Fertility decisions and the norm of intergenerational support to aging parents

1 INTRODUCTION

This paper examines the impact of the social norm of intergenerational support on fertility decisions among parents in developing economies. The traditional expectation of receiving support from adult children in old age has historically been a significant factor in the decision to have children, particularly in East Asian and developing countries. However, recent research suggests a weakening of this norm, with important implications for demographic and economic outcomes.1

Unlike developed economies with well-established pension and social security systems, the financial incentive for having children to ensure support in old age is often more salient in developing countries where social security is underdeveloped, and many individuals have low-paying and insecure jobs, making it challenging to accumulate sufficient assets for a comfortable old age. For example, İmrohoroğlu and Zhao (2018a) mentioned that family support for the elderly is an important factor to consider when studying the social security reform in China. Pei and Cong (2020) suggested that family support for the elderly is still important in rural China although it is weaker in urban areas. Cai et al. (2006) found private transfers from children as an important insurance for the elderly when the pension coverage is low.

To explore the relationship between the social norm of intergenerational support and fertility decisions, we construct a life cycle model that endogenizes fertility choices and incorporates the expectation of intergenerational transfers from children based on the filial responsibility norm. In the model, individuals would like to have children not only because children bring happiness (utility), but also because they anticipate a financial return from having children. We use Indonesia as an illustrative example to conduct our quantitative exploration. We also employ household survey data from Indonesia to estimate earnings profiles and uncertainties and ensure that our model accurately captures the financial considerations of individuals throughout their life cycle. Specifically, we use data from the Indonesia Family Life Survey (IFLS) conducted between 2000 and 2014, which is a comprehensive socioeconomic survey that provides detailed information on a wide range of economic outcomes and behaviors of Indonesian households.

To gauge the strength of the filial responsibility norm in Indonesia, we estimate the transfers from adult children to their elderly parents based on the IFLS data. Our findings suggest that although over 70% of elderly parents (age 60+) in Indonesia receive transfers from their adult children, the average net transfer is relatively low, accounting for only 3% of the average earnings of young adults. This result points to a relatively weak norm of old-age support, which seems consistent with the declining fertility rate in Indonesia over time. In fact, the total fertility rate (TFR) in Indonesia has decreased from above 5 in the 1960s to below 2.5 in the 2000s (see Figure 1).2

In addition to directly estimating the above mentioned model parameters from data, we estimated the other parameters by minimizing the distance between data and model moments, such as the average fertility rate and fertility differentials across income groups.

Using our calibrated benchmark model for the Indonesian economy in the 2000s, we conducted counterfactual experiments to explore the impact of the filial responsibility norm on fertility decisions in response to financial incentives for parents. Our findings suggest that a stronger norm of intergenerational support to aging parents, at a level of 10% of average earnings, would result in an increase in the TFR by 0.6 from the current level. Specifically, if this norm was in place in the 1960s, it could have explained up to about 25% of the fertility decline observed in Indonesia from the 1960s to the 2000s. In addition, we also find that if an economic slowdown is expected, the fertility would be further reduced. Moreover, the impact of the social norm on fertility is significant across wealth categories, especially when parents expect their children's income to be significantly higher than their own. Low-income households are more affected by the norm in case of zero income growth due to higher returns from children relative to their own savings. However, in economies with persistent growth (e.g., Indonesia), the norm provides financial incentives even for wealthy parents.

The literature analyzing fertility choice and its impact on demographic change began with Becker (1960), which introduced the concept of the quantity-quality tradeoff concerning children. Subsequent research, including studies by Barro and Becker (1989), Becker et al. (1990), Galor and Weil (1996), Zhang (1997), Doepke (2004), Doepke and Zilibotti (2005), and Liao (2011), aimed to associate fertility and demographic shifts with economic growth based on this quantity-quality tradeoff mechanism. Furthering the scope of these studies, Jones et al. (2008) delved into alternative theories such as the opportunity cost of time and the quantity-quality tradeoff, discussing their ability to explain the relationship between fertility and income. Complementing these works, this paper emphasizes the exploration of the role in family transfers from children under the lens of the filial responsibility norm.

Moreover, studies examining family transfers in developed countries tend to focus on transfers from parents to children. Altonji et al. (1992, 1997), for instance, employed the Panel Study of Income Dynamics (PSID) to test whether the observed family transfers from parents to children are consistent with the standard altruism model. Their findings reject the hypothesis that the transfers are motivated by altruism. Greenwood et al. (2003), also developed a model of parents' fertility decisions and human capital investments in their children to evaluate the child tax credit policy.

However, as mentioned earlier, the financial incentive for having children to ensure support in old age is often more valued in developing countries. In contrast to the aforementioned studies, which focus on caring/altruism from parents toward children, some macroeconomic studies have considered old-age support from children as the incentive for parental fertility decisions. Boldrin and Jones (2002), for example, developed a three-period overlapping-generations model with endogenous decisions on fertility and support to old-age parents to explain the fertility transition in historical data. Zhao (2011) extended the framework in Boldrin and Jones (2002) to study the effects of social security on fertility and earnings distribution through the channel of differential fertility change across income groups, illustrating that the fertility channel could be quantitatively important in accounting for the distributional effect of social security. Boldrin et al. (2015) further compared two types of models with endogenous fertility: the Barro and Becker (1989) type model, where rearing children is due to parental altruism, and the Boldrin and Jones (2002) type model, where rearing children is due to children's filial support in old age. They found that the latter can better account for the negative relationship between old-age pension and fertility observed from Europe/US data.

In addition to the endogenous fertility literature, İmrohoroğlu and Zhao (2018a) also showed that it is important to consider family support from children to the elderly when studying social security issues in developing economies. İmrohoroğlu and Zhao (2018b) studied the saving rate dynamics in China and found that transfers from children to parents provide additional insurance value against parents' old-age risks, suggesting that the saving rate would have been significantly lower if the one-child policy had not been implemented.

In this study, we consider both incentives shaping parental fertility decisions: the altruistic valuation of children and the expectation of support in old age. Drawing upon the life-cycle framework developed by Daruich and Kozlowski (2020), originally crafted to explore transfers from parents to children in support of educational attainment within developed countries, this study focuses on the filial support for elderly parents in developing economies and uses the model to investigate the fertility dynamics given the trend of changes in intergenerational transfers.

One limitation of our paper is that we assume the norm is exogenously given, without specifying any evolution of the norm in the model. In future research, it would be interesting to explore the dynamics of the norm and its interactions with other institutional changes, such as social security development. In addition, the model developed in the paper is a partial equilibrium life-cycle model because the earnings age profile, earnings shocks and the interest rate are all estimated from the data. The general equilibrium effect is not captured in this model. Overall, our findings highlight the importance of considering the role of intergenerational transfers and social norms in shaping fertility decisions and demographic outcomes in developing economies.

The rest of the paper is organized as follows: Section 2 presents the model, Section 3 details the data sources, estimation techniques, and parameter selection, Section 4 provides a quantitative analysis of the role of the norm of old-age support from children, and Section 5 concludes.

2 MODEL

The model economy is populated with ex-ante identical individuals, who become heterogeneous ex-post due to income shocks over the life cycle, differentiated choices of fertility and uncertainty of transfers. As is typical in the literature, we assume asexual reproduction, meaning that we model only a generic parent and abstract from marriage. Similarly, we do not differentiate between genders and abstract from gender differences in our model.

3 DATA, ESTIMATION/CALIBRATION AND THE BENCHMARK MODEL

In this paper, we use Indonesia as an example to characterize a large developing economy. We utilize a household level data set in Indonesia that provides rich information to quantify our benchmark model.

3.1 Data and measurement

To perform a precise quantitative analysis, this study utilizes data from the Indonesia Family Life Survey (IFLS) spanning the years 2000 to 2014, with a 7-year interval between waves. The IFLS is a comprehensive socioeconomic survey that offers extensive data on a wide range of behaviors and outcomes of Indonesian households. This data includes information on demography, employment, labor force participation, education, health, income, expenditure, housing, fixed assets and durable goods, involvement in poverty alleviation programs, general economic conditions, agricultural production, local infrastructure and transportation, and social problems. These surveys enable the construction of panel data that is well-suited for the purpose of our estimation.4

3.1.1 Family transfers

From IFLS data in the period 2014, we focus on adult child at aged 25 and over and construct a sample consisting of pairs of child-parent. The information on transfer flow between children and parents is based on those following questions:

Based on IFLS data, we have constructed a sample consisting of pairs of adult children and their parents, with the adult children aged 25 or over. To gather information on transfers between children and parents, we utilize the following questions in the survey:

During the last year, did you receive any money transfer from your children? About how much were those transfers?

and During the last year, did you send any transfer to your children? About how much were those transfers?

These questions specifically inquire about support from children, which is advantageous compared to other surveys that do not specify the relationship or purpose of the transfers. We collected data on both upstream and downstream transfers for respondents and their non-coresiding children. Figure 2 shows that more than 70% of old parents received positive transfers from children.

3.2 Parameter estimation and calibration

The model parameters are determined through a two-stage process. During the first stage, we estimate/choose the parameters that can be identified outside of the model, such as earnings age profiles, earnings shocks, and the strength of the norm of old-age support from children (measured by net transfer to income ratios), either from data or by adopting values from previous studies. In the second stage, we calibrate the remaining parameters to match targeted moments.

3.2.1 First-stage estimation/parameterization

Earnings

We adopt the following regression model of log earnings ($$y$$) of household $$i$$ with age $$j$$ in period $$t$$:

$$$\mathrm{ln}({y}_{i,j,t})={\beta }_{0}+{\beta }_{1}(j)+{\beta }_{2}{(j)}^{2}+f({X}_{i,j,t})+{z}_{i,j,t},$$$()

where $${X}_{i,j,t}$$ controls for other factors such as time, location, education, and so forth, and $${z}_{i,j,t}$$ is the idiosyncratic shock. We model the unobserved shock $${z}_{i,j,t}$$ as the sum of two independent components

$$${z}_{i,j,t}={\alpha }_{i}+{u}_{i,j,e,t}+{m}_{i,j,t},$$$()

where $${u}_{i,j,t}$$ is a persistent shock assumed to have an AR(1) structure

$$${u}_{i,j,t}=\rho {u}_{i,j-1,t-1}+{v}_{i,j,t}$$$()

$$${v}_{i,j,t}\unicode{x0007E}N(0,{\sigma }_{v})$$$()

and $${m}_{i,j,t}\unicode{x0007E}N(0,{\sigma }_{m})$$ is a transitory shock. The initial draw is $${u}_{i,0,t}\unicode{x0007E}N(0,{\sigma }_{{u}_{0}})$$. The parameters $${\beta }_{0}$$, $${\beta }_{1}$$, $${\beta }_{2}$$, $$\rho ,{\sigma }_{{u}_{0}},{\sigma }_{v}$$, and $${\sigma }_{m}$$ are used in the model and estimated from the IFLS data. Please see the appendix for the details of measuring earnings.

Figure 3 displays the predicted earnings age profile from the estimation, which shows a hump-shaped pattern as observed in the literature. We normalize earnings at age 40 to one. Regarding shocks, our estimation results suggest values of $$\rho =0.99$$, $${\sigma }_{{u}_{0}}=0.52$$, $${\sigma }_{v}=0.14$$, and $${\sigma }_{m}=0.51$$. While the persistence and variances of persistent and fixed shocks are close to those observed in the United States (Storesletten et al., 2004), the variance of the transitory shock in Indonesia is higher. Figure 4 displays the predicted variance of earnings by age, which exhibits an increasing pattern with age.

Strength of the old-age support norm

We calculate the average net transfers to parents from individuals aged 25 and above using data from the IFLS 2014 and use the fraction of average earnings to represent the current strength of the norm of supporting aging parents (i.e., $$b$$ in the model). We find that $$b$$ is relatively low, at $$3 \% $$, which is close to that estimated for China in Curtis et al. (2015).

Other parameters

Survival rates $${\rho }_{j}$$ are identified from the life tables of the World Health Organization. The interest rate $$r$$ is based on the average real interest rate in the 2000s, which was 6%, and thus the annual time discount factor $$\beta $$ is set to 0.94. We follow Daruich and Kozlowski (2020) and set the value of intertemporal elasticity $$\mu $$ at 0.5, which is commonly used in the literature studying fertility choices with OLG models, for example, Greenwood et al. (2003) and Roys and Seshadri (2017). Manuelli and Seshadri (2009) have estimated this parameter and also obtained roughly this value. In addition, we assume parents expect an increase in children's earnings $${y}_{c}=(1+\tilde{g})y$$. Because one period is 30 years in the model, $$1+\tilde{g}={(1+g)}^{30}$$ where $$g$$ is the average annual growth rate. We utilized the average real GDP per capita growth rate in Indonesia from 1980 to 2010, obtained from the World Bank, to approximate $$g$$ as our benchmark scenario. The annual average growth rate ($$g$$) during this period is 3.4%. Table 1 summarizes the parameters in the benchmark model.

Table 1. Parameters of the benchmark model.

	Description	Value	Source/targeted moment
Parameters identified outside of the model
$$y(z,j)$$	1) Earnings age profile	see text	Estimated (IFLS)
	2) Earnings shocks	see text	Estimated (IFLS)
$$b$$	Strength of norm	0.03	Estimated (IFLS)
$${\rho }_{j}$$	Survival rates	see text	Estimated (WHO)
($$r$$, $$\beta $$)	Interest rate/discount factor (annual)	(0.06, 0.94)	World Bank
$$\mu $$	Relative risk aversion	0.5	Daruich and Kozlowski (2020)
$$g$$	Real GDP growth rate (annual)	0.034	1980–2010, World Bank
Parameters calibrated in the model
$$\lambda $$	Utility parameter	0.49	Match average TFR and
($${c}_{1},{c}_{2},{c}_{3}$$)	Child rearing cost	(0.17, 0.68, 0.81)	Fertility rates by wealth level

3.2.2 Second-stage calibration

We calibrate four parameters, including the utility parameter $$\lambda $$ and child-rearing cost parameters, to match the average TFR and fertility by wealth groups as reported in the 2002–2003 Indonesia Demographic and Health Survey (IDHS). We adopt the functional form of child-rearing cost with taking into account parents' income and number of children from Daruich and Kozlowski (2020): $$C(y,n,j)={c}_{1}{(y)}^{{c}_{2}}{n}^{{c}_{3}}$$ for $$j\le {J}_{o}$$. $${c}_{1}$$ is to capture the basic cost of having a child as a fraction of income, while $${c}_{2}$$ captures the curvature of parents' time opportunity cost based on income levels. Additionally, $${c}_{3}$$ accounts for the marginal cost of having more than one child. By minimizing the distance between the model and data moments, we obtain values of $$\lambda =0.49$$, $${c}_{1}=0.17$$, $${c}_{2}=0.68$$, and $${c}_{3}=0.81$$. Our calibrated parameters for the cost function are close to those found in Daruich and Kozlowski (2020), except for $${c}_{1}$$, which reflects lower child-raising costs in developing countries.

4 QUANTITATIVE ANALYSIS—ROLE OF THE FILIAL NORM OF SUPPORTING AGING PARENTS

We perform several counterfactual experiments by varying the strength of the norm ($$b$$) in the benchmark model. A larger value of $$b$$ indicates a stronger norm of support from children to aging parents (as children transfer $$b$$ proportion of their average earnings to their parents). In the benchmark, $$b$$ is relatively small, indicating a weak norm in the 2000s. Since direct data on family transfers in the 1960s, when the fertility rate was high, is unavailable, we consider several scenarios with different values of $$b$$ for the experiments.

Table 3 presents the results with $$b$$ at 5% and 10%, in addition to the benchmark ($$b=3 \% $$), under an annual income growth rate scenario of $$g=3.4 \% $$. Our findings suggest that a stronger norm of old-age support increases fertility. Specifically, the TFR would increase from 2.6 in the benchmark to 3.2 if $$b$$ increases to 10%.

Furthermore, we observe that the impact of the social norm change on fertility is significant across wealth when parents expect their children's income to be significantly higher than their own (i.e., $$g=3.4 \% $$ in the benchmark, implying an average income of children in the next 30 years to be 2.7 times that of parents'). For instance, increasing $$b$$ to 10% leads to a rise in the fertility rate from 3.2 to 4.0 for those in the bottom 40% of the wealth distribution, while those in the top 40% also experience an increase from 1.8 to 2.2.

Given that income growth may vary considerably across countries and over time, we conduct additional experiments with alternative income growth rates $$g$$, in addition to the benchmark setting of $$g=3.4 \% $$. Table 3 presents two additional scenarios: one with a higher growth case ($$g=6 \% $$) and another with a zero growth case ($$g=0 \% $$).

In the case where parents expect a higher income growth (e.g., $$g=6 \% $$), even if the filial norm remains unchanged, the average fertility rate would increase from 2.6 in the benchmark to 2.9 (see the 4th column of Table 3). Conversely, if parents expect the economy to stop growing (e.g., $$g=0 \% $$), the fertility rate would decrease to 2.2 even with the filial norm remaining the same (see the 7th column, the 3% case under $$g=0 \% $$ in Table 3).

One noteworthy finding is that the fertility decisions of poor households are more strongly affected by the norm than those of rich households in the case of zero income growth. This stronger impact on low-income parents can be attributed to the higher returns from children compared to their own savings, as they have lower opportunity costs of rearing children. In contrast, high-income parents face higher opportunity costs associated with having children, and the return on future transfers from children may not necessarily exceed the return on savings. Consequently, the channel of old-age support is more valuable for poorer households.

However, in economies experiencing persistent growth, such as Indonesia, the filial norm with a significant increase in children's income brings nontrivial financial incentives even for wealthy parents.

Our quantitative analysis suggests that a stronger norm of intergenerational support to aging parents, at a level of 10% of average earnings, would increase the TFR by 0.6 from the benchmark level (representing the 2000s). Specifically, if this norm had been in place in the 1960s, it could have explained about 25% of the observed fertility decline in Indonesia from the 1960s to the 2000s. Furthermore, our experiments also show that with higher income growth, the impact of the norm on fertility decisions will be more pronounced. If parents expect an economic slowdown, they would further reduce the number of children.

5 CONCLUSION

This paper examines the impact of the social norm of intergenerational support on fertility decisions in developing economies, using Indonesia as an illustrative example. The study constructs a life cycle model that endogenizes fertility choices and incorporates the expectation of intergenerational transfers from children based on the filial responsibility norm. Using the Indonesia Family Life Survey data, we estimate earnings profiles and uncertainties and ensure that the benchmark model accurately captures the financial considerations of individuals throughout their life cycle. We also estimate transfers from adult children to elderly parents and find that the strength of the norm of old-age support is relatively weak in Indonesia in the 2000s. With the costs and benefits of having children calibrated to match the average fertility in Indonesia in the 2000s, we have a plausible platform for conducting counterfactual experiments. The results show that a stronger norm of intergenerational support to aging parents would result in an increase in the TFR. If a more stringent norm, which required children to transfer 10% of their income to support their aging parents, had been in place in the 1960s, the decline in social norm could have accounted for approximately 25% of the decline in fertility that was observed in Indonesia between the 1960s and the 2000s. Furthermore, our experiments also show that with higher income growth, the impact of the norm on fertility decisions will be more pronounced. It implies that an economic slowdown may also lead to a decline in fertility. The study suggests the importance of considering the role of intergenerational transfers in shaping fertility decisions and demographic outcomes in developing economies.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENT

This study exclusively involves theoretical analysis and utilizes publicly available data. No primary data collection was conducted, and the study does not involve any interaction with human or animal subjects.