What Motivates Gifts? Intra-Family Transfers in Rural Malawi
-
Upload
simon-davies -
Category
Documents
-
view
215 -
download
1
Transcript of What Motivates Gifts? Intra-Family Transfers in Rural Malawi
ORIGINAL PAPER
What Motivates Gifts? Intra-Family Transfers in Rural Malawi
Simon Davies
Published online: 8 July 2010
� Springer Science+Business Media, LLC 2010
Abstract This paper uses Family Transfers Project data
collected in rural Malawi during 1999 to ascertain the
motivation for gift-giving using discriminating hypotheses.
The study models monetary and monetized gifts sent and
received between the survey respondents and their parents,
their children, and their siblings as a function of sender and
receiver characteristics. Individual analyses are compared
with household level models to reveal that both individual
and household characteristics can matter in different cases.
OLS, Probit and Tobit models are compared to conclude
that, as with other similar studies, a wide range of moti-
vations exist including altruism, (co-)insurance, and an
inheritance motive. Motivations differ slightly depending
upon the relationship between the sender and receiver,
however, no single motive can be attributed to any given
relationship.
Keywords Africa � Altruism � Family transfers �Family economics � Gift-exchange � Insurance
Introduction
The growing importance of transfer flows and the realiza-
tion that these can be a means to both alleviate poverty and
moderate the impact of negative income shocks has
encouraged work attempting to understand the motivation
for remitting and the impact of transfers in developing
countries. Models of transfer motivations can be divided
into several main branches: (1) those modeling transfers as
altruistic behavior in which the utility of the sender is
influenced by the utility of the receiver; and (2) (partly)
self-interested models. It should be noted that in (2), we
have both the investment/exchange motive and the ‘‘warm
glow’’ motive.1
Self-interested models can be further split into subcat-
egories: (2a) suggests that transfers are primarily used for
investment purposes and will therefore respond to the
macroeconomic climate—for example, investing in busi-
nesses; (2b) views transfers as part of a ‘‘joint optimiza-
tion’’ agreement in which both the remitter and receiver
gain from risk-sharing. Different income sources shared
permit the reduction of income risk, and allow such
transfers to be termed ‘‘insurance payment’’2; (2c) views
transfers as repayments of implicit or explicit loans made
to the remitter in the past to say, fund education or
migration. This group of models can be termed ‘‘family as
bank’’ models; (2d) views transfers as a means of safe-
guarding inheritance.3
S. Davies (&)
Department of Economics and International Development,
University of Bath, Bath BA2 7AY, UK
e-mail: [email protected]; [email protected]
1 When even pure altruists would not have an incentive to give, but
are shown to do so, this is termed the ‘‘warm glow’’ motive to giving
(see e.g. Crumpler and Grossman 2008). We would like to thank an
anonymous referee for suggesting this as an alternative explanation.
Interestingly Cowley et al. (2004) find that some families give gifts to
their church which have a negative impact on the family’s financial
situation. This suggests either some social pressure or else non-
financial compensation such as the ‘‘warm glow’’ motive.2 This is particularly important in an agricultural economy such as
Malawi. For example, households living in different areas might
implicitly agree to send each other transfers whenever one has
suffered from drought and the other not. Alternatively, urban and
rural income may follow different patterns making it possible to share
income through implicit transfer agreements in order that when one
household suffers a negative shock, it is supported by the other.3 An additional motivation can be found by distinguishing between
giving to one’s children and the ultimate aim of ‘‘dynastic altruism’’
or the ‘‘survival of the gene’’ (see e.g. Horioka 2002; Fan 2005).
123
J Fam Econ Iss (2011) 32:473–492
DOI 10.1007/s10834-010-9216-1
Why study motivations for remitting? Untangling
motivations for transfers is an important component in
understanding transfer flows in general. Most importantly,
different motivations to remit imply different impacts.
Altruistic motivations may act as a counter-cyclical force
helping to reduce risk of poverty during negative shocks
(such as droughts or macroeconomic shocks). Altruistic
motivations combined with asymmetric information may
have the effect of reducing labor-market participation
which could reduce welfare (Azam and Gubert 2004).
Altruistic and ‘‘family as bank’’ motivations to remit could
increase consumption at the micro level, but have negative
‘‘Dutch Disease’’ effects at the macro level if the transfers
are from abroad. ‘‘Dutch Disease’’ is an effect whereby a
large amount of money coming from abroad causes the
value of the currency to appreciate as demand for it rises.
This makes goods produced in the country for export to
appear more expensive for foreign buyers making it more
difficult for certain industries to survive in the face of
international competition. This can hurt production and
jobs. See, for example, Sachs and Warner (2001). Insur-
ance motivations serve to reduce risk for the family, per-
haps encouraging greater risk taking (perhaps investment)
and increasing welfare. Such risk pooling may also
increase crop yields as households switch to riskier, higher
yield crops (Dercon 1996), and prevent the selling of
productive assets (Fafchamps et al. 1998). Finally, invest-
ment motivations suggest that transfers can help to improve
output and productivity with the caveat that investment in
existing housing stock and land may encourage a Balassa–
Samuelson effect with negative macro consequences. That
is, increases in the prices of these may pass through to
other areas of the economy causing some industries to
become less competitive in export markets. Any transfers
from abroad may impact on the exchange rate (Amuedo-
Dorantes and Pozo 2004).
Various authors have made attempts to understand the
motivations for remitting using discriminating hypotheses
to test between different motivations whilst others have
noted correlations between remittances and household
characteristics (e.g. Gupta and Hegde 2009). Despite the
importance of understanding transfers, there has been little
progress, with authors largely concluding that they are
unable to rule out certain motivations. This study extends
existing papers in the following ways: firstly we look at
transfer flows going in both directions—that is it looks at
both ascending (children to parents) and descending (par-
ents to children) transfers.4 See Sheng and Killian (2009)
for a study of inter-generational transfers. With the
exception of VanWey (2004) no study has, to the author’s
knowledge, analyzed transfer flows in each direction of the
transfer relationship in a developing country (See Koh and
MacDonald 2006 for an example in a Western context).
This is surprising given the potential importance of co-
insurance in remitting behavior, and the importance of
mutual gift-giving in many developing countries.
As well as analyzing monetary transfer flows in both
directions, this paper makes one further extension. We do
not model transfer flows only between a household of
origin and migrants, but between a central household and
their children, parents and siblings. It is likely that different
family members have different motivations for transfers
potentially allowing us to disentangle the motivations
which other authors have, so far, been unable to achieve.
Thus, for example, we might expect to find an inheritance
motivation amongst children of respondents, but not for the
parents of respondents. In addition to the contributions
stated above, this paper takes advantage of a previously
unused data set.
The remainder of this paper is organized as follows:
‘‘Theoretical and Empirical Setting’’ section reviews the
relevant theoretical models before summarizing empirical
findings from previous related studies in developing
countries. ‘‘Data’’ section discusses the descriptive data
and ‘‘Empirical Analysis’’ presents the econometric mod-
eling and discusses the results. Conclusions are drawn in
the last section.
Theoretical and Empirical Setting
Studies analyzing motivations to remit use either Tobit
models to estimate the value of transfers, Probit models to
estimate the probability of sending or receiving transfers or
OLS to estimate net transfers received. The independent
variables focus on the receiver’s and sender’s characteris-
tics. Regressions thus take the form:
Transfers ¼ f�Receiver characteristics;
Sender characteristics; X�
ð1Þ
where Transfers is value of transfers sent or received
(Tobit), net transfers (OLS), or whether or not transfers
were received (Probit) and f(.) is the relevant function
(Tobit, OLS, Probit). The X represents any other study
specific variables included. This paper concentrates on
monetary transfers and monetized value of physical gifts
such as food. It does not therefore look at other transfers
such as time (see Hayhoe and Stevenson 2007 or Cao 2006,
for a discussion of the link between time and financial
transfers).
Rapoport and Docquier (2006) provide an excellent
review of the theoretical and empirical literature on moti-
vations for remitting. They begin by illustrating an4 We would like to thank an anonymous referee for pointing this out.
474 J Fam Econ Iss (2011) 32:473–492
123
altruistic theory in which both the sender (s) and the
receiver (r) exhibit altruism towards each other and in
which the utility of the sender, UsðCs;CrÞ is a weighted
average of his/her felicity derived from his/her own con-
sumption,5 VsðCsÞ and the utility of the receiver,
UrðCr;CsÞ.UsðCs;CrÞ ¼ ð1� bsÞVsðCsÞ þ bsUrðCr;CsÞ ð2ÞUrðCr;CsÞ ¼ ð1� brÞVrðCrÞ þ brUsðCs;CrÞ ð3Þ
where felicity exhibits diminishing marginal return in
consumption, V0[ 0 and V00\ 0 and the 0� bi� 1=2giving the degree of altruism. If bs ¼ 1=2 then the sender
values the receiver’s happiness resulting from consumption
as much as he values his own. Values above � in which the
sender values the receiver’s happiness above his/her own
are excluded, and a value of bs ¼ 0 results in a purely
selfish model in which the sender (usually a family
migrant) does not consider the utility of his/her family at
all. It should be noted that this form of analysis can be used
to study both ascending (from children to parents) and
descending (parents to children) transfers. This paper
extends previous literature by looking at both.
Rapoport and Docquier (2006) incorporate transfers by
re-writing consumption as equal to income, I, less transfers,
T. In addition, they rule out the possibility of negative
transfers from the sender to the receiver, and impose a
felicity function satisfying V0[ 0 and V00\ 0, V(.) = ln(.)
and solve for the optimal level of transfers from the sen-
der’s perspective. The resulting altruistic model has several
properties: (1) Transfers are increasing in the sender’s
income; (2) Transfers are falling in the receiver’s income;
(3) Transfers are increasing in the sender’s altruism; and
(4) Transfers are falling in the degree of altruism of the
receiving household.
This provides several testable hypotheses, but these
results could also be generated by other factors. Rapoport
and Docquier (2006, p. 12) note that the ‘‘main testable
implication of the altruistic model is that transfers cannot
increase with the recipient’s income’’.
Agarwal and Horowitz (2002) formally model a sug-
gestion by Funkhouser (1995) that under altruism, transfers
from any one remitter (a migrant in their model) should
decline in the number of remitters (migrants), but this
should not be the case under an insurance hypothesis. This
provides an additional testable implication.
In their two period models, a transfer sender faces cer-
tain income in period 1, equal to Is. In period 2, s/he faces
uncertainty with high income, IsG, with a probability of
1� p and low income, IsB with a probabilityp ð0\p\1Þ.
S/he can choose to remit to a receiver an amount T in the
first period and receive an actuarially fair indemnity
(s ¼ T=p) in the case of a negative shock in the second
period. The receiver (insurer) is assumed to face no
uncertainty. Denoting, as before VsðCsÞ and VrðCrÞ the
sender’s, s, and receiver’s, r, felicity functions, the sender’s
expected utility (EU) is denoted:
EU ¼ VsðIs � TÞ þ ð1� pÞVsðIsGÞ þ pVsðIs
B þ sÞ ð4Þ
where felicity functions are kept constant across time and
state and the sender’s utility depends only on his/her own
consumption, and not that of the receiver, unlike in the
altruistic model.
Using log utility as before, which satisfies decreasing
marginal utility to consumption and risk aversion, the
optimal level of transfers or transfers, T* can be shown to
be:
pð1þ pÞ½I
s � IsB� ¼ T� ð5Þ
This insurance model has several properties:
(1) Transfers are increasing in the sender’s first period
income (as with previous altruistic model);
(2) Transfers are decreasing in the sender’s bad state
income;
(3) Transfers are increasing in the probability of a bad
state (potentially proxied empirically by education,
unemployment or legal status if abroad).
Agarwal and Horowitz (2002) go onto extend the altru-
istic model sketched above to include the fact that house-
holds receive transfers from several senders (migrants). The
model provides one further testable implication: ‘‘[u]nder
pure insurance (or other self-interest) motives, the number
of other migrants would not affect own-transfers. On the
other hand, under altruism where migrants are concerned
with the welfare of the nonmigrating household, the pres-
ence of multiple remitting migrants will affect the average
transfer level’’ (p. 2036). Rapoport and Docquier (2006)
point out that this assumes the exogeneity of the number of
remitters a household benefits from (those with more vol-
atile income or that are more risk averse may ensure they
have more transfer relationships). In addition, they note if
household income is affected by moral hazard then house-
hold income might not necessarily be assumed exogenous.
Moral hazard and transfers are modeled by Azam and
Gubert (2004). Rapoport and Docquier (2006) also sketch a
version of Cox (1987) in which transfers are viewed as
payment for services and Laferrere and Wolff (2006)
describe in detail gift exchange when viewed as intra-family
transfers.
5 ‘‘Felicity’’ is used here in order to distinguish between total utility
(derived from both one’s own consumption and the consumption of
the other) and the utility derived only from one’s own consumption,
which has been called ‘felicity’. For example, the Giver’s Util-
ity = f(Utility of Receiver; His/Her own Felicity) where his/her own
felicity is derived from his/her own consumption.
J Fam Econ Iss (2011) 32:473–492 475
123
On an empirical level finding suitable discriminating
hypotheses is challenging. Authors have used a wide range
of methods drawn from the theory described above to draw
conclusions. Principal results are summarized in Table 1.
Despite efforts to disentangle motivations behind
transfer flows, it is important to note that motivations are
not mutually exclusive. For example, it could be that a
threat to disinherit a child enforces an insurance payout to
be made. Given this, it is not surprising that most studies
have difficulties in concluding unambiguously in favor of a
single motivation. This study aims however to shed further
light on transfer motivations by comparing and contrasting
evidence of different transfer motivations according to
familial relationship between the sender and the receiver.
Data
Descriptive Statistics
This study uses the rural Malawian Family Transfers Pro-
ject (FTP) dataset collected by the University of Pennsyl-
vania Population Studies Center (Social Networks) for the
purpose of analyzing gifts and transfer flows from a
number of different perspectives. The survey was carried
out in three rural areas of Malawi (Balaka in the southern
region, Mchinji in the centre and Rumphi in the north)
between June and August, 1999. The three areas in which
the survey was conducted are both similar and broadly
representative of rural areas in Malawi in socioeconomic
terms and with regards to commercial activities (markets,
banks,…) and institutions (post offices, clinics,…)
(Weinreb 2001, 2002). Unfortunately the data remain
under-used from an econometric point of view. One
exception is Mtika and Doctor (2002) who study differ-
ences in transfer behaviour between matrilineal and patri-
lineal tribes in Malawi and conclude that wealth exchange
in matrilineal tribes are biased towards female relatives
whilst the inverse is true under patriliny.
After cleaning, there are 616 females and 501 males in
the main households surveyed. There are more females due
to lower response rates amongst males, polygamy and
absence. The sampling deliberately targeted working-age
households and made efforts to interview both the house-
hold head and his wife. Males and Females reported
transfers sent and received on an individual (not house-
hold) level. This has several advantages and disadvantages.
The major disadvantage is that transfers can be viewed as a
part of a household activity. That is, transfers might be
given to one member but used for the benefit of several
members, or, alternatively, that the household has decided
as a whole to send a migrant away to earn money and to
pool income. This would make the household level more
appropriate, and is supported by literature stemming from
the New Economics of Labor Migration (Bloom and Stark
1985). This however makes an assumption which is not
entirely appropriate for the current data. Firstly, it is not
clear that all households that receive transfers do so
because of an economic decision to migrate. Rather, other
(cultural) factors are likely to predominate—notably
Table 1 Summary of key findings from studies on motivations to remit
Study Key conclusions
Lucas and Stark (1985)—Botswana Positive association between remittance receipts from children and per capita household
income (altruism). Sons remit more the wealthier is the household (inheritance)
Ilahi and Jafarey (1999)—Pakistan Return Pakistani migrants remit less to their immediate family, the more they have borrowed
from extended family (repayment of past loans)
Agarwal and Horowitz
(2002)—Guyana
The more migrants in the household, the less a migrant will remit (altruism). Lower household
income is associated with higher remittance receipts (altruism)
Naufal (2008)—Nicaragua As the number of migrants increase, remittances from any one sender decline (altruism). As
income risk of the household increases, remittances increase (altruism)
Amuedo-Dorantes and Pozo
(2006)—Mexico/United States
Mexican migrants in the U.S. remit more home to Mexico as their income risk increases
(insurance). Larger home households increase remittances (altruism)
De la Briere et al.
(2002)—Dominican Sierra
Remittances are increasing in work day losses due to sickness for the home household
(altruism, insurance, reverse causality?). Remittances are increasing in inheritable land, but
decreasing in the number of heirs (inheritance)
Van Dalen et al. (2005)—Morocco,
Egypt and Turkey
Higher remittances as home households perceives its financial situation to be ‘insufficient’
(altruism)
VanWey (2004)—Thailand Male migrants more likely to remit to landless households (altruism) and both male and female
migrants remit less as the number of migrants in the household increases (altruism). Female
migrants remit less the more land the household owns (altruism)
Grigorian and Melkonyan
(2008)—Armenia
High unemployment discourages remittance flows to a region (undefined selfish motives)
476 J Fam Econ Iss (2011) 32:473–492
123
marriage or, in the case of children, individual career
aspirations. This is particularly the case for siblings who
have left the same household as the respondents (that of
their parents). Secondly, studies which analyze transfers on
a household level tend to use the head’s characteristics.
Here information on transfers and characteristics were
collected on an individual level allowing me to use the
husband’s or wife’s characteristics, as appropriate. One
possibility would be to include the characteristics of both
the husband and the wife, but this would result in all
observations for which data is absent on one of the partners
(single heads or no-response) being dropped. In any case,
as much information as is used in most studies is used in
estimating the empirical models. Given these data limita-
tions, we proceed with studying transfers from the indi-
vidual perspective.
There are 1145 potential transfer flows between the male
or female and their parents; 522 between the male or
female and their children, and 3945 with siblings. Transfer
flows are studied for all relationships described separately,
and only for adult relations who do not reside in the same
household as the respondent, explaining the small number
of children. Each potential transfer relationship between
two individuals is termed a ‘‘transfer dyad’’. Thus, one
dyad is between the respondent and her father, and another
dyad is the respondent and her mother. We therefore ana-
lyze transfer flows that are inter-household but intra-fam-
ily. Table 2 reports summary statistics for the respondents.
The average age of respondents was around 33 years with
males being on average around 6 years older than females.
Almost all respondents are married and respondents have
3.7 years of education. Interestingly, there is little differ-
ence between males and females in this respect. (Amongst
their parents however, only around 50% of mothers had
any education compared with over 80% of fathers, showing
key generational difference.) Self-reported health status
was, on average over 8/10 although 28% of respondents
reported having suffered from ill health during the previous
month. Average weekly wage income was aver MK300 for
men compared with around MK100 for females, and on
average, men had a far higher asset index score In addition,
many respondents reported looking after the children of the
relatives with whom they have transfer relationships.
Table 3 shows summary statistics for parents. On
average, parents are 60 years (fathers 64 years and mothers
57 years), and respondents rated their parents’ health at
around 6.2/10 on average (little difference between fathers
and mothers). Parents had an average of 7.9 heirs, and 66%
of them reported having some schooling (54% of mothers
and 82% of fathers). Over a third of parents live in the
same village as the interviewed child. Nearly 7% of
respondents reported looking after a sibling.
Since working age respondents were purposively target,
and this study analyses only intra-household transfers, very
few respondents had adult children living outside of the
family home. There are thus only 522 potential dyads, or
transfer relationships. Table 4 presents summary statistics
for children.
The average reported health level of the children was
around 8.4/10, and children were on average around
23 years of age. More were daughters, probably due to
daughters leaving the home to marry at a younger age than
sons, indeed on average daughters were younger than sons.
Around two-thirds of children were married, and 27% had
moved to a city or abroad. Respondents reported having an
average of 4.4 children, and receive transfers from an
average of 1.35 of these. Over 17% of respondents reported
having a grandchild in their household but the data to
not permit me to ascertain to which child the grandchild
belongs.
Table 2 Respondents’ characteristics
Obs Mean Std. Dev.
Age 1010 33.29 (10.64)
Years education 1131 3.71 (3.32)
Married 1166 97.00%
Female respondent 1166 55.06%
Eldest child 1166 19.47%
Health rating (1 = lowest to
10 = highest)
1165 8.14 (1.94)
Health problem in last month 1166 28.22%
Paid income last week (MK) 1117 222.33 (1087.00)
Asset index (males)a 501 0.05 (1.73)
Asset index (females)a 616 0.01 (1.52)
Sibling in house 1166 6.35%
Nephew/Niece in house 1166 12.18%
Grandchild in house 1166 9.35%
Matrilineal ethnicity 1166 23.76%
Patrilineal ethnicity 1166 37.14%
Mixed (Chewa) ethnicity 1166 39.11%
a Created using Principle Components Analysis and includes own-
ership of bed, radio, bike, lamps, pit latrines, cattle, goats, pigs,
poultry, land and quality of housing material
Table 3 Parents’ characteristics
Obs Mean Std. Dev.
Age parents 1144 60.09 (11.88)
Health parents 1139 6.23 (2.35)
Heirs 1147 7.89 (2.74)
Sibling in respondent’s house 1147 6.80%
Schooling 1147 66.17%
Parent lives in same village as respondent 962 37.63%
J Fam Econ Iss (2011) 32:473–492 477
123
Respondents’ siblings are on average around 31.8 years
and respondents rated their health at 8.14 on average.
Around half of siblings are sisters and half brothers. Nearly
a quarter of all siblings live either abroad or in a city inside
Malawi. In nearly 13% of cases, the respondents reported
looking after a sibling’s child. Again, unfortunately the
data do not permit me to ascertain to which sibling the
child belongs. Table 5 presents summary statistics for
siblings.
Transfer Flows
Although both husbands and wives were interviewed in the
majority of cases, respondents reported transfers sent and
received on an individual bases. Thus, transfers sent from
the male to a son is not the same as that sent from the
female to the same son. Respondents reported transfers
sent and received since the end of the previous growing
season—a period of around 3 months. Detailed informa-
tion is given regarding the transfers, and estimated values
of goods (collected in the field) are used to value physical
(as opposed to cash) gifts. Thus, transfer flows analyzed in
this paper include cash and the monetized value of physical
gifts. During the course of the survey, the interview team
regularly verified the market value of the gifts given at the
place of receipt. It was possible to monetize quantities in
local market places or in the nearest market towns. This
was done regularly in order to ensure the current market
values were correct. Secondi (1997) reports very similar
determinants for cash and physical gifts.
There are 1145 potential ‘‘transfer dyads’’ between the
respondents and their parents (see Table 6). Of these,
respondents reported remitting to parents since the last
agricultural season in around 65% of cases, and received
transfers from parents in 50% of cases. Excluding zero
transfer flows, the average transfers sent to parents was
MK245 and the average value of transfers received from
parents was MK205. (MK is Malawi Kwacha, the local
currency unit. At the time of the survey US$1 & MK70.)
Between respondents and children, there are 522
potential transfer relationships with respondents giving to
children in just over half of all cases an average amount of
MK300, and received transfers from children in around
38% of cases with the average amount received being
around MK200.
Respondents reported remitting to siblings in 35% of the
3945 potential cases, and received from them in around a
quarter of cases. Average transfers sent and received are
similar at around MK185.
It is possible that there is some degree of reporting bias
since more respondents reported receiving than giving in
all cases. However, due to sampling, this seems realistic in
the case of parents and children. It is perhaps less likely
with regard to siblings, and it is interesting to note that
there is little difference in the value of transfers sent and
received in this case, and the gap between the number of
respondents reporting sending and receiving transfers is
smaller for siblings than for parents and children.
Empirical Analysis
Econometric Modeling
The main household surveyed is thus the household on
which we focus and transfer functions are estimated for
each of the three transfer relationships. Since transfers are
estimated in dyads, it is possible one person to appear in
the regression more than once, and members of the same
family will appear multiple times in a single regression.
This introduces a potential problem of clustering of errors
into the regressions. All regressions are therefore corrected
for potential clustering at the household level. In addition,
all standard errors are corrected for potential heteroske-
dasticity using White (1980).
We begin by estimating OLS regressions for net transfer
receipts by the respondents in the central household sur-
veyed for each of the three relationships, where net
Table 4 Children’s characteristics
Obs Mean Std.
Dev.
Health of son/daughter (1 = lowest,
10 = highest)
522 8.43 (1.78)
Number of children parents have 522 4.38 (1.89)
Number of sons/daughters remitting to
parents
522 1.35 (1.36)
Age of son/daughter 522 22.94 (5.76)
Eldest son/daughter 522 45.02%
Parents have one of children’s children in
household
522 17.43%
Daughter (not son) 522 59.77%
Son/Daughter lives in city or abroad 522 27.01%
Son/Daughter married 522 64.56%
Table 5 Siblings’ characteristics
Obs Mean Std. Dev.
Age of sibling 2916 31.78 (11.49)
Health of sibling 3945 8.14 (1.95)
Sister 3945 50.37%
Eldest sibling 3945 12.19%
Sibling lives abroad or in city 3945 24.41%%
Respondent household has nephew/niece 3945 12.75%%
478 J Fam Econ Iss (2011) 32:473–492
123
transfers are the amount received by the respondent minus
the amount the remitted to the same person. Thus, we
estimate:
Ri ¼ Xibþ ei ð6Þ
where Ri is net transfer receipts by the individual from the
sender, Xi the regressors indicating sender and receiver
characteristics, and b are the parameter estimates. The error
term, ei is assumed is corrected for potential heteroske-
dasticity and clustering at the household level as discussed.
We next go onto extend the analysis by estimating a
series of Probit regressions indicating both whether or not
transfers were sent and received from a potential transfer
partner.
PðRi ¼ 1jXi ¼ xiÞ ¼ UðXbÞ ð7Þ
where U represents the cumulative distribution function of
the normal distribution. Standard errors are again corrected
for heteroskedasticity and clustering.
We present other estimates in addition to net transfer
flows because, although net flows capture an important
component of transfer relationships, there is little distinc-
tion between those who do not engage in a potential
transfer relationship and those who give and receive a
similar amount. For this reason, probit estimates are pro-
vided alongside net transfers.
In addition, this study reports results from Tobit models
estimating the value of transfers sent/received by each
dyad, and signs and significances of variables are very
close to the Probit models presented.
Discriminating hypotheses for the respondent–parent
transfer perspective are shows in Table 7 and are based on
previous studies. This table summarizes the key focus
of the discussion in the ‘‘Econometric Results and
Discussion’’ section, and similar discriminating hypotheses
are used for other relationships.
Economic specifications used are based on theory which
indicates that motivations for remitting can be inferred
from the impact of certain sender and receiver character-
istics on remittance flows. Of particular interest are any
indicators of income, wealth, recent shocks and any ser-
vices provided by the receiver for the sender. Our data
provide information on recently earned income, household
assets (wealth), recent health shocks and general health as
well whether the respondent household looks after any
children of remittance partners (a service). The model is
augmented by variables known to be important at the
household level in Malawi—for example, we include
whether a household has matrilineal or patrilineal heritage
as this can be important for the inheritance motivation.
Table 8 shows the signs expected on different variables
under different motivations and the specification used for
each model and justification is given below:
Under altruism, remittance receipts from a parent would
be decreasing in the respondent’s wealth, as indicated by
the theoretical model presented above (Rapoport and
Docquier 2006). If remittance flows are altruistic, a parent
would also increase remittance flows (or the likelihood of
remitting) if the respondent suffers from a health shock and
decrease remittance flows as the respondent’s health
improves.
If a parent’s remittances are an insurance premium, they
are likely to increase as the respondent become a more
reliable insurer. Thus, they are increasing in the respon-
dent’s wealth and health. In addition, they can be expected
to increase as the likelihood that a parent suffers from a
negative shock increases, as shown in the theoretical
insurance model above (Agarwal and Horowitz, 2002).
Table 6 Incidence and values
of transfer flowsObs Mean Std. Dev. Min Max
Respondent $ Parents
Respondent ? Parent 1145 64.37%
Parent ? Respondent 1145 50.57%
Value of transfers to parent (excl. zeros) 737 244.57 435.6 1 5000
Value of transfers from parent (excl. zeros) 579 205.13 489.84 5 6350
Respondent $ Children
Respondent ? Children 522 51.72% 0 1
Children ? Household 522 37.74% 0 1
Value of transfers to children (excl. zeros) 275 300.24 455.16 4 4000
Value of transfers from children (excl. zeros) 198 200.82 420.65 4 5000
Respondent $ Siblings
Respondent ? Sibling 3945 35.18%
Sibling ? Respondent 3945 26.84%
Value of transfers to sibling (excl. zeros) 1389 180.03 418.73 1 7503
Value of transfers from sibling (excl. zeros) 1061 188.4 572.3 2 15000
J Fam Econ Iss (2011) 32:473–492 479
123
Ta
ble
7D
efin
itio
nan
dd
iscu
ssio
no
fv
aria
ble
s
Par
ents
Ch
ild
ren
Sib
lin
gs
Res
po
nd
ents
’ch
arac
teri
stic
s
Inco
me
Res
po
nd
ents
earn
edin
com
e(i
nM
alaw
ian
Kw
ach
a)in
wee
kp
rio
rto
the
surv
ey.
Am
easu
reo
fw
ealt
hb
ut
also
cap
ture
scu
rren
tin
com
eto
wh
ich
may
be
‘in
sure
d’
dif
fere
ntl
yfr
om
oth
eras
sets
Ass
etin
dex
(far
m)
Co
nst
ruct
edu
sin
gP
rin
cip
leC
om
po
nen
tsA
nal
ysi
s.C
aptu
res
farm
wea
lth
.R
emit
tan
ces
may
resp
on
dd
iffe
ren
tly
tofa
rmw
ealt
hco
mpar
ed
too
ther
wea
lth
soth
isis
ente
red
sep
arat
ely
Ass
etin
dex
(no
n-f
arm
)C
on
stru
cted
usi
ng
Pri
nci
ple
Co
mp
on
ents
An
aly
sis
Ed
uca
tio
nY
ears
of
edu
cati
on
of
the
resp
on
den
t.A
nin
dic
ato
ro
fq
ual
ity
of
the
resp
on
den
tas
anin
sure
r
Ag
eA
ge
iny
ears
of
the
resp
on
den
t
Mal
eE
qu
alto
1if
the
resp
on
den
tis
mal
e,0
oth
erw
ise
Mar
ried
Eq
ual
to1
ifth
ere
spo
nd
ent
ism
arri
ed,
0o
ther
wis
e
Ho
use
ho
ldsi
zeN
um
ber
of
ho
use
ho
ldm
emb
ers
Res
po
nd
ent
eld
est
chil
dA
nin
dic
ato
rfo
rw
het
her
or
no
tth
ere
spo
nd
ent
isan
eld
est
chil
d.
Po
ten
tial
lyim
po
rtan
tw
ith
reg
ard
sto
rem
itta
nce
sw
ith
par
ents
Hea
lth
(1–
10
)S
ub
ject
ive
var
iab
lein
dic
atin
gcu
rren
th
ealt
h(1
bei
ng
low
and
10
bei
ng
hig
h).
Rel
evan
tb
oth
asan
ind
icat
or
of
qu
alit
yo
fin
sure
r(r
elat
ed
toin
sura
nce
pre
miu
ms)
and
po
ten
tial
of
rece
ivin
gin
sura
nce
ind
emn
itie
so
ral
tru
isti
cre
mit
tan
ces
Hea
lth
pro
ble
min
last
mo
nth
Ind
icat
or
var
iab
leeq
ual
to1
ifre
spo
nd
ent
rep
ort
edb
ein
gsi
ckin
the
pre
vio
us
mo
nth
,0
oth
erw
ise.
Var
iab
lein
dic
ates
are
cen
th
ealt
h
sho
ck
Rem
itta
nce
par
tner
’sch
ild
inh
ou
seIn
form
atio
no
nw
het
her
resp
on
den
t’s
ho
use
ho
ldlo
ok
saf
ter
po
ten
tial
rem
itta
nce
par
tner
’sch
ild
.A
nex
amp
leo
fa
serv
ice
pro
vid
edb
yth
e
resp
on
den
th
ou
seh
old
Giv
entr
ansf
ers
tore
mit
tan
cep
artn
erIn
dic
ates
wh
eth
ero
rn
ot
giv
enre
mit
tan
ces
top
ote
nti
altr
ansf
erp
artn
er.
Cap
ture
sre
cip
roci
ty
Rec
eiv
edtr
ansf
ers
fro
mre
mit
tan
cep
artn
erIn
dic
ates
wh
eth
ero
rn
ot
rece
ived
rem
itta
nce
sfr
om
po
ten
tial
tran
sfer
par
tner
.C
aptu
res
reci
pro
city
Rem
itta
nce
par
tner
char
acte
rist
ics
Ag
eo
fre
mit
tan
cep
artn
erA
ge
iny
ears
of
rem
itta
nce
sp
artn
er
Nu
mb
ero
fch
ild
ren
rem
itti
ng
Ind
icat
esh
ow
man
ych
ild
ren
are
rem
itti
ng
toth
ere
spo
nd
ent.
Incl
ud
ed
bas
edo
nth
eory
sep
arat
ing
altr
uis
tic
fro
mo
ther
mo
tiv
atio
ns
for
rem
itti
ng
(see
Ag
arw
alan
dH
oro
wit
z2
00
2)
Hea
lth
of
rem
itta
nce
par
tner
Th
esu
bje
ctiv
ew
ealt
ho
fth
ere
mit
tan
cep
artn
er.R
elev
ant
bo
thfo
rth
eq
ual
ity
of
insu
rer
the
par
tner
mig
ht
mak
ean
dan
ind
icat
or
of
rece
nt
hea
lth
sho
cks
Eld
est
chil
d/s
ibli
ng
An
ind
icat
or
equ
alto
1if
the
chil
dis
anel
des
tch
ild
,0
oth
erw
ise.
Eld
est
chil
dre
nm
ayh
ave
mo
re
resp
on
sib
ilit
ies
vis
-a-v
isth
eir
par
ents
than
oth
ers
An
ind
icat
or
var
iab
leeq
ual
to1
ifth
e
sib
lin
gis
the
eld
est
sib
lin
g.
Po
ten
tial
lyre
lev
ant
for
tran
sfer
rela
tio
nsh
ips
bet
wee
nth
ere
spo
nd
ent
and
the
sib
lin
g
Par
ent
has
som
esc
ho
oli
ng
Du
mm
yv
aria
ble
equ
alto
1if
the
par
ent
has
som
esc
ho
oli
ng
,0
oth
erw
ise
Sib
lin
g/c
hil
dli
ves
abro
ado
rin
city
Eq
ual
to1
ifre
mit
tan
cep
artn
erli
ves
abro
ado
rin
aci
ty,
0o
ther
wis
e.R
elev
ant
for
insu
ran
cem
od
els
inan
agri
cult
ura
lec
on
om
yin
wh
ich
tho
sew
ho
liv
ein
citi
esan
d
abro
adh
ave
inco
me
less
corr
elat
edw
ith
resp
on
den
tsth
anth
ose
wh
oli
ve
nea
rby
480 J Fam Econ Iss (2011) 32:473–492
123
Thus, as a parent ages, s/he is likely to increase remit-
tances, and parents in better health will remit less. Under
the co-insurance/income pooling hypothesis, we are likely
to observe remittance flows in each direction. Therefore,
there will be a positive association between the likelihood
that a parent remits to the respondent, and the likelihood
that a respondent remits to his/her parent. This variable
might also capture traditional gift-sharing motivations
(Mauss 1990).
If remittances from a parent are insurance payouts, a
respondent is more likely to receive or to receive more
remittances as wealth decreases. Having suffered from a
health shock will increase remittance flows under this
motivation.
The data permit the analysis of one service provided by
the respondent for their parents: that of looking after a
sibling. If remittances are payment for this service then
remittances or likelihood of receiving remittances will be
higher for respondents who reported looking after a sibling.
If respondents remit to parents for altruistic motivations,
the number of heirs (potential remitters) will have a neg-
ative impact on remittances as other people are likely to
ensure that the parents have a good quality of life (Agarwal
and Horowitz 2002). Respondents will be less likely to
remit or remit less to their parents if their parents are in
better health.
Econometric Results and Discussion
Respondent–Parent Transfer Flows
Regression results for respondent–parent transfers are
reported in Table 9. Net transfers from parents are
decreasing in the asset index, that is, those with lower
wealth receive more net transfers from their parents than
wealthier individuals. This is consistent with parental
altruism. The Tobit model in column 5 reveals that this is
largely driven by the fact that wealthier individuals are
more likely to give to their parents than their poorer
counterparts, with non-farm assets being highly significant
at the 1% level. Van Dalen et al. (2005) and De la Briere
et al. (2002) show that this is consistent with both altruism
and inheritance motivations.
Net transfers from parents are higher for individuals
who reported suffering from a health problem during the
previous month. This is consistent with both altruistic
motivations and insurance payouts from parents to chil-
dren. However, this is significant only at the 10% meaning
it is difficult to draw strong conclusions from this result.
Although there are no data on whether or not a parent
recently suffered from a health problem, a similar result
can be seen with respect to parents’ general health withTa
ble
7co
nti
nu
ed
Par
ents
Ch
ild
ren
Sib
lin
gs
Par
ent
liv
esin
sam
ev
illa
ge
Du
mm
yv
aria
ble
equ
alto
1if
par
ent
liv
esin
sam
ev
illa
ge
asre
spo
nd
ent
Nu
mb
ero
fh
eirs
par
ent
has
Rel
evan
tfo
rin
her
itan
cem
oti
vat
ion
s
Rem
itta
nce
par
tner
fem
ale
An
ind
icat
or
var
iab
leeq
ual
to1
ifth
ep
ote
nti
alre
mit
tan
cep
artn
eris
fem
ale,
0o
ther
wis
e
Ch
ild
mar
ried
Eq
ual
to1
ifth
ech
ild
ism
arri
ed,
0
oth
erw
ise.
Mar
riag
ere
du
ces
lin
ks
wit
hth
eh
ou
seh
old
of
ori
gin
and
is
lik
ely
tocr
eate
oth
erre
spo
nsi
bil
itie
s
mak
ing
this
anu
sefu
lco
ntr
ol
var
iab
le
Join
tch
arac
teri
stic
s
Mat
rili
nea
l/p
arti
lin
eal
her
itag
eA
nim
po
rtan
tin
dic
ato
rv
aria
ble
ina
cou
ntr
yin
wh
ich
wea
lth
flo
ws
and
lik
elih
oo
do
fin
her
itan
ced
iffe
rb
ytr
ibe
Vil
lag
ein
dic
ato
rsD
um
my
vil
lag
ev
aria
ble
sar
ein
clu
ded
toca
ptu
resy
stem
atic
beh
avio
ral
dif
fere
nce
sb
etw
een
vil
lag
es
J Fam Econ Iss (2011) 32:473–492 481
123
respondents being less likely to remit to parents who are in
better health.
The results show that individuals who receive transfers
from their parents are more likely to remit in turn, and vice
versa. A large proportion of transfer relationships are
therefore bidirectional. This result is consistent with trans-
fers serving as insurance or income pooling as part of a
survival strategy. This result is a strong one with highly
significant coefficients in all cases, and is in line with
‘‘balanced reciprocity’’ as mutual insurance discussed by
Platteau (1997).
Net transfers from parents who live in the same village
as the respondent tend to be lower than for those who live
in another village. The Probit and Tobit models in columns
4 and 5 reveal that this is due to the fact that respondents
are more likely to remit to parents who live in the same
village. In addition, the coefficients on these models are
highly significant. This could be as a result of transaction
costs of remitting to parents who live further afield.
Equally, social pressure to support parents can be stronger
when both live in the same village.
The Probit models in columns 2 and 4 reveal that the
more heirs a parent has, the more likely respondents are to
remit to them and the more likely they are to receive from
them. It is interesting that this appears to influence only
likelihood of remitting or receiving and not the amount, as
the Tobit model coefficients are insignificant. However, the
Probit model coefficients are significant only at the 10%
level for sending and the 5% level for receiving. Although
there is no information on parental wealth, this is tentative
evidence of an inheritance motivation for remitting.
The respondent’s education is positive and significant at
the 1% level in both the Probit and Tobit for receiving
transfers from their parents. Better educated respondents
are both more likely to receive transfers from their parents,
and receive more from them. This is in line with the theory
that parents ‘‘insure’’ themselves with children who make
the best insurers such as those with better education.
Interestingly, parents who have some schooling are also
both more likely to send transfers to the respondents—their
children (as revealed by the Probit model in column 2) and
remit more (see Tobit model in column 3). Combined with
the fact that parents remit to better educated respondents,
this might be a further indication of a balanced reciprocity
amongst those who are more likely to be able to help each
other. This has the interesting and important economic
consequences that those who are able to ‘‘insure’’ each
other, do so, whilst those who are not able to insure others
might lack insurance themselves in the event of a shock,
causing them to remain in a poverty trap.
Motivations for respondent–parent transfer flows are
mixed. Nonetheless, there is strong support for altruistic
giving in both directions. In addition, the results are also
consistent with co-insurance motivations.
The low pseudo r-squared are disappointing but not too
much cause for concern. Such results are normal in the
literature (e.g. Cao 2006), and often for Probit and Tobit
models in other fields.
Table 8 Discriminating hypotheses from respondent-parent transfer perspective
Altruism Co-insurance
(premium)
Co-insurance
(indemnity)
Inheritance Implicit payment
for services
Parent’s motivations for remitting to Respondent
Respondent’s wealth Negative Positive Negative – No direct impact
Respondent’s general health Negative Positive No direct impact – No direct impact
Respondent suffered health shock Positive No direct impact Positive – No direct impact
Respondent looks after sibling No direct impact No direct impact No direct impact – No direct impact
Respondent sends transfers No direct impact Positive No direct impact – Positive
Parent’s age No direct impact Positive No direct impact – No direct impact
Parent’s health No direct impact Negative No direct impact – No direct impact
Parent lives in same village No direct impact Negative No direct impact – No direct impact
Respondent’s motivation for remitting to parent
Parent’s age No direct impact Negative No direct impact Positive –
Parent’s heirs Negative No direct impact No direct impact Positive/Negative –
Parent’s health Negative Positive Negative/No direct
impact
Negative –
Parent’s schooling No direct impact Positive No direct impact Positive –
Parent sends transfers No direct impact Positive No direct impact No direct impact –
Parent lives in same village No direct impact Negative No direct impact ? –
482 J Fam Econ Iss (2011) 32:473–492
123
Table 9 Respondent–Parent transfer flows
Net transfers received
from parent
Received transfers
from parent
Value of transfers
from parent
Sent parent
transfers
Value of transfers
sent to parent
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Income 0.007 -0.000 -0.022 -0.000 -0.029***
(1.347) (-0.986) (-1.257) (-1.556) (-3.025)
Asset index (farm) -22.725 -0.069 -16.322 0.063 29.257**
(-1.557) (-1.644) (-0.936) (1.427) (2.342)
Asset index (non-farm) -37.262** 0.030 -2.441 0.061 47.103***
(-2.561) (0.920) (-0.150) (1.637) (3.025)
Education 1.738 0.062*** 30.682*** -0.036* 9.231
(0.188) (3.605) (3.364) (-1.858) (1.062)
Age -1.980 -0.010 4.766 0.000 5.830
(-0.220) (-0.304) (0.386) (0.012) (0.572)
Age square 0.044 -0.000 -0.137 -0.000 -0.108
(0.324) (-0.220) (-0.729) (-0.046) (-0.677)
Maleb -7.428 -0.480*** 29.235 0.682*** 247.020***
(-0.154) (-3.626) (0.467) (4.949) (5.339)
Married -478.170 0.348 -502.684 0.659* 123.774
(-1.618) (1.124) (-1.376) (1.926) (1.202)
Household size -16.636* -0.049* -8.628 -0.004 19.163*
(-1.830) (-1.768) (-0.709) (-0.146) (1.778)
Respondent eldest child 8.789 0.074 25.688 0.051 20.249
(0.176) (0.639) (0.446) (0.430) (0.413)
Health (1–10) 6.748 0.033 11.142 0.040 -0.430
(0.728) (1.100) (0.969) (1.229) (-0.039)
Health problem in last month 53.859* 0.195 60.140 -0.046 -52.851
(1.679) (1.610) (1.436) (-0.345) (-1.373)
Sibling in house 64.391 0.063 -18.262 0.114 -83.139*
(1.435) (0.477) (-0.323) (0.702) (-1.719)
Given transfers to parent 0.661*** 245.675***
(6.618) (3.857)
Received transfers from parent 0.653*** 98.823***
(6.651) (2.776)
Parent’s age 3.585 0.022** 9.391** -0.010 -1.997
(1.312) (2.438) (1.984) (-1.101) (-0.594)
Parent’s age square -0.041 -0.000** -0.123** 0.000 0.020
(-1.296) (-2.544) (-2.232) (1.445) (0.547)
Parent’s health -0.879 0.017 4.762 -0.069*** -9.482
(-0.136) (0.857) (0.546) (-3.213) (-1.309)
Parent lives in same village -84.438** 0.144 -10.092 0.554*** 119.718***
(-2.324) (1.483) (-0.225) (4.795) (3.238)
Parent has some schooling 39.246 0.223** 111.516*** -0.006 5.757
(1.054) (2.124) (2.646) (-0.062) (0.129)
Number of heirs parent has 0.873 0.032** 7.120 0.030* 4.828
(0.212) (1.967) (1.146) (1.662) (0.901)
Motherb -151.606** 0.158 -74.225 0.126 95.845*
(-2.472) (1.248) (-0.954) (1.020) (1.948)
Matrilineal heritagea -26.418 0.122 22.384 -0.790*** -123.757
(-0.392) (0.642) (0.308) (-4.047) (-1.549)
J Fam Econ Iss (2011) 32:473–492 483
123
Respondent–Children Transfer Flows
We report regressions for respondent–child transfers in
Table 10. Net transfer receipts from children are decreas-
ing in income and this is significant at the 1% level. The
Probit and Tobit models in columns 2 and 3 reveal that
respondents do not receive more money from children as
their income increases. Instead, the Tobit model in column
5 shows that this result appears to be driven by the fact that
respondents with higher income send their children more
money. The theoretical models discussed above indicate
that this result is consistent with altruism towards their
children on the part of the respondent. Interestingly, the
coefficient on income in the Probit in column 4 is insig-
nificant suggesting that income does not influence whether
or not the respondent sends transfers to their children, but
only the amount sent.
Although income is significant in some models, wealth
is insignificant with only one exception: farming wealth is
positive and significant at the 5% level in column 5. This
suggests that those with greater farm assets send more to
their children than others. However, farm wealth appears to
influence only the amount of transfers sent, and not the
decision to send or not, since the coefficients on wealth in
the Probit model in column 4 remain insignificant.
Although children do not appear to ‘‘insure’’ health
shocks, net transfer receipts and probability of receiving
transfers from children is increasing in respondents’ health.
In addition, the better the child’s health and this is signif-
icant at the 5% level. That is, the lower net transfers
respondents receive. This is consistent with the hypothesis
that children choose to insure themselves with their par-
ents; parents in better health make better insurers, and
children in better health are less likely to require their
parents’ insurance.
Respondents receive less net transfers from their chil-
dren if they look after their grandchildren. Although the
data do not permit me to know to which son/daughter the
grandchild belongs, the evidence indicates transfers are not
a payment for this service. Indeed, respondents who look
after their grandchildren receive less than others. Although
potentially initially surprising, this should not be unex-
pected within the Malawian context where 14% of the
population is HIV positive (Morah 2007). HIV dispropor-
tionally affects those of working and child-bearing age and
grandparents are often left to care for the children (Conroy
et al. 2006). This result could therefore be interpreted as
evidence of altruism of grandparents towards their grand-
children and is consistent with theoretical models which
posit dynastic altruism
Table 9 continued
Net transfers received
from parent
Received transfers
from parent
Value of transfers
from parent
Sent parent
transfers
Value of transfers
sent to parent
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Patrilineal heritagea -6.595 0.199 -102.043 0.008 -105.696
(-0.097) (1.075) (-1.011) (0.042) (-1.472)
Northern village (Mwankhunikira)c -58.019 -0.485* -34.805 -0.511* 26.398
(-0.568) (-1.803) (-0.331) (-1.828) (0.214)
Northern village (Mwahenga)c -16.800 -0.501** 12.386 -0.134 68.606
(-0.180) (-2.222) (0.114) (-0.580) (0.727)
Central village (Mkanda)c 25.401 -0.150 -1.067 -0.415** -81.220
(0.509) (-0.900) (-0.017) (-2.406) (-1.397)
Mother—Female transfer flowb 164.198** -0.211 80.066 0.370** 10.353
(2.486) (-1.403) (1.044) (2.414) (0.195)
Constant 385.078* -1.373** -255.275 -0.703 -365.624
(1.658) (-1.973) (-0.848) (-0.900) (-1.454)
Standard error of regression 536.491*** 450.169***
(6.297) (8.582)
N 1083 1083 1083 1083 1083
(Pseudo) r2 0.083 0.114 0.013 0.161 0.016
Clusters 483 483 483 483 483
Goodness of fit F: 2.889 Chi 2: 148.590 F: 1.863 Chi 2: 203.302 F: 3.724
Notes: t-stats in parentheses below coefficients. *, ** and *** indicate significance at 10%, 5% and 1%, respectivelya Excluded dummy is Mixed Heritage (Chewa tribe); b Indicated dummy variables capture all transfer flow relationships between them
(excluded Father-Male transfer flow); c Excluded village dummy is Southern Village (Kalembo)
484 J Fam Econ Iss (2011) 32:473–492
123
Table 10 Respondent-Children transfer flows
Net transfers received
from child
Received transfers
from child
Value of transfers
from child
Gave transfer
to child
Value of transfer
sent to child
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Income -0.063*** -0.000 -0.011 0.000 0.054***
(-5.450) (-0.350) (-0.369) (0.578) (3.217)
Asset indexd -31.560 -0.007 -31.707
(-0.967) (-0.083) (-0.967)
Asset index (farm)d 0.116 63.068**
(1.436) (2.093)
Asset index (non-farm)d -0.005 7.730
(-0.076) (0.282)
Education 3.815 0.064* 29.003** -0.006 9.985
(0.374) (1.930) (2.154) (-0.197) (0.795)
Age 0.769 -0.002 -8.666 -0.047 -10.963
(0.035) (-0.022) (-0.355) (-0.560) (-0.299)
Age square 0.023 -0.000 -0.009 0.000 -0.015
(0.098) (-0.182) (-0.037) (0.205) (-0.039)
Maleb -99.572 0.159 116.401 0.411 199.189*
(-1.007) (0.660) (0.881) (1.394) (1.722)
Married 54.570 -0.715** -63.622 1.133*** 404.216**
(0.636) (-2.170) (-0.566) (2.785) (2.299)
Household size 15.529 0.078 38.944 -0.033 -3.903
(0.809) (1.541) (1.572) (-0.718) (-0.163)
Health (1-10) 35.574** 0.088** 34.675** -0.016 -28.947
(2.470) (2.308) (2.081) (-0.304) (-1.394)
Health problem in last month 66.623 0.199 167.573 -0.199 -74.805
(0.829) (1.066) (1.457) (-1.073) (-0.847)
Grandchild in house -186.738** -0.196 -101.821 0.112 143.362
(-2.185) (-0.722) (-0.995) (0.427) (1.290)
Given transfer to child -0.087 -73.483
(-0.531) (-1.121)
Received transfers from child 0.327** 1.096
(2.149) (0.018)
Number of children -7.365 -0.215*** -35.756** 0.078 10.380
(-0.351) (-3.437) (-2.210) (1.511) (0.372)
Number of children remitting 57.762** 0.841*** 224.886***
(2.251) (9.961) (3.471)
Number children remitting*Asset index -6.818 -0.016 9.228
(-0.469) (-0.283) (0.574)
Child’s age -2.707 0.073* 24.633 -0.033 8.052
(-0.198) (1.736) (1.474) (-0.926) (0.515)
Child’s age square 0.046 -0.001* -0.209 0.000 -0.078
(0.381) (-1.712) (-1.408) (1.046) (-0.559)
Eldest child -12.499 0.412* 115.390 0.152 67.992
(-0.201) (1.895) (1.517) (0.927) (0.961)
Child’s health -24.444** 0.024 -20.433 -0.003 27.701*
(-2.315) (0.441) (-1.083) (-0.075) (1.756)
Child lives city/abroad -11.353 -0.236 -28.035 -0.223 -9.205
(-0.147) (-1.255) (-0.316) (-1.214) (-0.123)
J Fam Econ Iss (2011) 32:473–492 485
123
The coefficients on number of children and number of
children remitting in the Probit model in column 2 are of
particular interest. The more children a respondent has, the
less likely s/he is to receive transfers from any one of the
children with this variable being significant at the 1% level.
Agarwal and Horowitz (2002) show that this can be
interpreted as evidence of altruism as the more children
there are to look after the parents the less responsibility
there is on any one of them to do so. In contrast, the more
children there are actually remitting, the more likely the
respondent is to receive transfers from any one of them.
This coefficient is also highly significant at the 1% level
and suggests that there is competition amongst children to
be seen to assist their parents.
This is consistent with inheritance motivations to remit
since, in Malawi, inheritance does not automatically go to
the eldest child, but is decided by a committee of surviving
senior relatives on the basis of who is seen to have fulfilled
their duty towards their parents (Takane 2007). To assess
this hypothesis further we interact the number of children
remitting with the asset index. We expect a positive coef-
ficient on this since the children will be prepared to com-
pete more strongly the more they stand to benefit. This
coefficient is however, insignificant for the pooled sample.
It does however turn significant for male transfer receipts
when male and female are analyzed separately (results
available on request). On average, the male asset index is
significantly higher than that for females.
Respondent–children transfers show evidence of chil-
dren insuring their parents against income (but not health)
shocks, parents showing altruism towards grandchildren,
children exhibiting both altruism and inheritance motiva-
tions. No one motivation appears to dominate in this case
however.
Table 10 continued
Net transfers received
from child
Received transfers
from child
Value of transfers
from child
Gave transfer
to child
Value of transfer
sent to child
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Child married 173.701** 0.813*** 332.894** -0.342* -213.014**
(2.315) (3.127) (2.022) (-1.917) (-2.360)
Child daughterb -48.032 -0.620** -199.928** 0.012 34.566
(-0.709) (-2.455) (-2.383) (0.062) (0.383)
Matrilineal heritagea -4.333 0.543** 147.043 -0.307 -26.153
(-0.049) (2.098) (1.529) (-0.948) (-0.204)
Patrilineal heritagea 150.050* 0.288 282.936* 0.339 42.837
(1.709) (1.060) (1.917) (1.156) (0.477)
Northern village (Mwankhunikira)c -543.996*** -0.873 -448.529** -0.099 359.716*
(-2.979) (-1.571) (-1.980) (-0.232) (1.660)
Northern village (Mwahenga)c -138.781 -0.228 -126.777 -0.469 10.891
(-1.248) (-0.841) (-0.902) (-1.297) (0.111)
Central village (Mkanda)c -84.811 -0.154 -64.259 -0.083 74.216
(-1.170) (-0.591) (-0.577) (-0.296) (0.846)
Daughter–Mother transfer flowb -49.813 0.372 59.683 0.333 99.883
(-0.484) (1.218) (0.568) (1.164) (0.737)
Constant -327.070 -2.936 -1140.232* 1.280 -223.096
(-0.694) (-1.540) (-1.786) (0.625) (-0.290)
Standard error of the regression 510.692*** 508.374***
(3.323) (6.952)
N 430 430 430 430 430
(Pseudo) r2 0.1976 0.394 0.051 0.101 0.023
Clusters 144 144 144 144 144
Goodness of fit F: 10.345 Chi 2: 251.285 F: 2.981 Chi 2: 75.163 F: 6.079
Notes: t-stats in parentheses below coefficients. *, ** and *** indicate significance at 10%, 5% and 1%, respectivelya Excluded dummy is Mixed Heritage (Chewa tribe); b Indicated dummy variables capture all transfer flow relationships between them
(excluded Son-Male transfer flow); c Excluded village dummy is Southern Village (Kalembo); d A pooled asset index is included in some
regressions due to the Asset * Remitters interaction term. Results do not differ substantially if separate farm and non-farm asset indexes are
entered
486 J Fam Econ Iss (2011) 32:473–492
123
Respondent-Sibling Transfer Flows
Results from Respondent-Sibling regressions are presented
in Table 11. Net transfers from siblings are decreasing in
asset wealth of the respondent but increasing in income.
The probit regressions show that this is driven by transfers
sent by respondents to their siblings. Respondents appear to
share their (long term asset) wealth with siblings, but not
their short term income.
Net transfers from siblings increase as the respondents’
health declines suggesting siblings remit for altruistic
motives. In addition, siblings are more likely to remit to
respondents who have recently suffered from a health
problem. This is consistent with insurance payouts or
altruistic motivations. Interestingly, respondents who
reported suffering from a recent health problem were also
more likely to send transfers to their siblings. This could be
an indication that siblings provided help to sick respondents,
but that respondents ‘‘repaid’’ the help. The distinction
between gifts and loans may not be clear. Udry (1990) for
example finds that repayment of loans in rural Northern
Nigeria is conditional upon the borrower’s and lender’s
economic situation. A similar situation may exist in rural
Malawi. Both respondents and their siblings are more likely
to remit the better their own health is, and both are more
likely to send to each other the worse the other’s health is.
As with parents, there appears to be a great deal of
reciprocity with those giving to siblings more likely to
receive from them, and vice versa. This is consistent with
co-insurance/income pooling motivations for remitting.
Net transfers from siblings residing in a city or abroad are
higher than from those residing in rural Malawi. The Tobit
model indicates that total transfers received are higher from
siblings living in cities or abroad, but transfers sent to them
(and the likelihood of remitting to them) are lower. Since
such siblings are likely to earn more, this could be evidence
of altruism (they give because they can afford to), or of a
family survival strategy (sending some people to work in the
city), or repayment of implicit loans to fund the migration.
Alternatively the transaction costs involved in such transfers
are prohibitive for rural Malawians but not for those residing
in cities or abroad. Without additional information, it is not
possible to untangle this further.
Respondent-Sibling transfer flows reveal a strong
altruistic component. In addition, there is evidence of co-
insurance. Additional information is required to understand
the rural–urban linkage discussed.
Transfers From a Household Perspective
It is useful for comparison to re-run the basic regressions at
a household level in order to permit comparison with the
previous results. This is shown in Table 12 for net remit-
tances received from each of the three sets of remittance
partners. Several choices have to be made regarding
inclusion of variables for such an analysis.
Firstly, for the respondent household we are able to
include either male or female characteristics for the
household as a whole or else include both. In the latter
case, any single households or those for which data was not
collected on one of the partners will be lost from the
sample, however, it will make it possible to study whether
it is male or female characteristics which drive transfer
behavior. In the former case, it is not obvious which set of
characteristics should be included and excluding relevant
variables risks leading to missing variable bias. We have
chosen to include both male and female characteristics
leading to a reduction in the number of observations, but
enabling additional information to be extracted from the
study.
Secondly, regarding the transfer partner households, we
are able to include all remittance transfers with relevant
characteristics attached to the partner. Alternatively, it is
possible to sum all remittance flows and assume the
characteristics of, for example, the eldest sibling or child.
We have opted for the former option in the results
reported.
Children are a special case as they may remit to either
parent whereas, (at least in our data), parents and siblings
tend to remit to their own children/siblings. At the house-
hold level therefore, total remittances have been summed
and a transfer to the mother is no longer considered as
different to that of a remittance to the father.
Male and Female Characteristics
Before comparing household level results, it is instructive
to note the different impacts of male and female charac-
teristics. These results are particularly interesting for chil-
dren and siblings. Net remittances from children fall their
mother’s income rises, but do not appear to be influenced
by father’s income. This is some evidence that children
behave altruistically towards their mother.
In contrast, net transfers from siblings do not respond to
the female’s income, but rather respond positively to the
male’s income. As the male respondent earns more, the
household is likely to receive more remittances from sib-
lings. This could be the result of a co-insurance premium.
In addition, siblings’ transfers decrease as the male’s
education increases and decrease as his health improves.
Increased remittances when sick may be the result of a
co-insurance pay-out. It is not clear why siblings appear to
exhibit insurance behavior with regards to income with the
male but not the female.
J Fam Econ Iss (2011) 32:473–492 487
123
Table 11 Respondent-Sibling transfer flows
Net transfers received
from sibling
Received transfers
from sibling
Value of transfers
from sibling
Sent transfers
to sibling
Value of transfers
sent to sibling
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Income 0.011*** 0.000 0.010 -0.000*** -0.033***
(2.816) (0.888) (0.974) (-3.451) (-3.068)
Asset index (farm) -6.962 0.035 13.948 -0.011 10.841
(-0.896) (1.505) (1.423) (-0.455) (0.736)
Asset index (non-farm) -10.916* -0.043* -14.806 0.075*** 33.674***
(-1.784) (-1.833) (-1.479) (3.382) (2.795)
Education -2.171 -0.009 3.455 0.031** 16.687***
(-0.634) (-0.778) (0.561) (2.506) (2.824)
Age 0.447 -0.030 -16.588 -0.004 -7.566
(0.094) (-1.353) (-1.576) (-0.167) (-0.892)
Age square -0.052 0.000 0.138 0.000 0.171
(-0.879) (0.855) (1.060) (0.771) (1.500)
Maleb -32.353 0.019 26.672 0.249*** 128.396***
(-1.195) (0.199) (0.584) (2.739) (3.296)
Married -43.964 -0.085 -65.774 0.127 57.637
(-0.819) (-0.387) (-0.610) (0.504) (0.580)
Household size -1.218 -0.017 4.423 -0.003 9.626
(-0.257) (-0.909) (0.535) (-0.155) (1.366)
Respondent eldest child -10.254 -0.116 -0.427 0.038 43.107
(-0.348) (-1.234) (-0.010) (0.428) (0.950)
Health (1–10) -9.204** 0.006 -3.023 0.058*** 25.751***
(-2.203) (0.272) (-0.325) (2.685) (2.925)
Health problem in last month -8.812 0.194** 47.270 0.154* 47.453
(-0.481) (2.152) (1.308) (1.771) (1.480)
Nephew/Niece in house 36.670 0.008 31.465 0.010 -16.615
(1.278) (0.072) (0.550) (0.108) (-0.460)
Given transfers to sibling 0.563*** 199.164***
(9.175) (5.304)
Received transfers from sibling 0.585*** 187.620***
(9.338) (6.033)
Sibling’s age 8.211*** 0.100*** 44.126*** -0.050*** -14.441***
(3.627) (7.235) (4.960) (-3.977) (-2.638)
Sibling’s age square -0.065** -0.001*** -0.472*** 0.000*** 0.115*
(-2.349) (-5.890) (-4.514) (2.990) (1.775)
Eldest sibling -24.590 0.055 -19.817 -0.046 -13.680
(-1.294) (0.633) (-0.584) (-0.513) (-0.364)
Sibling’s health 6.461* 0.045*** 18.981** -0.048*** -15.836**
(1.835) (2.767) (2.530) (-3.011) (-2.163)
Sibling lives abroad or in city 82.477*** 0.071 109.691*** -0.610*** -200.600***
(4.601) (0.972) (3.075) (-8.474) (-4.263)
Sisterb 3.352 -0.109 -75.184** -0.146* -85.267**
(0.177) (-1.337) (-2.113) (-1.796) (-2.450)
Matrilineal heritagea 59.033* -0.007 18.968 -0.123 -98.337
(1.839) (-0.046) (0.285) (-0.822) (-1.602)
Patrilineal heritagea 34.014* 0.213* 102.160** -0.040 -40.518
(1.804) (1.768) (2.132) (-0.317) (-0.839)
488 J Fam Econ Iss (2011) 32:473–492
123
Comparing Household with Individual Transfers
The major changes can be seen with respect to transfers
between respondents and their children with little differ-
ences being observed with respect to other transfer
relationships.
Table 10 shows that children remit more net transfers
to respondents and respondents’ income declines.
Table 12 shows that this is particularly driven by the
mother’s income. In contrast, the household asset index is
insignificant at the individual level. This turns signifi-
cantly negative when remittances are viewed from the
household perspective. This suggests that children con-
sider more the household level of wealth and that it
matters less to whom they remit. The results suggest that
children transfer more to their parents the poorer the
parents are. Other results do not differ significantly from
the previous analysis.
The household level analysis does not show significant
changes regarding sibling characteristics. However, the
results reveal that it is male income that is the key (posi-
tive) driver of net transfers and siblings’ transfers respond
more to male sickness than female sickness. The household
non-farm asset index remains negative and significant
suggesting that siblings transfer more to respondents with
fewer non-farm assets.
Conclusion
This paper has aimed to disentangle motivations for
remitting by analyzing transfers separately for transfer
relationships between the respondents and their parents,
siblings and children. However, the results confirm the
conclusions of all papers to date, that is, that transfer
motivations are difficult to disentangle. One motivation to
remit does not preclude another, even at an individual
level.
A word of warning is in order when comparing the
results from the different transfer relationships due to the
different number of observations in each category. None-
theless, the results are robust to numerous alterations to
model specification, and the large number of significant
variables in each specification can be taken as a sign that,
despite the difference in observations, the results are
reliable.
All transfer relationships showed evidence of altruism.
For example, net transfers from parents and siblings
increase as respondents’ asset indexes decline. However,
there is also strong evidence of co-insurance, particularly
between respondents and their parents, and respondents
and their siblings with both of these groups more likely to
give if they receive and vice versa. In addition, children
appear to insure respondents’ short term income and
Table 11 continued
Net transfers received
from sibling
Received transfers
from sibling
Value of transfers
from sibling
Sent transfers
to sibling
Value of transfers
sent to sibling
(1) Robust OLS (2) Probit (3) Tobit (4) Probit (5) Tobit
Northern village (Mwankhunikira)c 31.070 -0.091 -24.076 0.014 -15.710
(0.775) (-0.538) (-0.331) (0.088) (-0.224)
Northern village (Mwahenga)c -27.945 -0.154 -82.149 -0.071 -4.788
(-0.869) (-1.055) (-1.315) (-0.484) (-0.075)
Central village (Mkanda)c 17.883 -0.196 -78.806 -0.101 -71.220
(0.751) (-1.560) (-1.571) (-0.776) (-1.248)
Sister-Female transfer flowb -11.143 0.077 47.816 0.256** 103.045**
(-0.385) (0.670) (0.943) (2.224) (2.147)
Constant -102.571 -2.261*** -1025.598*** 0.159 -134.272
(-0.813) (-4.402) (-3.978) (0.306) (-0.649)
Standard error of regression 474.482*** 468.252***
(5.721) (6.335)
N 2876 2876 2876 2876 2876
(Pseudo) r2 0.041 0.076 0.016 0.087 0.015
Clusters 540 540 540 540 540
Goodness of fit F: 2.562 Chi 2: 182.937 F: 2.846 Chi 2: 244.848 F: 2.771
Notes: t-stats in parentheses below coefficients. *, ** and *** indicate significance at 10%, 5% and 1%, respectivelya Excluded dummy is Mixed Heritage (Chewa tribe); b Indicated dummy variables capture all transfer flow relationships between them
(excluded Brother-Male transfer flow); c Excluded village dummy is Southern Village (Kalembo)
J Fam Econ Iss (2011) 32:473–492 489
123
parents and siblings increase their net transfers to respon-
dents who have suffered from a recent health shock. There
is also evidence of inheritance motivations in transfers
from children to respondents.
These mixed results could be the result of a large degree
of heterogeneity in the population with regards to familial
transfers, with different individuals and families having
different motivations. Alternatively, or in addition, it is
possible for one person to have more than one single
motivation when choosing to give. These facts mean it is
unlikely to be able to conclude in favor of one particular
motive.
One interesting observation is that individual and
household level analyses reveal similar results with respect
to remitter characteristics. However remitters may respond
more to male or female respondents suggesting that, at
least in part, the individual (and not just the household)
Table 12 Household level transfer flows
Children Siblings Parents
Household characteristics—female
Income -0.048*** 0.002 0.002
(-2.870) (0.818) (0.236)
Education -9.831 -2.625 3.311
(-0.766) (-1.252) (0.337)
Age -74.230 2.320 18.363
(-1.130) (0.459) (1.074)
Age square 0.819 -0.045 -0.205
(1.135) (-0.609) (-0.849)
Health (1–10) -20.766 -4.060 12.563
(-0.854) (-0.785) (1.277)
Health problem in last month 46.066 -3.234 25.609
(0.469) (-0.149) (0.572)
Household characteristics—male
Income 0.014 0.015*** 0.007
(0.328) (3.410) (0.774)
Education -8.767 -3.611* -5.689
(-0.779) (-1.823) (-0.928)
Age -0.201 -0.917 -27.428**
(-0.010) (-0.317) (-2.157)
Age square -0.082 -0.006 0.331**
(-0.356) (-0.160) (2.204)
Health (1–10) 6.463 -6.852** 15.358
(0.342) (-1.984) (1.308)
Health problem in last month 124.370 -3.742 60.962
(0.937) (-0.261) (1.352)
Other household characteristics
Respondent maleb -67.822 -48.514** 20.050
(-0.820) (-2.385) (0.369)
Respondent eldest child 5.806 -19.528
(0.265) (-0.357)
Household size 8.282 -3.669 -26.572**
(0.353) (-0.848) (-2.418)
Asset index (Farm) -99.918*** -8.733 -14.977
(-2.646) (-0.948) (-0.705)
Asset index (non-farm) 43.471 -9.226* -36.024**
(1.428) (-1.934) (-2.382)
Household looks after remittance
partner’s child
-241.557** 28.742 101.770**
(-2.111) (1.576) (2.087)
Number of children -8.325
(-0.303)
Number of children remitting 61.866***
(2.651)
Number of children remitting*asset index -5.951
(-0.310)
Remittance partner characteristics
Age 19.320 6.895*** 3.686
(1.173) (3.531) (1.102)
Age square -0.133 -0.062*** -0.050
(-0.886) (-2.719) (-1.250)
Eldest child -129.704 -9.695
(-1.274) (-0.497)
Health (1–10) -31.796* 3.518 -7.308
(-1.885) (1.101) (-0.919)
Table 12 continued
Children Siblings Parents
Lives in city/abroad 93.958 64.933***
(1.144) (4.375)
Lives in same village as household -84.375**
(-2.145)
Married 241.307**
(2.494)
Has some schooling 62.736
(1.378)
Number of heirs -1.689
(-0.368)
Remittance Partner Femaleb -136.496 2.713 -
185.314***
(-1.551) (0.139) (-2.589)
Joint Characteristics
Matrilineal heritagea -61.175 53.144* -38.063
(-0.515) (1.939) (-0.414)
Patrilineal heritagea 165.993 52.508** 10.654
(1.305) (2.436) (0.124)
Northern village (Mwankhunikira)c -301.322** 53.047 -81.755
(-2.102) (1.264) (-0.698)
Northern village (Mwahenga)c -169.410 0.342 -93.388
(-1.560) (0.009) (-0.814)
Central village (Mkanda)c -39.633 44.185 4.793
(-0.484) (1.552) (0.074)
Female to Female Transferb -26.801 -17.662 203.670***
(-0.236) (-0.746) (2.699)
Constant 1805.235 -89.656 36.664
(1.261) (-0.844) (0.158)
N 286 2283 846
r2 0.251 0.039 0.095
F 3.570 3.172 2.364
Notes: t-stats in parentheses below coefficients. *, ** and *** indicate significance at
10%, 5% and 1%, respectivelya Excluded dummy is Mixed Heritage (Chewa tribe); b Indicated dummy variables
capture all transfer flow relationships between them; c Excluded village dummy is
Southern Village (Kalembo)
490 J Fam Econ Iss (2011) 32:473–492
123
they are remitting to matters. One highly notable exception
to this is that children appear to remit to the household
rather than the individual when considering household
wealth. That is, if children support parents with low levels
of wealth they may give to either parent and it is necessary
to conduct the analysis at a household and not individual
levels to observe this result.
Although efforts were made to ensure the reliability of
survey responses, this cannot always be guaranteed. In
particular, it can be difficult to obtain accurate information
on income and wealth, with this likely to be somewhat
underestimated in Malawi for cultural reasons.
This study has contributed to the literature analyzing
motivations for remitting by extending the analysis to
different transfer relationships within the family unit with
the understanding that different people have different
motivations for remitting. In addition, we have been able to
study transfer flows in both directions of the relationship,
something the data used in such research rarely allows
users to do. Finally we have used a previously unused data
set to examine the issue in a previously unstudied country.
The complexity of gift exchange in a developing context
is clearly shown in this study where, although differences
in motivations for remitting can be discerned between
different groups, no single motivation can be ascribed to
each group. This study has found evidence of altruism, co-
insurance and, for transfers sent by their children to
respondents, inheritance motivations.
References
Agarwal, R., & Horowitz, A. W. (2002). Are international transfers
altruism or insurance? Evidence from Guyana using multiple-
migrant households. World Development, 30(11), 2033–2044.
Amuedo-Dorantes, C., & Pozo, S. (2004). Workers’ transfers and the
real exchange rate: A paradox of gifts. World Development,32(8), 1407–1417.
Amuedo-Dorantes, C., & Pozo, S. (2006). Transfers as insurance:
Evidence from Mexican immigrants. Journal of PopulationEconomics, 19, 227–254.
Azam, J. P., & Gubert, F. (2004). Those in Kayes: The impact of
transfers on their recipients in Africa. Institute d’Economie
Industrielle (IDEI) Working Paper Series, 308.
Bloom, D. E., & Stark, O. (1985). The new economics of labor
migration. American Economic Review, 75(2), 173–178.
Cao, H. (2006). Time and financial transfers within and beyond the
family. Journal of Family and Economic Issues, 27(2), 375–400.
Conroy, A. C., Blackie, M. J., Whiteside, A., Malewezi, J. C., &
Sachs, J. D. (2006). Poverty, AIDS and hunger: Breaking thepoverty trap in Malawi. New York: Palgrave MacMillan.
Cowley, E. T., Paterson, J., & Williams, M. (2004). Traditional gift
giving among pacific families in New Zealand. Journal ofFamily and Economic Issues, 25(3), 431–444.
Cox, D. (1987). Motives for private income transfers. Journal ofPolitical Economy, 95(3), 508–546.
Crumpler, H., & Grossman, P. J. (2008). An experimental test of
warm glow giving. Journal of Public Economics, 92(5–6), 1011–
1021.
De la Briere, B., Sadoulet, E., de Janvry, A., & Lambert, S. (2002).
The roles of destination, gender, and household composition in
explaining transfers: An analysis for the Dominican Sierra.
Journal of Development Economics, 68, 309–328.
Dercon, S. (1996). Risk, crop choice and savings: Evidence from
Tanzania. Economic Development and Cultural Change, 44(3),
485–514.
Fafchamps, M., Udry, C., & Czukas, K. (1998). Drought and saving in
West Africa: Are livestock a buffer stock? Journal of Develop-ment Economics, 55(2), 273–305.
Fan, C. S. (2005). Survival of the gene, intergenerational transfers and
precautionary saving. Journal of Development Economics, 77,
451–479.
Funkhouser, E. (1995). Remittances from international migration: A
comparison of El Salvador and Nicaragua. Review of Economicsand Statistics, 77(1), 137–146.
Grigorian, D. A., & Melkonyan, T. A. (2008). Microeconomic
implications of transfers in an overlapping generations model
with altruism and self-interest. IMF Working Paper WP/08/19.
Gupta, R., & Hegde, S. A. (2009). An exploratory study of financial
remittances among non-resident Indians in the United States.
Journal of the Family and Economic Issues, 30(2), 184–192.
Hayhoe, C. R., & Stevenson, M. L. (2007). Financial attitudes and
inter vivos resource transfers from older parents to adult
children. Journal of the Family and Economic Issues, 28(1),
123–135.
Horioka, C. Y. (2002). Are the Japanese selfish, altruistic, or
dynastic? Japanese Economic Review, 53, 26–54.
Ilahi, N., & Jafarey, A. (1999). Guestworker migration, transfers and
the extended family: Evidence from Pakistan. Journal ofDevelopment Economics, 58, 458–512.
Koh, S.-K., & MacDonald, M. (2006). Financial reciprocity and elder
care: Interdependent resource transfers. Journal of Family andEconomic Issues, 27(3), 420–436.
Laferrere, A., & Wolff, F. C. (2006). Microeconomic models of
family transfers. In S. Kolm & J. M. Ythier (Eds.), Handbook ofeconomics of giving, altruism and reciprocity 2 (pp. 889–969).
Amsterdam: Elsevier.
Lucas, R. E. B., & Stark, O. (1985). Motivations to remit: Evidence
from Botswana. Journal of Political Economy, 93(5), 901–918.
Mauss, M. (1990). The gift: The form and reason for exchange inarchaic societies, Original title: Essai sur le don, Presses
Universitaires de France, 1950, Trans. Hall, W.D., London:
Routledge.
Morah, E. U. (2007). Are people aware of their HIV-positive status
responsible for driving the epidemic in sub-saharan Africa?
The case of Malawi. Development Policy Review, 25(2), 215–
242.
Mtika, M. M., & Doctor, H. V. (2002). Matriliny, patriliny, and
wealth flows in rural Malawi. African Sociological Review, 6(2),
71–97.
Naufal, G. S. (2008). Why remit? The case of Nicaragua. Institute for
the Study of Labor Discussion Paper, IZA DP 3276.
Platteau, J. P. (1997). Mutual insurance as an elusive concept in
traditional rural communities. Journal of Development Studies,33(6), 764–796.
Rapoport, H., & Docquier, F. (2006). The economics of migrants’
transfers, on the economics of giving, reciprocity and altruism. In
S. Kolm & J. M. Ythier (Eds.), Handbook of economics of giving,altruism and reciprocity 2 (pp. 1135–1198). Amsterdam: Elsevier.
Sachs, J. D., & Warner, A. M. (2001). Natural resources and
economic development: The curse of natural resources. Euro-pean Economic Review, 45, 827–838.
J Fam Econ Iss (2011) 32:473–492 491
123
Secondi, G. (1997). Private monetary transfers in rural China:
Are families altruistic? Journal of Development Studies, 33(4),
487–511.
Sheng, X., & Killian, T. S. (2009). Over time dynamics of monetary
intergenerational exchanges. Journal of Family and EconomicIssues, 30(3), 268–281.
Takane, T. (2007). Customary land tenure, inheritance rules, and
smallholder farmers in Malawi. Institute of Developing Econ-
omies Discussion Paper No.104.
Udry, C. (1990). Credit markets in northern Nigeria: Credit as
insurance in a rural economy. The World Bank EconomicReview, 4(3), 251–269.
Van Dalen, H. P, Groenewold, G., & Fokkema, T. (2005). Transfers
and their effect on emigration intentions in Egypt, Morocco and
Turkey. Tinbergen Institute Discussion paper, TI 2005-030/1.
VanWey, L. K. (2004). Altruistic and contractual transfers between
male and female migrants and households in rural Thailand.
Demography, 41(4), 739–756.
Weinreb, A. A. (2001). Substitution and substitutability: The effects
of kin availability on intergenerational transfers in Malawi.
In A. H. Gauthier, C. Chu, & S. Tuljapurkar (Eds.), The
distribution of private and public resources across generations.
Oxford: Oxford University Press/IUSSP.
Weinreb, A. A. (2002). Lateral and vertical intergenerational
exchange in rural Malawi. Journal of Cross-Cultural Gerontol-ogy, 1(38), 101–137.
White, H. (1980). Heteroscedasticity consistent covariance matrix and
a direct test for heterscedasticity. Econometrica, 48, 817–838.
Author Biography
Simon Davies is currently working as an advisor on macroeconomic
policy in the Ministry of Finance and Development Planning in
Lesotho. His research looks at household, individual and firm
behavior focusing on insurance, psychological choices, happiness,
intra-household dynamics, sexual attitudes related to HIV/AIDS and
investment decisions. He obtained his PhD from the University of
Bath in the UK in 2009 and is due to begin a position with the World
Bank in September 2010.
492 J Fam Econ Iss (2011) 32:473–492
123