Causal effects of family income on educational investment and child outcomes: Evidence from a policy reform in Japan

Highlights

• Causal effects of family income on educational investments and child outcomes examined.

• Child allowance policy reforms provide plausibly exogenous income variations.

• Positive income effects on educational spending even in the short run.

• No causal effects of family income on cognitive outcomes in the short run.

• Interpretation of the results points to the need for analysis of long-term impacts.

Abstract

The Child Allowance Policy (CAP) in Japan, a nationwide cash transfer program for families with children, was designed to increase household expenditures toward children. Using unforeseen changes in the CAP that occurred due to the electoral results as a source of exogenous variation in income in the early 2010s, this paper examines the causal impact of family income on households’ private educational expenditures and child outcomes in the short-run, based on a longitudinal parent-child survey. The ordinary least squares (OLS) and first-differenced (FD) results show that family income is in most cases positively correlated with child's cognitive outcomes, and, to a lesser extent, with families’ educational expenditure on their children. Based on the FD instrumental variable (FD-IV) estimation, using unexpected changes in CAP payments as an instrument, we find positive income effects on educational expenditure in the short-run. However, we did not find statistically significant impacts on children's cognitive outcomes.

Introduction

Many countries have relied on cash transfer programs to improve the well-being of the children of low-income families and to circumvent the cycle of poverty trap across generations. The cash transfer program is supported by the belief that part of an additional income will be used as human capital investment in the health or education of children, which is difficult under the credit constraint created by low income.¹ However, because most previous research was based on programs of developing countries, it is not clear how much additional income would be invested in children and would effectively enhance their human capital in developed countries.

The Child Allowance Policy (CAP) in Japan, which is a nationwide cash transfer program to families with children, has also been designed for the purpose of increasing expenditure for these children. In this paper, we use CAP in order to identify the causal effect of income on a child's education expenditure and children's outcomes using the unique data. To our knowledge, this is the first paper that investigated the effect of CAP on children.

Since the Coleman Report (Coleman et al., 1966), much research effort has been devoted to determining the factors (both at school and among families) that influence the outcomes of children the most. One variable that has been the major focus of economic research is household income. Virtually all previous research has found a positive association between household income and children's outcomes (Duncan and Brooks-Gunn, 1997; Mayer, 1997). The simplest human capital theory, however, predicts that under the complete market assumption, the amount of educational investment does not directly depend on the household's income level (Becker, 1962). Many researchers have questioned the complete market assumption and have examined whether households with low incomes face borrowing constraints and thereby have difficulty investing in their children's human capital. These studies have typically focused on children's college enrollment (Carneiro and Heckman, 2002; Keane, 2002; Cameron and Taber, 2004; Belley and Lochner, 2007).

The major challenge in attempting to identify the causal impact of family income on educational investment and child outcomes is the endogeneity of income. Family income is generally correlated with parental educational achievement, which is likely to be correlated with parents’ and children's unobserved abilities (Duncan and Brooks-Gunn, 1997; Mayer, 1997). Furthermore, some important factors contributing to child outcomes, such as home environments, parenting practices, and cultural backgrounds, are not always observable to researchers. Failing to control for these unobserved factors, which are correlated with family income as well as educational investment and children's outcomes, results in a bias in the estimation of the effect of family income.

In a vast literature, researchers have sought to disentangle the effects of unobservable factors and family income using various methods. Except for a few experimental studies,² most researchers have used panel data of the same families and children that include information on family backgrounds, such as family income, and children's outcomes. Early studies used child fixed effects to identify the effect of transitory income on child outcomes (Duncan et al., 1998; Blau, 1999). However, family income can be endogenous even after controlling for child fixed effects. For example, if a child has a developmental problem, parents might increase the time spend on their child and reduce their working hours. Changes in parental time allocation would then coincide with a reduction in family income, causing reverse causality issues.

Currently, a more prominent approach is to find an exogenous variation of family income due to a policy reform or economic event that is uncorrelated with unobservable factors. This approach has generated conflicting results. Among others, Løken (2010) used the economic boom of the 1970s and 1980s—caused by the discovery of oil fields in Norway—as an instrumental variable (IV) for an increase in income and found no effect on children's outcomes. Akee et al. (2010) used the opening of a casino in North Carolina that benefitted a subpopulation in the area (Native American tribes) and found that it raised children's educational achievements and lowered their probability of committing crimes. Changes in tax and welfare policies have also been used as exogenous sources of variation in income. For example, Dahl and Lochner (2012) used changes in the Earned Income Tax Credit (EITC) in the United States and showed that increased earnings improved the test scores of children in families subject to this policy change.

The main contribution of this paper is the estimation of the causal effects of family income on the households’ private educational expenditure and child outcomes (academic test scores).³ We used changes in the Child Allowance Policy (CAP) in Japan that occurred in the early 2010s as an exogenous source of income variation. The CAP is unique because it is a nationwide program that is not limited to low-income households or working households, allowing us to examine the income effects for entire households with children from a broader range of income distribution. This is in contrast to previous studies that investigated the impact of transfer policies, such as the EITC in the US, with a primary focus on low-income households (Gregg et al., 2006; Dahl and Lochner, 2012). Addressing the effects of income changes on each child's educational expenditures and the child's outcomes is another unique feature of this study as few previous studies have estimated both pathways on a common data set.⁴ Although we do not attempt to estimate the causal effect of educational expenditure on children's outcomes, obtaining the causal effects of income on both expenditure and outcomes provides valuable information regarding the channel through which family income may influence children, facilitating the interpretation of the results.

Another contribution of this paper is the policy evaluation of the recent changes in the CAP from the perspective of child outcomes. Several previous studies have examined the effect of the CAP on consumption patterns (Kobayashi, 2011; Unayama, 2011), the mental health of parents (Takaku, 2015), the labor supply (Bessho, 2018), and household wealth accumulation (Stephens and Unayama, 2015). To our knowledge, however, no previous research has examined the effect of the CAP on households’ educational expenditures and direct measures of children's educational outcomes, which are the primary targets of the policy. More broadly, this paper presents the first empirical evidence on the effect of an unconditional cash transfer program on households’ educational expenditures and child outcomes in the Japanese context.⁵

Our identification strategy relies on the largely exogenous, discontinuous changes in the CAP in Japan that took place between 2010 and 2012. Before 2010, a child aged 12 or younger was eligible for the allowance. A monthly allowance was determined based on children's age and birth order. In 2010, the government substantially expanded the amount and scope of the allowance. A child aged 15 or younger became eligible for the allowance, which was 13,000 JPY (approximately, $US120) per month, independent of the child's age. The current policy, from 2012, provides a monthly allowance again based on children's age and birth order.

The policy changes described above provide a nearly ideal situation for identifying the income effects on children's outcomes for several reasons. First, policy changes were most likely unanticipated by families since they were a result of a regime change (from the Liberal Democratic Party (LDP) of Japan to the Democratic Party of Japan (DPJ)) after the national lower-house election during one of the most volatile and unstable political periods after World War II in Japan. Second, a monthly allowance payment is solely determined based on the number and ages of the existing children in the family, which cannot be controlled by individual households. Under other redistribution policies, such as the EITC in the United States, families can endogenously change their behavior to manipulate the benefits or subsidies they receive from the policy. Altogether, we believe that changes in the CAP can provide an exogenous source of variation in income.

The ordinary least squares (OLS) results show that, in most cases, family income is positively correlated with families’ educational expenditure and children's cognitive outcomes. By controlling for child fixed effects, the first-differenced (FD) results show that the income coefficient is positive and statistically significant only for mathematics and combined test scores. Based on the FD instrumental variable (FD-IV) estimation, using unexpected changes in CAP payments as an instrument, we found significantly positive income effects on educational expenditure even in the short-run but did not find statistically significant impacts on children's cognitive outcomes.⁶

The remainder of the paper is organized as follows. Section 2 summarizes the institutional background of the CAP in Japan. Section 3 provides a brief description of the dataset used in our empirical analysis. Section 4 sets out our identification strategy based on exogenous variations in family income due to policy changes and provides the empirical specification. Section 5 presents our empirical results, and discussions and concluding remarks follow in Section 6.