The Relationship between Housing Price and Low Fertility Rate --- The Case of Taiwan as An Example 1
3. Research Method and Data Information
3.1 Research Method
The purpose of this study is to explore the impact of fertility factors and the lead-lag relations between TFR and other variables by empirical results. For the former section, we use Johansen cointegration method to analyze TFR and other variables respectively. As for the latter section, granger causality test is employed to examine the lead-lag relations.
3.1.1 Cointegration
The main time series variables of this study are housing price (HPIt) and fertility rate (TFRt). The regression can be written as follows.
TFRt = c0 + a1HPIt+ εt (1) In this equation, c0 and a1 represent the coefficient, TFRt represents total
fertility rate, HPIt represents housing price index, and εt represents the error term.
9
If εt is stationary, housing price and fertility rate are thus cointegrated. Therefore, we should test whether the residual series has a unit root or not.
After Engle and Granger proposed two–step method in 1987, Johansen and Juselius addressed maximum likelihood estimation in 1990. In general, Engle and Granger (1987) can effectively avoid spurious regression, but the results could be influenced under the situations of small amount of samples. The Johansen cointegration approach therefore provides a more flexible and robust way to interpret the equilibrium relationship between variables.
3.1.2 Granger Causality Test
We aim at the relationship between housing price and low fertility rate, especially for the lead-lag relations. Previous studies showed that there is two-way relationship between housing price and fertility rate. This conclusion is confusing since low TFR is expected to cause low housing prices. This study therefore conducts Granger causality test to investigate the causalities both of TFR on housing price and of housing price on TFR. By causality, we mean the existence of lead-lag relations between housing price and TFR, which implies the ability of one variable to predict another. The definition of Granger causality is that there exists two variables, Xt
and Yt, Xt= bijYt−k and Yt= aijXt−k. When past value of Xt can be explained by current and past values of Yt, Xt is said to be Granger-caused by Yt. In other words, Yt helps in the prediction of Xt. The hypothesis is expressed as:
H0：aij = bij = 0, represents there’s no causality.
H1：aij ≠ 0 or bij ≠ 0, represents there exists causality.
3.2 Data Information
3.2.1 Variable Selection
(1) Housing Price
When households make fertility decisions, the opportunity costs of fertility are often taken into account. Those may widely cover from housing price, household
10
income, government subsidy to tuition fee, among which housing price could be the most expensive. Especially in recent years, the housing price has risen sharply, shown in Figure 3. In the same period, the decline of fertility has been more rapid and pervasive than expected. Rising housing price will deter household from fertility behavior. In other words, the escalation of housing price will change household’s budget constraint and may naturally induce a demographic transition.
Therefore, we suggest that the house price and low fertility rate should have
equilibrium relationship in the long-run. Additionally, taking the opportunity cost of fertility behavior into account, housing price is likely to be associated with fertility decisions. Hence, this study anticipates that housing price may be negatively correlated with fertility rate.
(2) Female labor force participant
Higher female education will postpone the timing of female childbearing; and further, it will enhance the female labor force participation rate. This implies that female education and female labor force participation rate will affect each other, so we choose one of them for empirical analysis. We decided to focus on the female labor force participant rate because the research period of female education is limited.
Most of the industrialized countries have witnessed an increase in female labor force participant rate with a simultaneous decrease in fertility. As discussed, the ability to work will increase the opportunity cost of the fertility decision-making, so women with jobs are likely to be the cause of the reducing number of children.
Besides, as seen in Figure 3, the trend of female labor force participant rate is negatively related to fertility rate. This study thus expects that the increase of female labor force participant rate will reduce the fertility rate.
(3) Household income
The increase in household income may also reduce fertility rate by expanding the opportunity cost. Yet it may contrarily increase fertility rate due to the
increasing income, household can thus afford costs of childbearing. Especially, wealthy people will choose to increase fertility in order to inherit the family business.
As shown in Figure 3, there’s no obvious connection between income and fertility.
Consequently, this study considers the increase of income may affect fertility rate at both positive and negative signs.
11
(4) Unemployment rate
High unemployment rate is often accompanied by the overall economic downturn, thus households will choose to work more and postpone or stop childbearing. The unemployment rate is therefore expected to have negative relationship with fertility rate. In Figure 3, it seems that the unemployment rate is negatively related to fertility rate.
(5) Consumer price index growth rate
Lastly, related literatures rarely discuss the relationship between fertility and consumer price index since it could be interfered by other economic elements easily.
Real estate is traditionally treated as a hedging tool against inflation risks, therefore, this study adopts the growth rate of household income to eliminate the interference of consumer price. After adjusting the variable, we decide to explore the impact of consumer price on fertility behavior. As shown in Figure 3, the consumer price index appears to have no connection with fertility rate. Nonetheless, taking
opportunity cost into consideration; we presume that increases of consumer price would lead to declines in fertility rate.
12
Figure 3 The Trend of TFR and other Variables in Taiwan
3.2.2 Data Source
This study employed the data in Taiwan from the first quarter of 1991 to the first quarter of 2011 with 371 samples in total. The descriptions and sources of each variable are shown in Table 1. In fact, we faced a number of data limitations.
Although low fertility rate phenomenon traced decades ago, the housing price data are not transparent in Taiwan now. As a result, we can only conduct the research through the limited data mentioned above.
Table 1 The Variable Descriptions
-0.0500
1991Q1 1992Q4 1994Q3 1996Q2 1998Q1 1999Q4 2001Q3 2003Q2 2005Q1 2006Q4 2008Q3 2010Q2
TFR household income growth rate
0.0000
1991Q1 1992Q3 1994Q1 1995Q3 1997Q1 1998Q3 2000Q1 2001Q3 2003Q1 2004Q3 2006Q1 2007Q3 2009Q1 2010Q3
TFR Unemployment rate
-2.0000
1991Q1 1992Q4 1994Q3 1996Q2 1998Q1 1999Q4 2001Q3 2003Q2 2005Q1 2006Q4 2008Q3 2010Q2
TFR CPI growth rate
1991Q1 1992Q4 1994Q3 1996Q2 1998Q1 1999Q4 2001Q3 2003Q2 2005Q1 2006Q4 2008Q3 2010Q2
TFR HPI
1991Q1 1992Q3 1994Q1 1995Q3 1997Q1 1998Q3 2000Q1 2001Q3 2003Q1 2004Q3 2006Q1 2007Q3 2009Q1 2010Q3
TFR female labor force
13
Varia
bles Descriptions Sourc
e
The average number of children that would be born alive to 1,000 women during their lifetime if they were to pass their
childbearing ages 15-49 year experiencing the age-specific fertility rate prevailing in that year.
the
The base period is the fist quarter of 1991.
The female labor force population accounts for a percentage of the women civilian population over
Unem A percentage by dividing DGB