The Financial Burden of Having Children and Fertility Differentials Across Development and Life Stages: Evidence from Satisfaction Data

Abstract

Comparing the financial burden of having children across countries accompanies various types of measurement issues. The present study employs financial satisfaction to overcome the measurement issues and examines how the financial burden of having children differs across development stages. The challenge in this approach lies in detecting the impact of having children on financial satisfaction. To address this challenge, we focus our attention on the peculiar movement of satisfaction in the financial domain of life, which is measured by standardizing financial satisfaction by overall life satisfaction, and perform regression analyses using World and European Integrated Values Survey. The results show that the negative impact of having an additional child on satisfaction becomes particularly greater in the financial domain as income increases and total fertility rate (TFR) decreases. The results also indicate that having children offers a sense of financial security to the elderly in high TFR countries while this is not the case in lower TFR countries. These results support the general idea that the heavier financial burden of having children is a major cause of fertility decline and provide policy implications to find a way out of extremely low fertility.

Appendix

1.1 Descriptive Statistics, List of Countries, and Country-specific Means

See Tables 5 and 6.

Table 5 Descriptive statistics

Table 6 List of countries and country-specific means

1.2 Validity of Regression Results

1.2.1 Endogeneity Test

We perform the over-identification test (Hansen 1982) by adding an instrument, “importance in life: friends” for high and middle TFR countries, and “the importance in life: politics” for low and very low TFR countries. The results are presented in Table 7. As Hansen’s J-statistic demonstrates, all instruments can be treated as exogenous.

Table 7 Regression results (dependent variable: SFS. Categorized by TFR)

1.2.2 OLS versus IV

To assess the severity of the endogeneity bias, we estimate the same model with OLS. The results, presented in Table 8, show that the difference in the coefficients between IV and OLS estimations is significant for high TFR countries, suggesting that the endogeneity bias is more severe for high TFR countries. Performing the Wu–Hausman test (Hausman 1978), we obtain the results supporting the use of IV at the 1 % level for high TFR countries, and at the 15 % level for low and very low TFR countries. Thus, for comparing the results across various TFR levels, the IV method provides more consistent results.

Table 8 Regression results (dependent variable: SFS. categorized by TFR)

1.2.3 Robustness

To further assess the robustness of the estimation results, we regress the logarithm of SFS, which is technically a more accurate measure than SFS itself. We used SFS in the main analysis because the interpretation of estimated coefficients is more intuitive and because it can sufficiently reduce the endogeneity bias.

To see the logic behind the use of the logarithm of SFS, assume that financial satisfaction, FS, takes a log-linear form such that

$$\ln \left( {FS} \right) = a_{FS} CHILD + {\mathbf{xb}}_{{{\mathbf{FS}}}} + \theta_{FS} + \varepsilon_{FS}$$

where $\theta_{FS}$ and $\varepsilon_{FS}$ are respectively the unobserved individual characteristics and the usual error term and where $\alpha_{FS}$ and ${\mathbf{b}}_{{{\mathbf{FS}}}}$ are coefficients. In reality, however, $\theta_{FS}$ is unobservable and the error term in the regression equation contains not only $\varepsilon_{FS}$ but also $\theta_{FS}$. Thus, as discussed in Sect. 2, CHILD and the error term in the regression equation correlate with each other, and the estimated coefficient of CHILD becomes biased. Using IDEAL as the instrument does not solve the issue because IDEAL is also subjective data and expected to correlate with $\theta_{FS}$.

Assume also that life satisfaction, LS, takes the same functional form. Then, by taking the difference, the logarithm of SFS can be written as

$$\ln \left( {SFS} \right) = \ln \left( {FS} \right) - \ln \left( {LS} \right) = \left( {a_{FS} - a_{LS} } \right)CHILD + {\mathbf{x}}\left( {{\mathbf{b}}_{{{\mathbf{FS}}}} - {\mathbf{b}}_{{{\mathbf{LS}}}} } \right) + \left( {\theta_{FS} - \theta_{LS} } \right) + \varepsilon$$

where $\varepsilon = \varepsilon_{FS} - \varepsilon_{LS}$. Thus, by regressing ln(SFS), we can take the difference between $\theta_{FS}$ and $\theta_{LS}$, both of which share the same individual characteristics, and substantially reduce the endogeneity bias. In addition, this method allows us to use IDEAL as the instrument to further control the endogeneity bias.¹⁸

Table 9 presents the regression results for the financial burden over the life course. The estimated results share the same characteristics as the ones presented in Table 4. Generally speaking, the results are, as expected, more robust than regressing SFS itself. In high TFR countries, the coefficient of CHILD increases with parent’s age. In middle TFR countries, it again turns from negative to positive at the late middle age whereas it is now insignificant even at the 10 % level at any age. In low TFR countries, while the coefficient becomes significant at a young age, it dips again at the late middle age with a higher level of significance. These results point to the robustness of the results of the present study.

Table 9 Regression results (dependent variable: ln(SFS). categorized by TFR and Age)

1.3 Data Sources

EVS (2011). “European Values Study 1981–2008, Longitudinal Data File.” GESIS Data Archive, Cologne, Germany, ZA4804 Data File Version 2.0.0 (2011-12-30) DOI:10.4232/1.11005.

WVS (2009). “World Value Survey 1981–2008 official aggregate v.20090902, 2009.” World Values Survey Association (www.worldvaluessurvey.org). Aggregate File Producer: ASEP/JDS Data Archive, Madrid, Spain.

Feenstra, Robert C., Robert Inklaar, and Marcel P. Timmer (2013). “The Next Generation of the Penn World Table.” available for download at http://www.ggdc.net/pwt.

UN (2013). “World Population Prospects, The 2012 Revision.” available for download at http://esa.un.org/unpd/wpp/index.htm.