A causal test of the demonstration effect theory

A causal test of the demonstration effect theory

Economics Letters 103 (2009) 52–54 Contents lists available at ScienceDirect Economics Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r...

177KB Sizes 8 Downloads 137 Views

Economics Letters 103 (2009) 52–54

Contents lists available at ScienceDirect

Economics Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e c o l e t

A causal test of the demonstration effect theory☆ Andreea Mitrut a,b, François-Charles Wolff c,d,e,⁎ a

Department of Economics, University of Gothenburg, Box 640, SE-405 30, Gothenburg, Sweden ASE, Bucharest, Romania c LEMNA, Université de Nantes, BP 52231 Chemin de la Censive du Tertre, 44322 Nantes Cedex 3, France d CNAV, Paris, France e INED, Paris, France b

a r t i c l e

i n f o

Article history: Received 19 March 2008 Received in revised form 13 January 2009 Accepted 23 January 2009 Available online 3 February 2009

a b s t r a c t We propose a causal test of the demonstration effect theory using gender differences in the provision of upstream transfers. Using Romanian data, we find that young daughters affect their parents' contact with the grandparents differently than young sons do. © 2009 Elsevier B.V. All rights reserved.

Keywords: Demonstration effect Upstream transfer Romania JEL classification: D10 J14

In memory of my brother, René Wolff, who passed away on November 7, 2008.

1. Introduction Since the seminal work of Cox (1987), economists have put a lot of effort into understanding the motives behind transfer decisions within the family. It has been suggested that donors could behave in an altruistic way, meaning that they simply care about the well-being of the recipients. On the other hand, private transfers may also be motivated by some kind of exchange if donors expect services from their children when giving them money or expect to be reimbursed later. These two models obviously lead to different predictions, and there is currently no clear empirical consensus on the true motive (see Laferrère and Wolff, 2006).

☆ We are indebted to an anonymous referee for helpful comments. The usual disclaimer applies. ⁎ Corresponding author. LEMNA, Université de Nantes, BP 52231 Chemin de la Censive du Tertre, 44322 Nantes Cedex 3, France. Tel.: +33 2 40141779; fax: +33 2 40141700. E-mail addresses: [email protected] (A. Mitrut), [email protected] (F.-C. Wolff). URL: http://www.sc-eco.univ-nantes.fr/~fcwolff (F.-C. Wolff). 0165-1765/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2009.01.029

Another motive, originally presented in Cox and Stark (1996), is the so-called demonstration effect which involves three generations. Let us consider a family with one grandparent, one parent and one child, where the parent will provide time and attention to the grandparent in order to elicit the same support in the future from the child. Interestingly, this model is not based on a direct exchange between a donor and a recipient; instead it involves the presence of a third person and is rooted in indirect reciprocity. Consequently, parents are expected to have more contact with their elders when they have children. Also, they should devote more time to the elders when their young children are around (as early experiences strongly shape behaviors during adulthood) and they should rely on visible transfers. A few empirical papers have shown that some predictions of the demonstration effect theory were supported by the data. Cox and Stark (1996, 2005) show that U.S. parents have more contact with their parents in the presence of their children. Wolff (2001) finds a similar result for France, although the increased attention stems mainly from childcare services provided by grandparents. Using Spanish data, Giménez et al. (2007) also show that the presence of grandchildren increases the time devoted to elder care activities. On the other hand, Yamada (2006) finds no support for the demonstration effect in Japan. These correlations between contact with parents and number of grandchildren do not mean that children have a causal impact on upstream transfers. In this paper, we provide a new test of the demonstration effect theory. As shown in Jellal and Wolff (2000), gender-specific returns exist when investing in the demonstration

A. Mitrut, F.-C. Wolff / Economics Letters 103 (2009) 52–54

53

effect, with women having longer life expectancy. Thus, daughters should affect transfer behavior differently than sons. We rely on the exogeneity of a child's gender to investigate whether sons and daughters have an influence on how often parents visit their own elders, and provide an illustration using 2005 Romanian data. The remainder of this note is organized as follows. Section 2 emphasizes the gender differences when investing in the demonstration effect and presents our empirical strategy. We then describe the Romanian data and discuss our econometric results in Section 3. Section 4 concludes. 2. Theoretical background and empirical strategy According to the demonstration effect theory, parents will help their own elders because they expect to receive services and attention from their own children in return, after they have grown up. In the theoretical framework of Cox and Stark (1996), there is an exogenous probability for children to reproduce observed parental actions, although the children may also adopt a different attitude. Jellal and Wolff (2000) introduce uncertainty in the life expectancy of the parents in this model, and show that parents who live longer receive more support from their children. From an empirical perspective, the demonstration theory has two main implications. First, since women generally live longer than men, women have an incentive to provide more attention to their elders. Second, since women are, on average, expected to do more for their parents, parents should invest more in the demonstration effect when having daughters than when having sons. With daughters, the indirect exchange is more secure, while the delayed repayment (through transfers from the child to the parent after the former has grown up) is more uncertain with sons. Many empirical studies have pointed to gender-specific attitudes when caring for parents (Jellal and Wolff, 2002). So, a simple implication of the demonstration effect theory is that, relative to having boys, having girls should have a greater influence on the decision of a parent to care for his/her elders. A difficulty here is that fertility decisions may be made simultaneously with the decision to help grandparents. Following the original approach of Angrist and Evans (1998), we choose to rely on the gender composition of the offspring to test whether children significantly affect upstream transfer decisions made by their parents. As the gender of a child is random, and given gender-specific returns in the indirect exchange, a higher coefficient for girls than for boys in the upstream transfer equation is evidence of a causal effect and is consistent with parents relying on the demonstration effect theory. 3. Empirical analysis 3.1. The Romanian survey We propose an empirical application using Romanian data. While many studies have focused on family transfers in either developing or developed countries, less is known about transition countries (Cox et al., 1997; Mitrut and Nordblom, 2008). Romania offers an interesting setting. Formal transfers are very limited, while private transfers are sizeable and quite common.1 Since people often rely on family transfers (see Amelina et al., 2004), there is scope for understanding whether parents have incentives to rely on the demonstration effect in order to secure transfers from their children once they are old.

1 In 2001, almost three out of ten Romanians were poor, while one out of ten was extremely poor. Moreover, the public pension system (PAYG) faced a chronic deficit (about 1% of the GDP), and the average monthly pension was about 1.4 million lei (roughly 40 USD) (see more on the World Bank: Romania Poverty Assessment, 2003).

Fig. 1. The pattern of visits between parents and grandparents.

Specifically, we use the 2005 Living Conditions Survey, a national household-based survey administered by the Statistics Romania covering 9703 households. It includes information on all members living in the household. Also, for each spouse, there are questions on whether his or her parents living outside the household are alive or not: “Do you have any first-degree relative living outside your household?” If yes, we inquire about the frequency of contact with that parent: “How often do you see each other?” There are five possible responses: (1) “very rare or never,” (2) “a few times/year,” (3) “at least once a month,” (4) “at least once a week,” or (5) “every day.” Following Cox and Stark (1996) and all previous studies, we rely on a measure of attention between the two generations as this is a more visible form of family transfer. Since our dependent variable is categorical, we turn to an ordered Probit model. We control for respondent characteristics like gender, age, living arrangements (single vs. cohabitation), health status, education, ethnicity, working status, income and wealth, living in an urban area, and number of parents. Our key variables, which are related to the children, are defined in the following way. First, we make a distinction between young children (aged 0–10) and older children (aged 11–18). All studies have pointed out that preferences have to be shaped when children are young, meaning that the effect of the children variable should be much more pronounced when they are young. Second, for each age group, we compute the number of boys and the number of girls. A shortcoming of the data is that we have almost no characteristics related to the grandparents. For each household, we select the respondent and the spouse (if any) aged 25–60 with at least one parent/parent-in-law who is alive and not co-residing.2 After deleting missing values, the sample consists of 5030 observations. Fig. 1 indicates a positive correlation between the number of young children (aged 0–10) and the parents' contact with the grandparents. On average, respondents and spouses who have young children visit their own parents much more often, which is consistent with the demonstration effect. It could also be that parents have more contact with their elders because the latter care for the grandchildren (Wolff, 2001). However, note that this explanation should not lead to gender-specific effects, unless grandparents care more for girls than for sons. 3.2. Econometric results Let us now comment on the results of our ordered Probit models, described in Table 1. To assess the relevance of the demonstration

2 We choose to exclude households with co-residing elders as co-residence is in itself a form of family transfer, albeit it is difficult to know whether this is an upward or a downward transfer. In the same way, we choose not to control for distance separating respondents and their parents as this covariate is highly endogenous.

54

A. Mitrut, F.-C. Wolff / Economics Letters 103 (2009) 52–54

Table 1 Ordered Probit estimates of contact between parents and grandparents Variables Male respondent In couple Age Age squared (/100) Poor health Education (ref: Primary-grades 1 to 8) Secondary lyceum/vocational school High school University or higher Ethnicity (ref: Romanian) Hungarian Gypsy Work Income (log) Wealth Living in urban area Number of parents Number of children 0–10 Number of children 11–18 Number of boys 0–10 Number of girls 0–10 Number of boys 11–18 Number of girls 11–18 F-test: number of boys 0–10 = number of girls 0–10 (value; prob) Observations Log likelihood

(1)

(2)

Mean

Coef.

t-value

Coef.

t-value

−0.171⁎⁎⁎ −0.076 −0.038⁎⁎ 0.040⁎ −0.009

(5.44) (1.51) (2.16) (1.86) (0.20)

−0.173⁎⁎⁎ −0.076 −0.039⁎⁎ 0.041⁎ −0.008

(5.48) (1.51) (2.20) (1.89) (0.19)

0.539 0.887 39.92 16.69 0.184

0.008

(0.16)

0.005

(0.10)

0.315

0.195⁎⁎⁎ 0.204⁎⁎⁎

(3.80) (3.08)

0.193⁎⁎⁎ 0.200⁎⁎⁎

(3.75) (3.02)

0.397 0.127

0.319⁎⁎⁎ 0.072 0.013 −0.044⁎ 0.146⁎⁎ −0.560⁎⁎⁎ −0.139⁎⁎⁎ 0.065⁎⁎⁎ −0.003

(5.40) (0.75) (0.32) (1.93) (2.42) (15.23) (3.81) (2.89) (0.14)

0.317⁎⁎⁎ 0.072 0.012 −0.043⁎ 0.146⁎⁎ −0.561⁎⁎⁎ −0.140⁎⁎⁎

(5.38) (0.75) (0.31) (1.88) (2.43) (15.26) (3.84)

0.071 0.027 0.781 11.092 0.068 0.640 1.747 0.463 0.475 0.233 0.230 0.249 0.226

0.023 (0.73) 0.102⁎⁎⁎ (3.30) 0.022 (0.71) −0.027 (0.84) (3.12; 0.078)

5,031 −6908.4

5,031 −6906.2

5,031

Source: Authors' calculations using ACOVI, 2005. Estimates from Ordered Probit models, with absolute values of t-statistics in parentheses. Significance levels are 1% (⁎⁎⁎), 5% (⁎⁎), and 10% (⁎) respectively. Each regression also includes a set of threshold values (not reported).

effect, we initially include in Column 1 the number of children between 0 and 10, and between 11 and 18, respectively. Consistent with previous studies, we find a positive relationship between contact and the number of young children, significant at the 1% level, while the number of older children turns out to be insignificant. This pattern is in favor of the demonstration effect theory since it is more valuable to invest in helping the elderly in the presence of young children, but there is no evidence of a causal relationship at this stage. Next, we account for the exogenous child's gender and introduce in the regression both the number of boys and the number of girls, for each age group. As shown in Column 2, we find a positive effect for the number of girls between 0 and 10, significant at the 1% level, while the number of boys plays no role in the regression.3 The key issue here is whether the two gender coefficients are statistically different. Using a linear Wald test, we find a value of 3.12 for the corresponding statistic

3 Concerning the older age group, the coefficients associated with the number of boys and girls remain somewhat low and not significant at any conventional level.

(with one degree a freedom) and a probability of 7.8%. Our results thus show that girls increase contact between respondents and their parents more than boys. We interpret these different effects of boys and girls as a causal influence of children on attention given to grandparents, which is consistent with the demonstration effect theory. Additionally, our regression reveals some other interesting findings. The number of contacts is less frequent when the respondent is a man, living in couple is not significant and respondent' age yields a U-shaped profile. Hungarians provide more contacts with their parents compared to the Romanian respondents, while the Gypsy dummy does not turn out to be significant. Attention increases with the education of the middle generation and with household's (permanent) wealth, proxied here by ownership of a secondary home, a holiday house or stocks, while it is negatively related to household's income. Also, living in an urban area strongly increases the provision of parental attention. Finally, we observe a negative effect of the number of grandparents. 4. Conclusion Several empirical studies have interpreted the positive effect of number of children on contact with elderly as evidence of the demonstration effect. Using exogeneity of the child's gender, we have proposed here a new causal test of that transfer motive based on the different effects of boys and girls on upstream attention. In Romania, we find evidence in favor of the demonstration effect as it is young girls and not young boys that influence parental attention. References Amelina, M., Chiribuca, D., Knack, S., 2004. Mapped in or mapped out? The Romanian poor in inter-household and community networks. The World Bank, Washington. Angrist, J.D., Evans, W., 1998. Children and their parents' labor supply: evidence from exogenous variations in family size. American Economic Review 88, 450–477. Cox, D., 1987. Motives for private income transfers. Journal of Political Economy 95, 508–546. Cox, D., Stark, O., 1996. Intergenerational transfers and the demonstration effect. Mimeo, Department of Economics, Boston College. Cox, D., Stark, O., 2005. On the demand for grandchildren: tied transfers and the demonstration effect. Journal of Public Economics 89, 665–1697. Cox, D., Jimenez, E., Okrasa, W., 1997. Family safety nets and economic transition: a study of worker households in Poland. Review of Income and Wealth 43, 191–209. Giménez, J.I., Marcén, M., Molina, J.A., 2007. How does the presence of children affect dependent care? A psycho-economic approach. IZA Discussion Paper, n° 2726. Jellal, M., Wolff, F.C., 2000. Shaping intergenerational relationships: the demonstration effect. Economics Letters 68, 255–261. Jellal, M., Wolff, F.C., 2002. Cultural evolutionary altruism: theory and evidence. European Journal of Political Economy 18, 241–262. Laferrère, A., Wolff, F.C., 2006. Microeconomic models of family transfers. In: Kolm, S.C., Mercier Ythier, J. (Eds.), Handbook on the Economics of Giving, Reciprocity and Altruism. Elsevier, North-Holland. Mitrut, A., Nordblom, K., 2008. Motives for Private Gift Transfers: Theory and evidence from Romania. Mimeo, Department of Economics, Gothenburg University. Wolff, F.C., 2001. Private intergenerational contact in France and the demonstration effect. Applied Economics 33, 143–153. Yamada, K., 2006. Intra-family transfers in Japan: intergenerational co-residence, distance and contact. Applied Economics 38, 1839–1861.