks as a basis for social

control: Remittance incentives among Senegalese in France and Italy. Regional

Science and Urban Economics 42(5): 858–874. doi:10.1016/j.regsciurbeco.

2012.02.001.

Clemens, M.A. and Pritchett, L. (2008). Income per Natural: Measuring Development

for People Rather Than Places. Population and Development Review 34(3): 395–

434. doi:10.1111/j.1728-4457.2008.00230.x.

Coast, E., Hampshire, K., and Randall, S. (2007). Disciplining anthropological

demography. Demographic Research 16(16): 493–518. doi:10.4054/DemRes.

2007.16.16.

De Brauw, A. and Carletto, G. (2008). Improving the Measurement and Policy

Relevance of Migration Information in Multi–Topic Household Surveys

Washington, DC: International Food Policy Research Institute.

Deaton, A. (1997). The Analysis of Household Surveys: A Microeconometric Approach

to Development Policy. Baltimore: Johns Hopkins University Press for the

World Bank. doi:10.1596/0-8018-5254-4

El Qorchi, M., Maimbo, S.M. and Wilson, J.F. (2003). Informal Funds Transfer

Systems: An Analysis of the Informal Hawala System. Washington:

International Monetary Fund. doi:10.5089/9781589062269.084.

Erdal, M.B. (2012). Who is the Money for? Remittances within and beyond the

Household in Pakistan. Asian and Pacific Migration Journal 21(4): 437–457.

Fantom, N. (2009). Improving the Measurement of International Remittances. Paper

presented at the Rome International Conference on Remittances. Rome: The

World Bank and the Italian Ministry of Foreign Affairs. 9 September.

Filmer, D. and Pritchett, L. (2001). Estimating wealth effects without expenditure data

or tears: An application to educational enrollments in states of India.

Demography 38(1):115–132.

Fisher, M., Reimer, J.J., and Carr, E.R. (2010). Who Should be Interviewed in Surveys

of Household Income? World Development 38(7): 966–973. doi:10.1016/

j.worlddev.2009.11.024.

Foddy, W. (1993). Constructing Questions for Interviews and Questionnaires. Theory

and Practice in Social Research. Cambridge: Cambridge University Press.

doi:10.1017/CBO9780511518201.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1272 http://www.demographic-research.org

Fransen, S., Kuschminder, K., and Siegel, M. (2012). Implementation of Cross-Country

Migration Surveys in Conflict Affected Settings: Lessons from the IS Academy

Survey in Burundi and Ethiopia Maastricht: UNU–MERIT (2012–019).

Groves, R.M., Fowler, F.J., Couper, M.P., Lepkowski, J.M., Singer, E., and

Tourangeau. R. (2004). Survey Methodology. Hoboken: Wiley.

Hagen–Zanker, J. and Siegel, S. (2007). The determinants of remittances: A review of

the literature. Maastricht: MGSOG (2007/3).

Hernandez, E.E. (2002). Power in remittances: Remaking family and nation among

Salvadoreans [Thesis]. Irvine: University of California, Faculty of Social

Science.

Horst, C., Erdal, M.B., Carling, J., and Afeef, K. (2014). Private money, public

scrutiny? Contrasting perspectives on remittances. Global Networks. 14(4): 514–

532. doi:10.1111/glob.12048.

International Organization for Migration (2005). Mainstreaming migration into

development policy agendas. Geneva: IOM.

Kasinitz, P., Waters, M. C., Mollenkopf, J. H., and Anil, M. (2002). Transnationalism

and the children of immigrants in contemporary New York. In: Levitt, P. and

Waters, M. C. (eds.). The changing face of home. The transnational lives of the

second generation. New York: Russell Sage Foundation: 96–122.

Levitt, P. (1998). Social remittances: migration–driven local–level forms of cultural

diffusion. International Migration Review 32(4): 926–948. doi:10.2307/

2547666.

Lindley, A. (2009). The Early-Morning Phonecall: Remittances from a Refugee

Diaspora Perspective. Journal of Ethnic and Migration Studies 35(8): 1315–

1334. doi:10.1080/13691830903123112.

Loschmann, C. and Siegel, M. (2014). The influence of vulnerability on migration

intentions in Afghanistan. Migration and Development 3(1): 142–162.

doi:10.1080/21632324.2014.885259.

Lucas, R.E.B. and Stark, O. (1985). Motivations to Remit: Evidence from Botswana.

Journal of Political Economy 93(5): 901–918. . doi:10.1086/261341.

Demographic Research: Volume 31, Article 41

http://www.demographic-research.org 1273

Mazzucato, V. (2008). Simultaneity and Networks in Transnational Migration: Lessons

Learned from a Simultaneous Matched–Sample Methodology. In: DeWind, J.

and Holdaway, J. (eds.). Migration and Development Within and Across

Borders: Research and Policy Perspectives on Internal and International

Migration. Geneva and New York: International Organization for Migration and

Social Science Research Council: 69–100.

McKenzie, D. and Sasin, M.J. (2007). Migration, Remittances, Poverty and Human

Capital: Conceptual and Empirical Challenges. Washington D. C.: World Bank

Policy Research.

Millis, B., Orozco, M., and Raheem, Z. (2008). Mining remittance data: Practical

considerations on survey design and administration. United States Agency for

International Development (MicroREPORT #119).

Moser, C. (1998). The asset vulnerability framework: Reassessing urban poverty

reduction strategies. World Development 26(1): 1–19. doi:10.1016/S0305-

750X(97)10015-8.

Osili, U.O. (2007). Remittances and savings from international migration: Theory and

evidence using a matched sample. Journal of Development Economics 83(2):

446–465. doi:10.1016/j.jdeveco.2006.06.003.

Parry, H. and Crossley, H.M. (1950). Validity of Responses to Survey Questions.

Public Opinion Quarterly 14(1): 61–80. doi:10.1086/266150.

Plaza, S., Navarrete, M., and Ratha, D. (2011). Migration and Remittances Household

Surveys in Sub–Saharan Africa: Methodological Aspects and Main Findings.

Background paper for the African Migration Project.

Randall, S. and Koppenhaver, T. (2004). Qualitative data in demography: The sound of

silence and other problems. Demographic Research 11(3): 57–94.

doi:10.4054/DemRes.2004.11.3.

Randall, S. Coast, E., and Leone, T. (2011). Cultural constructions of the concept of

household in sample surveys: Population Studies 65(2): 217–229.

doi:10.1080/00324728.2011.576768.

Rapoport, H. and Docquier, F. (2006). The Economics of Migrants' Remittances. In:

Kolm, S.C. and Ythier , J.M. (eds.). Handbook on the Economics of Giving,

Reciprocity and Altruism, Applications Volume 2. Amsterdam: Elsevier: 1136–

1195.

Brown et al.: Methodological and conceptual issues for survey designers and data analysts

1274 http://www.demographic-research.org

Rumbaut, R.G. (2002). Severed or sustained attachments? Language, identity and

imagins communities in the post–immigrant generation. In: Levitt, P. and

Waters, M.C. (eds.). The changing face of home. The transnational lives of the

second generation. New York: Russell Sage Foundation: 43–95

Schatz, E. (2012). Rationale and procedures for nesting semi–structured interviews in

surveys or censuses. Population Studies 66(2): 183–195. doi:10.1080/

00324728.2012.658851.

Sen, A. (1999). Development as Freedom. Oxford: Oxford University Press.

Singh, S. and Cabraal, A. (2013). Contested representations of remittances and the

transnational family. South Asia 36(1): 50–64. doi.10.1080/00856401.

2012.710304

Taylor, J.E. (1999). The new economics of labour migration and the role of remittances

in the migration process. International Migration 37(1): 63–86. doi:10.1111/

1468-2435.00066.

Thai, H.C. (2014). Insufficient Funds. The Culture of Money in Low–Wage

Transnational Families. Stanford: Stanford University Press.

Tourangeau, R. and Smith, T.W. (1996). Asking sensitive questions – The impact of

data collection mode, question format, and question context. Public Opinion

Quarterly 60(2): 275–304. doi:10.1086/297751.

Tourangeau, R. and Yan, T. (2007). Sensitive questions in surveys. Psychological

Bulletin 133(5): 859–883. doi:10.1037/0033-2909.133.5.859.

Unger, B. and Siegel, M. (2006). The Netherlands–Suriname Corridor for Worker's

Remittances: Prospects for Remittances when Migration Ties Loosen. Utrecht:

Dutch Ministry of Finance and the World Bank.

US Census Bureau (2008). Current Population Survey. Washington, DC: Department

of the Census. (Immigration/Emigration Supplement File)

Young, H., Osman, A., and Dale, R. (2007). Darfurian Livelihoods and Libya: Trade,

Migration, and Remittance Flows in Times of Conflict and Crisis. International

Migration Review 41(4): 826–849. doi:10.1111/j.1747-7379.2007.00101.x.