CEEAplA WP No . 02 / 2006 Inter-Household Private Transfers and Underlying Motives : Evidence for Bulgaria

The e¤ects of interactions between private transfer behavior and income redistribution policies depend on the motives underlying private income transfers. This paper tests for two di¤erent potential motives: pure altruism versus simple exchange, in the presence of capital market imperfections. Using household survey data for Bulgaria, microeconometric evidence is found that both motives are in e¤ect. We also nd evidence that capital market imperfections are likely to be binding for consumption smoothing, and hence are an important cause of private transfers. The results indicate that social security benets crowd in the incidence of private transfers, but not the amounts transferred.


Introduction
In the presence of operative inter-household private transfers, the e¤ects of income redistribution policies become uncertain and dependent on the motives underlying inter-household private transfers.For instance, Becker (1974Becker ( , 1993) ) shows that if inter-household private transfers are operative and are an outcome of altruistic feelings satisfaction, households can completely neutralize the effects of income redistribution policies, by adjusting the levels of their transfers (see also Altonji, Hayashi and Kotliko¤ (1997)).However, if households are motivated by exchange, that is, if households give because they expect something in return, this result does not hold (Altonji, Hayashi and Kotliko¤ (2000), Bernheim, Shleifer and Summers (1987) and Cox (1987Cox ( , 1990))).Hence, if one is interested in anticipating the outcome of a given income redistribution policy, one must understand the motives behind inter-household private transfers.This paper does just that, looking at the case of Bulgaria.
In particular, this paper tests the empirical relevance of two hypotheses -the altruism hypothesis and the exchange hypothesis -which have been accepted in the literature as the main driving forces behind inter-household private transfers (see Cox et al. (1998)).As its name suggests, under the altruism hypothesis households give to satisfy their altruistic feelings.In turn, under the exchange hypothesis, households give because they expect something in return, namely a future repayment.While it is likely that both motives are at work, it is plausible that they may not work to the same extent.Therefore, it is important to test which motive dominates at an empirical level in order to anticipate the outcome of a given income redistribution policy.This empirical question has important implications for a number of policies.If the altruism hypothesis is the main driving force behind inter-household transfers, then households may neutralize not only income redistribution policies, but also tax and debt policies, as several authors in the macroeconomics and public …nance literature have argued (see, for instance, Barro (1974)).
Inter-household private transfers are also important for reallocating resources.Cox and Jimenez (1990) document for a large sample of countries that more than half of the households engage in private transfers.In addition, and quite interestingly, Cox and Jimenez also show that it is often the case that the amount of transfers received is large in the sense that it constitutes an economically important fraction of the household's overall income.Hence, studying inter-household transfer behavior is important not only for a better understanding of allocation of resources but also to how safety nets work.Therefore, the empirical work in this paper also sheds light on important relationships that structural models of the household must rationalize.
We use microeconometric data to carry out our empirical work, for Bulgaria, collected by the World Bank, which allows us to control for an interesting number of household economic and demographic characteristics.We follow Altonji, Hayashi and Kotliko¤ (2000), Cox, Jimenez and Okrasa (1997), Cox, Jimenez and Eser (1998), Cox, Hansen and Jimenez (2004), among others, and estimate a microeconometric model of the determinants of the incidence of transfers and of the amount of transfers received.Our contribution is, thus, empirical.
The paper is organized as follows.Section 2 presents a simple model that guides the econometric work.Section 3 describes the data.Section 4 documents the empirical work.Finally, Section 5 concludes with policy implications.

Theory on Motives for Inter-Households Private Transfers
This section presents a simple model of the decision to transfer income between households in order to motivate and guide the empirical work.To test which motive -altruism vs. exchange -is the most important driving force underlying inter-household transfers one must look at the relationship between the recipient's pre-transfer income and the transfer amounts received.Under the pure altruism hypothesis this relationship is always negative.The exchange hypothesis, in turn, is not inconsistent with either a positive or a negative relationship between these two variables.Moreover, and still under the exchange hypothesis, and in the presence of capital market imperfections, transfer amounts received should rise with low levels of pre-transfer income and decline for high levels of pre-transfer income.That is, transfer amounts received is non-linear and concave in pre-transfer income, and hence non-monotonic.
Capital market imperfections are likely to be a strong cause of private transfers.If we consider households who wish to smooth their real consumption levels over their life-cycle, then if capital market imperfections bind, they will be unable to achieve their …rst-best real consumption path (Cox (1990)).This fact may prompt households to engage in private transfers with other households.This observation becomes clear if we assume for simplicity that capital markets are "perfectly imperfect".In particular, young households whose actual income is lower than their permanent income cannot borrow against their potentially higher income that they will receive while middle aged.Similarly, middle aged households whose actual income is higher than their permanent income cannot save for their retirement.What actions can these households take to ease the restrictions that they face?To answer to this question, we present below simple models of both the altruism hypothesis and of the exchange hypothesis that help us in setting up the empirical work.Admittedly, we do not fully develop the models as they are developed elsewhere.Our goal here is to provide enough intuition to develop our empirical tests.
Altruism Consider …rst altruistically motivated private transfers.The model presented to illustrate this hypothesis features utility interdependence and is due to Becker (1974).Suppose that parents care about their children, so that when children's income is low enough, as it would be early in the lifecycle, parents transfer income to their children.In addition, children care about their parents'well being, so that when the parents'earning power is low -i.e.retirement years -children transfer income to their parents.Formally, this utility inter dependence setting can be expressed by the following set of equations: where U denotes parental utility, a positive function of parental consumption c p and children's utility V .Since we assume that altruism is mutual, there is an analogous expression for the children's well being: where c k denotes children's consumption (k is a mnemonic for kids).The following budget constraints capture capital market imperfections: where T j denotes transfers received, net of transfers given by person j, and I j denotes person j's pre-transfer income.
Assume that parents and children overlap for 2 periods, 1 and 2. Period 1 is youth period for children and middle age for parents and period 2 is middle age for children and retirement for parents.In terms of pre-transfer income con…guration, we have the following pattern: The main insight of altruistically motivated private transfers is very simple: Private transfers can help overcome capital market imperfections, as parents transfer income to children in the …rst period and children transfer income to parents in the second period.A key prediction of this model is that an increase in pre-transfer income is always associated with a decline in transfers.Children with higher I k;1 require smaller T k to attain the level of consumption that is optimal from the parents'perspective.This results holds for T p in the second period.In terms of derivatives, we have @T k @I k;1 < 0 and @Tp @Ip;2 < 0 regardless of income levels.
Note that transfer behavior has two dimensions.The …rst is to transfer or not (the decision) and the second, contingent on deciding to transfer, is the amount to transfer.An increase in I k;1 reduces the parents'marginal utility of transferring income to the children and thus we expect a negative relationship between the incidence of transfers (likelihood of being a net receiver) and the recipient's pre transfer income under the altruism hypothesis.
Exchange This section presents a simple model of the exchange hypothesis taken from Cox et al. (1998).Suppose that parents and children realize the potential to engage in mutually bene…cial income transfers.Parents transfer to children in the …rst period and are paid back in the second period.Assume Nash bargaining.The parent's and children's lifetime utilities are de…ned as follows: where is the subjective rate of time preference, which for simplicity is assumed to be the same for parents and children.The parental loan is denoted by T and the repayment is denoted by R. Altruism is not dispensed in this particular bargaining framework.But this depiction of altruism di¤ers from the altruism hypothesis above, in which one agent implicitly dominates the bargaining arrangement.The levels of utility that parents and children can obtain on their own -the threat points -are given by: As usual, the solution to the Nash bargaining problem is given by: max The implications of the bargaining solution are easiest to see with a simulation exercise.Consider logarithmic functional forms for equations ( 4)-( 7) and suppose that I k;2 = 150, I p;1 = 150, I p;2 = 20, = 0:25 and = = 0:30.Figure 1 displays the results of varying I k;1 from 1 to 30 on the value of …rst period transfers T .Transfers initially rise with I k;1 , which contradicts the results from the altruism model.When I k;1 increases two e¤ects take place.The …rst e¤ect is that the children's liquidity constraint is eased, which reduces the …rst period transfer.The second e¤ect is that the children's threat point utility rises.This second e¤ect causes an increase in transfers, because the terms on which the children can borrow improve: The implicit interest rate for intergenerational loans, (R T )=T , declines as I k;1 rises.If the second e¤ect dominates the …rst e¤ect, @T @I k;1 is positive.Furthermore, since the second e¤ect is stronger at lower levels of I k;1 , @ 2 T @I 2 k;1 is negative under the exchange hypothesis.
Under the exchange hypothesis, an increase in the recipient's pre-transfer income reduces the chances that intergenerational lending is mutually bene…cial.Thus, the incidence of transfers is inversely related to own pre-transfer income, just as under the altruism hypothesis.However, while the exchange hypothesis implies that an increase in the income of potential recipients should decrease the likelihood of receiving transfers it can increase the amounts transferred.2468 households and 7199 individuals.The interviews took place in May 1995.Households constitute the unit of analysis.Households with missing information for age, education, and gender of the head of the household, and households with no residents were deleted from the sample.The …nal sample has 2427 observations.Income variables are presented and analyzed on a monthly basis.
Almost 30% of the sample engaged in private transfers, or about 700 households.Of these, about 15% received a private transfer, while 13.7% gave private transfers.Only 50 households both donated and received transfers.For the subsample that received a transfer, private transfers averaged 2602 leva, roughly 23% of this same group average pre-transfer income.Social security bene…ts averaged 2194 leva for all sample.From these descriptive statistics, one can see that private transfers may play a crucial role in poverty alleviation, income redistribution and their interaction with public policies is, thus, potentially intense. 1ouseholds were asked to specify the sources of transfers received and destinations of transfers given.The table  Perhaps as expected (Altonji, Hayashi and Kotliko¤ (1996)), the bulk of transfers occurred between parents and children.The main source of transfers was from parents to children (68%).The second most important source of transfers was from children to parents (17%).Transfers among non-relatives occurred only in 2% of the cases.

Empirical Model
In order to learn about the determinants of inter-household private transfers behavior -incidence and volume -we follow the literature (see Cox, Hansen and Jimenez (2004), Cox, Eser and Jimmenez (1998) and Cox, Jimenez and Okrasa (1997), among others) and estimate an ordered probit model and a Heckman selection model (see Greene (2003) for details on both models).With respect to the former, we estimate an ordered probit model to learn more about the incidence of transfers, encompassing not only net-receivers but also net-givers and households who do not engage in private transfers.More formally: where h is a normally distributed disturbance, cut 1 and cut 2 are ancillary parameters estimated by MLE, and X 1h is a vector containing the covariates whose e¤ect on predicted probabilities we are interested in.Table 1 summarizes the main results from the ordered probit model, which we comment in the next section.
We also estimate a regression model of the amount of transfers received.As usual, and since there is scope for a potential selection problem, we estimate a Heckman selection model.The selection equation reads: where h indexes households, t h is the latent variable, T h is the actual amount of transfers received, I h is pre-transfer income and X 2h is a collection of socioeconomic variables, including age, education and other demographic variables and e h is an error term.The estimating equation for transfer amounts received reads: and h and is a random error component.Table 2 summarizes the main results from the Heckman selection model.The exchange model predicts an inverted U-shaped relationship between recipient's income and transfer amounts received that can be tested by a quadratic form in income.The altruism model, in turn, predicts a monotonically decreasing relationship between those variables.Pre-transfer income also enters in quadratic form in the selection equation because, although neither model predicts a de…nite sign for a 2 , they do not imply a linear relationship between pre-transfer income and the incidence of transfers.This way, hence, less structure is imposed.The model is estimated by MLE, with STATA, using as starting values the values obtained from Heckman's 2-step procedure.Identi…cation of the model is guaranteed by the fact that the relevant X 2h is a subset of X 3h (see Cox, Eser and Jimmenez (1998) for more on the identi…cation strategy).

Results
Since both altruism and exchange models are derived under the assumption of capital market imperfections, before proceeding one must investigate if capital market imperfections are likely at place.One way to analyze this issue is to consider the case of perfect capital markets.If capital market imperfections do not matter, the position of the household over her life-cycle should not matter for the probability of receiving a transfer.Only the present value of lifetime wealth would matter.This contradicts the results illustrated in Figure 2, constructed from the ordered probit analysis presented on Table 1.The probability of being a net-receiver for a household with average characteristics in all aspects other than age …rst declines and then increases with household age (proxied by the head's age).Middle aged households (with higher earning power) are the less likely to receive a transfer, where the youngest are the most likely.Capital market imperfections are, hence, very likely to bind.
Table 2 summarizes the results from joint estimation of ( 10) and (11).The reason for joint estimation as an MLE problem is to correct the amounts equation coe¢ cients for a possible selection problem.However, as Table 2 documents, there is no signi…cant selection problem.The point estimate for (the coe¢ cient associated with the Inverse Mill's Ratio) is only 0.08, which is statistically equal to zero, at conventional con…dence levels (asymptotic t stat. is 0.638).
For the structural amounts equation, pre-transfer income has a positive sign and pre-transfer income squared has a negative sign, as predicted by the exchange or bargaining model.However, these coe¢ cients are both statistically insigni…cant, at conventional con…dence levels.For the selection (feeder probit) equation, the signs of these variables are reversed, which is consistent with both models.The relationship between pre-transfer income and transfers is not clear from these statistics.Both models are not rejected and a plausible explanation is that both motives matter to a fairly equally important matter.More light can be shed into this issue when we combine information from both the structural and the selection equations to obtain Figure 3.In Figure 3 the value of expected receipts for a household with average education and demographic characteristics, monotonically decreases with pre-transfer income (which is allowed to change in percentiles, in 60 equal steps, from the 1st to the 99th percentile).This result should be interpreted with caution since the pattern may be in ‡uenced by a declining probability rather than an overall negative relationship between recipient's pre-transfer income and transfers.Nevertheless, it is contradictory with the exchange model and consistent with the altruism model.
An interesting result in its own right concerns the empirical relation between public transfers and inter-household private transfers.There is no evidence of crowding out between private transfers and public transfers, in the form of social security bene…ts.Both the probit equation of the Heckman selection model (Table 2) and the ordered probit (Table 1) indicate that the likelihood of receiving a transfer increases if the household receives social bene…ts.Using the ordered probit analysis, the probability of being a net receiver increases by 4.4 percentage points, when evaluated at sample means, if the household receives social bene…ts. 2emale headed households are also more likely to be net recipients.Using the ordered probit analysis once more, one can quantify this gender e¤ect at 3.8 percentage points (at sample means).Marital status also increases the probability of being a net-receiver by 5.3 percentage points, while having no working people in the household also increases the probability of being a net receiver by 4.1 percentage points (both evaluated at sample means).Households where there are sick or injured people are more likely to be net-receivers.However, this e¤ect is merely 1.7 percentage points (at sample means).Finally, households who live in urban areas are also more likely to be net receivers, by 5.9 percentage points than their rural counterparts (at sample means).

Conclusions
The motives underlying inter-household private transfers are important for a number of important phenomena, including the e¤ects of income redistribution, tax and debt policies, allocation of resources, safety nets and structural models of the households.We provide empirical microeconometric evidence on inter-household private transfers for Bulgaria and use our results to evaluate the empirical relevance of the altruism model and of the exchange model of interhousehold private transfer behavior.We …nd that neither model is strongly rejected by the data and that both motives are likely to be at work to the same extent.Hence, it is not likely that households may neutralize the e¤ects of income redistribution policies.We …nd that inter-household private transfers may play an important role as a safety net, given their incidence and volume.Our results suggest that capital market imperfections bind for consumption smoothing.Household demographic characteristics matter to predict the incidence and volume of inter-household private transfers, and, hence, income redistribution policies should take into account such household demographic characteristics.Finally, and quite interestingly, we …nd that public transfers crowd-in interhousehold private transfers, even after controlling for a plethora of household characteristics.

Table 1 -Ordered Probit Analysis
In selection equation dependent variable is transfer receipt as binary variable (1 if receives transfers; 0 otherwise).(b)In structural or regression equation dependent variable is net transfer amount received.-.-means that the variable was not used as a regressor