O R I G I N A L P A P E R
Fertility and Crime: Evidence from Spatial Analysis of TaiwanTsung Huang Jr.• Tsun-Feng Chiang•
Springer Science+Business Media New York 2015
Abstract Current literature indicates the possible effects of crime on family-related demographic events, including fertility. However, there is no empirical study that confirms this effect. The research purpose of this study is to fill this gap. Using county-level panel data during 2000–2010 from Taiwan, the estimated results of the fixed-effect spatial error models discovered that crime did have a significantly negative effect on fertility, although the magnitude was modest. Furthermore, fertility has spatial dependence: that is, some unobservable factors that promote fertility in a county will likely have positive impacts on the fertility of neighboring counties. From a general viewpoint, the esti-mated results of other control variables showed that eco-nomic factors were still the primary determinants of fertility.
Keywords Crime rate General fertility rate Fixed-effect spatial error model Taiwan
Among many subjects in the fields of population and household economics, fertility has been a primary focus in developed countries in recent years. Fertility is related not only to the development of the next generation through the supply of human capital, but also to the benefits of this generation by its interaction with pension plans. For ex-ample, Fenge and Meier (2005) and van Groezen et al. (2003) determined that the relationship between fertility and pensions was causal: that is, fertility affected future pension funds, and pensions affected today’s fertility rate. For Taiwan, the unusually low fertility rate has drawn public attention, and more studies are needed to interpret this scenario and to infer policy implications.
Figure1 shows the annual general fertility rate (GFR, hereafter) and total fertility rate (TFR, hereafter) in Taiwan from 1971 to 2011.1 The downward movement of both fertility rates is very obvious. Except for some small peaks in the year of the dragon,2i.e., 1976, 1988 and 2000, which were considered by some Taiwanese couples to be good years to intentionally give birth, the GFR and TFR declined each year. The lowest point was in 2010, when the TFR was below 900.
T. Huang Jr.
Department of Public Finance, National Chengchi University, No.64, ZhiNan Road, Sec.2, Wenshan District, Taipei 116, Taiwan
e-mail: firstname.lastname@example.org T.-F. Chiang (&)
School of Economics, Henan University, Kaifeng 475004, Henan, China
e-mail: email@example.com J.-N. Pan
General Education Center, Kainan University, No. 1 Kainan Road, Luchu, Taoyuan 33857, Taiwan
1 The general fertility rate is defined as the number of births per 1000
women of childbearing age (15–49). The total fertility rate is defined as the sum of the age-specific birth rates for 1000 women per cohort from age 15 to age 49 multiplied by 5.
2 The Chinese animal zodiac, literally translated as ‘‘born
resem-bling,’’ is a rotating cycle of 12 years, with each year being represented by an animal and its reputed attributes. The 12 animals are, in order, rat, ox, tiger, rabbit, dragon, snake, horse, goat, monkey, rooster, dog and pig. Generally speaking, the dragon is the most popular animal from the Chinese viewpoint.
Currently, Taiwan has the lowest fertility rate in the world. Taiwan’s government has acknowledged possible negative consequences of the low fertility rate in terms of population replacement, aging, and fiscal policies and has applied many incentive policies to promote the fertility rate. Chen (2012) summarized these incentive policies, including cash bonuses, childcare subsidies, education subsidies, etc. Moreover, tax reduction policies were em-ployed at the national as well as local level, such as by Taipei City.
The media often blames Taiwan’s low fertility on stagnant economic performance, the high cost of child-bearing, and the change in viewpoint regarding a woman’s role in the family.3Several economic factors and structural changes in society have been confirmed as contributing to the decreasing fertility rate in Taiwan (Chen2013; Poston
2000). Previous studies have implied that similar to other developed countries, when Taiwan achieves a certain level of social and economic development, a low fertility rate becomes inevitable. However, in addition to the socioe-conomic variables that would intuitively affect fertility, this study explores whether there are any other potential determinants overlooked by the empirical literature but crucial to low fertility in Taiwan.
Crime was treated as the most important factor in this study. Some studies have mentioned that crime could affect women’s childbirth either by changing a household’s life course (Wilson and Daly 1997) or by damaging men’s economic value in families and marriage markets (South and Messner 2000). However, to our knowledge, no em-pirical study has been conducted to support this hypothesis. Therefore, the purpose of this study is to fill this gap by quantifying the effect of crime on fertility under the con-sideration of the spatial dependence of fertility. This study
adopted panel data from 20 counties and cities in Taiwan over the period from 2000 to 2010 and used spatial panel econometric models to consider the spatial dependence of fertility.
This study was organized as follows. The related lit-erature was reviewed in the next section. Then the data sources and spatial panel econometric models for analyzing the impact of the crime rate on the GFR were introduced. After that, the estimation results of the empirical model were analyzed, and some specification tests were applied to examine the accuracy of the empirical model, followed by conclusions in the last section.
Several studies have explored the determinants of the fer-tility rate in Taiwan. Freedman et al. (1994) concluded that the decline in Taiwan’s fertility rate in the 1980s was primarily due to late marriages. In addition, Poston (2000) found that changes in social and economic status con-tributed to a substantial reduction in the fertility rate in Taiwan. Huang (2002) indicated that the personal tax ex-emption had a statistically significant and positive impact on the fertility rate in Taiwan. Huang (2003) further con-cluded that the conception rate was negatively affected by the unemployment rate. Huang et al. (2006) determined that college tuition and fees had a significantly negative influence on regional GFR in Taiwan. Chen (2013) and Lo (2012) observed that housing costs and female opportunity costs determined childbearing decisions. Chen (2012) also discussed the trends in marriage and fertility and further explored the evolution of population policies using a methodology similar to Poston (2000), taking possible determinants of the fertility rate into the estimations. However, there is no study discussing the relationship be-tween crime and fertility in Taiwan.
In fact, studies in economics have often focused on the economic determinants of fertility (Ekert-Jaffe and Stier
0 20 40 60 80 100 120 1971 1976 1981 1986 1991 1996 2001 2006 2011 Year G e n e ra l F e rt ility Ra te : b ir th pe r 1, 000 w o m e n 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 T o ta l F e rtility Ra te : bi rt hs pe r 1000 w o m e n
General Fertility Rate Total Fertility Rate
Fig. 1 TFR and GFR of childbearing age women, 1971–2011. Data Source: Ministry of the Interior, Taiwan
3 For example, Sui (2011), a BBC News report, mentioned that some
Taiwanese women were reluctant to have children. In addition, Tso (2009), a report of Time, indicated that most women were afraid of losing their jobs if they took time out to have a child.
2009; Hakim2003; Sobotka et al.2011) and the causality between fertility and other economic variables, such as the relationship between fertility intention and women’s working hours (Shreffler and Johnson 2013). Meanwhile, studies in sociology have tended to discuss fertility in the context of life course and demographic background. Crime plays an important role by changing life course and thus also demographic characteristics. The issue regarding the influence of crime on life course and the population process has been extensively studied (Hagan and Wheaton 1993; Sampson and Laub2003; Yamaguchi and Kandel1987). In addition, some studies focused on marriage issues have confirmed that crime can change marriage patterns (Lichter et al. 1991; Tucker and Mitchell-Kiernan 1995). In fact, some studies have noted that fertility would also be af-fected by crime. South and Messner (2000) indicated that crime would lessen the desirability of men in the marriage market through the possibility that their economic status was diminishing. As a result, women’s marriage and fer-tility patterns would change because of crime. Wilson and Daly (1997) also found that crime shortened life ex-pectancy, which led to a compressed life horizon. The timing of parenthood would be earlier for neighborhoods with lower life expectancies. Because life-course theory does not predict the effect of crime on fertility, this study adopts the viewpoints by South and Messner (2000) and proposes the hypothesis that the crime rate could negatively affect the fertility rate.
To our knowledge, there have been few works, par-ticularly empirical studies, analyzing the impact of crime rates on fertility rates in the past decade. Furthermore, most studies discussed the relationship between crime and demographic characteristics without controlling other de-terminants; therefore, the influence of crime on demo-graphic characteristics cannot be estimated accurately. This study attempts to fill this gap by estimating econo-metric models to quantify the effect of the crime rate on the fertility rate, keeping other factors constant and under the consideration of the spatial dependence of fertility rates.
Data and Methodology
The data adopted in this study were the official county-level panel data from 20 counties and cities over the period from 2000 to 2010 in Taiwan.4The primary variables of interest were the fertility rate and the crime rate. The fer-tility rate was represented by the GFR, defined as births per
thousand women between the ages of 15 and 49.5Figure2
presents the average GFR for each county and city over the research period of 2000–2010. The darker the shade, the higher the GFR. Counties and cities with a high average GFR are agricultural regions; counties and cities with a low GFR are more industrialized and urbanized. The range of the GFR is between 50 (Hsinchu County) and 30 (Keelung City).
Data in Fig.2 indicate counties and cities clustered to-gether based on the level of the GFR. To investigate the spatial dependence of the GFR, this study employed Mo-ran’s I, proposed by Moran (1950). The formula for Mo-ran’s I can be described as follows:
I¼Pn n i¼1 Pn j¼1wij Pn i¼1 Pn
j¼1wijðyi yÞðyj yÞ
In Eq. (1), i = j, and wijwas the inversed geographical distance between region i and j or an element in the ith row and jth column of a spatial weight matrix that defined the geographical distance between any two regions. The null hypothesis of this test was H0: there is no spatial correla-tion. If the test result rejected the null hypothesis, it implied that observations feature spatial correlation. Generally speaking, Moran’s I has an asymptotically standard normal distribution with values ranging from -1 to 1; the higher the absolute value of Moran’s I was, the stronger the spatial correlation would be. The z-value of Moran’s I was 1.61 and the p value was 0.054, showing that the null hypothesis of no spatial correlation was rejected at a 10 % significance level, and that there was positive spatial correlation of the GFR. Because both the figure and Moran’s I supported the spatial dependence of the GFR, this study used a spatial econometric model to examine the influence of the crime rate on the regional GFR in Taiwan.
Because the cross-sectional observations represented regions contingent or close to each other and the GFR had spatial dependence, this study adopted spatial econometric models for the county-level panel data. Doing so included the possibility that observations in one region could affect other regions, after controlling other variables. Ignoring the spatial dependence of the GFR could cause inefficient or even biased estimates (LeSage and Pace2009). This study adopted two spatial panel econometric models: the spatial error model (SEM) and the spatial autoregression model (SAR). The advantage of the SEM was that it did not
4 Because some regional-level variables were not available after
2010, this study limited the research period from 2000 to 2010.
5 According to Whittington et al. (1990), the theoretical advantage of
the GFR is that it is not as sensitive to changes in the age and sex structure of the population as the birthrate. However, when compared with the TFR, the GFR is more volatile because the former is age-adjusted, and the number of women in each age group is assumed to be the same (Huang et al.2006). In empirical studies, the GFR is also a common dependent variable for testing the determinants of fertility (Huang et al.2006; Poston2000; Whittington et al.1990).
require a theoretical foundation for a spatial interaction process as indicated by Anselin et al. (1996) and Ku et al. (2014). However, the SAR is the most common spatial econometric model in the literature because it can inves-tigate the direct impact of one county’s fertility on the fertility of its neighboring counties and vice versa. This study also provides results estimated from non-spatial panel data models as a comparison.
The SEM specified that unobservable factors in a region had a spatial reaction with factors in other regions. More explicitly, let yit be the dependent variable for region i at time t; using the notations of Elhorst (2010), the SEM was described as follows: yit¼ aiþ x0itbþ uit /it ¼ q XN j¼1 wij/jtþ eit; i; j¼ 1; 2. . .; N ð2Þ
where aiwas a scalar representing the individual intercepts capturing region i’s specific characteristics; xit was the k 9 1 matrix of explanatory variables; and b was a k 9 1 vector of estimates. ujt was the spatially autocorrelated error term, whose magnitude of the effect was adjusted by wij. The larger the distance, the smaller wij, and the smaller
the effect conveyed by ujt to uit. The term q was a scale parameter representing the average strength of the spatial dependence of ujt.eit was an error term that independently and identically followed a normal distribution with 0 mean and variance r2. The parameters b, q, r2were needed to be estimated using the maximum likelihood method.
The specification of the SAR was that the GFR in a region had a spatial relation with that in other regions. Using the notations of Elhorst (2010), the SAR was de-scribed as follows:
yit¼ aiþ qWyitþ x0itbþ eit ð3Þ
where q was a scale parameter representing the average strength for the spatial dependence of the GFR.
To determine which was more suitable for describing the data and for use as the baseline model, Akaike’s in-formation criterion (AIC) and the Bayesian inin-formation criterion (BIC) were calculated. The baseline model would have the lower AIC and/or BIC. Then, the test proposed by Hausman (1978) was adopted to determine whether the fixed-effect or the random-effect model was better. If the fixed-effect model was better than the random-effect model, aicould also be estimated as an intercept for the ith region, but it did not affect the values of other parameters and is rarely the interest of research.
Regarding the primary explanatory variable, crime rate, there were two measures. One was the total crime rate defined as the number of crimes committed per 100,000 people, including some civil violations. The other was the rate of offense defined as the number of offenses com-mitted per 100,000 people. The rate of offense is more relevant because many studies emphasize major offenses, such as homicide in Lichter et al. (1991), Tucker and Mitchell-Kiernan (1995) and Wilson and Daly (1997), in which major offenses led to shorter expected life, thereby compressing the life course and implying higher fertility rates. Therefore, the impacts of the rate of offense on fer-tility are expected to be more significant than other types of minor crime.
Other independent variables were selected based on the literature (Chen2013; Poston2000; Yang and Liu2014) and included the junior-high-or-below ratio6to measure educa-tional level, two variables for household finances (household disposable income and the home ownership ratio), the fe-male labor force participation rate, unemployment rate, the infant mortality rate, and the female ratio. It is worth noting that in addition to controlling these independent variables, this study also considers the impacts caused by the spatial dependence of the GFR while the influence of the crime rate on the GFR was examined.
Fig. 2 Average GFR of County and City from 2000 to 2010. Data Source: Ministry of the Interior, Taiwan
6 Junior high school education has been compulsory in Taiwan since
Because the response of the fertility rate to changes in any explanatory variables cannot be instantaneous, all in-dependent variables were taken with a 1-year lag, i.e., xit-1 instead of xit in Eq. (2). In addition, the spatial weight matrix was created based on the inversed Euclidean dis-tance between each county’s centroid. Finally, the literacy rate replaced the junior-high-or-below ratio, and household savings replaced household disposable income in the fixed-effect SEM for the robustness check. The definitions and descriptive statistics for all variables are presented in Table1.
After calculating the AIC and BIC, AIC and BIC were lower in the SEM than in the SAR in both the fixed-effect and the random-effect models. Therefore, it is concluded that the SEM is better than the SAR for this study.7 In
addition, the Hausman test failed to reject the null hy-pothesis that the fixed-effect SEM was more efficient than the random-effect model.8 Thus, this study selected the fixed-effect SEM as the baseline model.
Models 1 and 2 are the baseline models of the fixed-effect SEM with the total crime rate and the rate of offense as the crime variables, respectively. Their counterparts in the traditional fixed-effect panel model are Models 3 and 4. According to Table2, the results of Model 1 are quite similar to those of Model 2. Therefore, the result of Model 1 is the case discussed in this study. It is shown in Table2
that the total crime rate has a statistically significant and negative effect on the GFR. As mentioned by South and Messner (2000), in a population with a high likelihood of committing a crime, the value of men diminishes in the marriage market and family, resulting in women’s un-willingness or inability to give birth. The estimated pa-rameter in the SEM is the marginal effect of the total crime rate on the GFR, meaning that when the total crime rate increases by 1000 crimes committed per 100,000 people, Table 1 Definitions and descriptive statistics of variables
Variables Definitions Max (Min) Mean Standard
deviation General fertility rate (GFR) The number of births per 1000 women of
childbearing age (15–49)
50.18 (30.00) 38 5.59 Measure of crime
Total crime Rate (TCR) The number of crimes, including some civil violations, committed per 100,000 people
2922.22 (951.14) 1885 519.62 Rate of offense (RO) The number of offenses committed per
2926.40 (954.48) 1868 536.36 Education
Junior high or below ratio The percentage of the population older than 14 years with a junior high school diploma or lower education level (%)
57.42 (18.15) 41 10.25
Literacy rate The percentage of the literate population older than 14 years (%)
98.62 (93.61) 97 1.41 Household finance
Household disposable income Yearly disposable household income (NTD) 1,249,802 (630,580) 823,142 160,612 Household savings Yearly household savings (NTD) 297,975 (126,060) 189,943 40,427 Home ownership ratio The percentage of households who own their
94.10 (80.42) 88 3.89 Female labor force participation rate The percentage of female population in the
labor force (%)
62.65 (49.11) 58 3.22 Unemployment rate The percentage of the labor force who are
4.80 (4.00) 4 0.23
Infant mortality rate The number of infants who die at less than 1 year old per 1,000 births
8.67 (4.16) 5 1.11
Female ratio The ratio of female to total population 0.51 (0.46) 0.49 0.01 Sources: Ministry of Interior, Taiwan
7 The AIC and BIC of the fixed-effect SEM were 900.1759 and
930.7186, respectively. The AIC and BIC of the fixed-effect SAR were 1034.578 and 1136.386, respectively.
8 The Chi square statistic and the p value from the Hausman test were
the GFR will decline by approximately 3 births per 1000 women of childbearing age (15–49) for a county or city. This result indicates that the increase in the total crime rate is not the primary cause of the low fertility rate in Taiwan. Other factors that have a significant impact on the GFR are the unemployment rate, the junior-high-or-below educational status ratio, household disposable income, the infant mortality rate, and the female ratio. The estimated coefficients of these variables have a negative or positive sign, which is consistent with our expectations. The impact of the unemployment rate on the GFR was statistically significant and negative at a 1 % significance level. This result implies that a higher unemployment rate will cause a lower GFR. This finding is consistent with Huang (2003), which found that household disposable income had a sta-tistically and significantly positive impact on the GFR, suggesting that children were normal goods in the eyes of the Taiwanese people. In addition, the estimated coefficient of the junior-high-or-below ratio is statistically and sig-nificantly positive, meaning that having more people with a junior high school diploma or a lower educational level would cause a lower GFR. There are several ways to in-terpret the mechanism behind the inverse effect of
education on the GFR: for example, schooling could change the traditional perspective of a woman’s role in a family, or it could change marriage patterns (Basu2002). Moreover, both the infant mortality and the female ratio have a statistically and significantly positive impact on the GFR. When the infant mortality rate is higher, people are more likely to have more births to compensate for the lost infants, further increasing the GFR. In regards to the positive impact of the female ratio, this study asserts that people have more opportunity to marry when the female ratio is larger, given that the gender structure is imbalanced toward females. Finally, the positive and significant esti-mated q indicates that some unobservable positive factors in a county will likely have a positive effect on the GFR of its neighboring counties.
In addition to the fixed-effect SEM, Table2 also pre-sents the estimated results of the traditional fixed-effect model without considering spatial effects for comparison. It is very obvious that without taking spatial effects into account, the negative impacts of the total crime rate, the rate of offense, and the unemployment rate were underes-timated. However, the positive impacts of the junior-high-or-below ratio, household disposable income, the infant Table 2 Empirical results of
the fixed effect SEM and panel model
p\ .10; * p \ .05; ** p \ .01; *** p \ . 001
Independent variables Fixed-effect SEM Fixed-effect panel model Model 1 Model 2 Model 3 Model 4 Total crime rate -0.0029**
(0.0006) -0.0014* (0.0006) Rate of offense -0.0027** (0.0006) -0.0013* (0.0006) Female labor force participation rate -0.0653
(0.0546) -0.0700 (0.0547) -0.1344* (0.0603) -0.1362* (0.0602) Unemployment rate -1.8860** (0.4515) -1.8611** (0.4518) -1.1583** (0.2179) -1.1539** (0.2179) Junior high or below ratio 1.3857***
(0.1605) 1.3934** (0.1600) 1.5217*** (0.1272) 1.5252*** (0.1269) Household disposable income 8.07 9 10-6*
(3.45 9 10-6) 8.10 9 10-6* (3.46 9 10-6) 8.55 9 10-6* (4.13 9 10-6) 8.58 9 10-6* (4.13 9 10-6)
Home ownership ratio 0.0038 (0.0553) 0.0026 (0.0556) 0.0666 (0.0661) 0.0652 (0.0661) Infant mortality rate 0.2286*
(0.1087) 0.2330* (0.1090) 0.3247** (0.1229) 0.3261** (0.1229) Female ratio 5.0730** (1.3602) 5.1556** (1.3580) 5.1722** (1.5766) 5.2211** (1.5687) q 0.6658** (0.1277) 0.6595** (0.1309) r2 3.2602*** (0.3466) 3.2757*** (0.3483) Log-likelihood -408.8701 -409.1674 R2 0.2585 0.2581 0.2618 0.2614
mortality rate, and the female ratio were overestimated when spatial effects were neglected. It is thus concluded that the existence of spatial dependence is crucial to the estimated results. If spatial dependence is ignored, the es-timations will be biased.
A robustness check was conducted to examine whether the influence of the crime rate and other explanatory variables were stable and robust. Models 5 and 6 used household savings to replace household disposable income in Models 1 and 2, respectively. The estimated results of Models 5 and 6 in Table3 were consistent with their counterparts in Models 1 and 2. In addition, Models 7 and 8 used the literacy rate to replace the junior-high-or-below ratio in Models 1 and 2.9The impact of both types of crime rate on the GFR is still statistically and significantly
negative. Therefore, the conclusion of a negative influence from the crime rate on the GFR is very robust, regardless of which educational and household finance variables are adopted in the empirical model.
The purpose of this study was to explore a rarely discussed but important issue regarding the impact of the crime rate on the fertility rate in Taiwan. This study used official county-level panel data from 20 counties and cities over the 2000–2010 period in Taiwan and estimated a fixed-effect SEM. The primary finding of this study was that that after controlling other factors and considering spatial depen-dence, the crime rate had a statistically and significantly negative impact on the GFR in Taiwan, regardless of whether total crime rate or the rate of offense was used as the crime variable in the empirical model. This conclusion Table 3 Robustness check for
the fixed-effect SEM
Symbol # indicates the independent variables used to conduct the robustness checks p\ .10; * p \ .05; ** p\ .01; *** p \ . 001
Independent variables Fixed-effect SEM
Model 5 Model 6 Model 7 Model 8 Total crime rate -0.0026**
(0.0006) -0.0033** (0.0006) Rate of offense -0.0025** (0.0006) -0.0031** (0.0006) Female labor force participation -0.0723
(0.0542) -0.0765 (0.0541) -0.0687 (0.0506) -0.0740 (0.0508) Unemployment rate -1.7760** (0.4297) -1.7507** (0.4279) -1.3838** (0.3680) -1.3685** (0.3674) Junior high or below ratio# 1.3390***
(0.1500) 1.3439*** (0.1496) Literacy rate# -4.2806** (0.3800) -4.2859** (0.3807) Household disposable income# 5.72 9 10-6
(3.33 9 10-6) 5.72 9 10-6 (3.34 9 10-6) Household savings# 9.41 9 10-6* (4.72 9 10-6) 9.69 9 10-6* (4.72 9 10-6) Home ownership ratio 0.0289
(0.0544) 0.0277 (0.546) 0.0341 (0.0563) 0.0332 (0.0540) Infant mortality rate 0.2095
(0.1078) 0.2135* (0.1077) 0.1838 (0.1018) 0.1877 (0.1021) Female ratio 4.7786** (1.3650) 4.8422** (1.3627) -1.1978 (1.2011) -1.1181 (1.2095) q 0.6442** (0.1242) 0.6389** (0.1263) 0.6800** (0.0938) 0.6759** (0.0969) r2 3.3029*** (0.3472) 3.3140*** (0.3482) 3.0424*** (0.3159) 3.0640*** (0.3185) Log-likelihood -409.5865 -409.7691 -402.3702 -402.9526 R2 0.2455 0.2456 0.3332 0.3310
9 The literacy rate had a negative effect on the GFR, as Caldwell
(1980) noted, by shifting the economy away from labor-intensive family-based production; mass education led to a lower level of fertility.
is reasonable and consistent with our expectations. In ad-dition, this study also supports the public’s opinion that economic factors, such as household disposable income and household savings, are essential to families’ decision to have children. The unemployment rate had a negative impact on the GFR, but the junior-high-or-below ratio, infant mortality, and the female ratio have a positive im-pact on the GFR. Finally, the positive and significant es-timated q indicates that some unobservable positive factors in a county will likely have a positive effect on the GFR of its neighboring counties. The conclusions drawn from the empirical results are robust, and once the spatial depen-dence is ignored, estimations will be biased. The first limitation of this study is that the county-data are not available before 2000 and after 2010. Therefore, it is not certain that the effect of crime on fertility is circumstance only of recent years or if the effect has existed for a longer period of time. The second limitation is the inability to identify the channel between crime and fertility. Since the theory predicts that by affecting marriage, crime could have an effect on fertility, household profiles might help to confirm this prediction. However, to our knowledge, a thorough household survey regarding marriage has still not been conducted in Taiwan.
Because the issue regarding low fertility rates is crucial to the economic development of many developed and de-veloping countries, especially Taiwan, the conclusions provided by this study can make some contributions in this regard. According to the results from the empirical models, this study suggests a pro-natal policy in which the gov-ernment increases efforts to reduce the total crime rate. Doing so will not only secure people’s lives but will also encourage people to have more children. Although the total crime rate is not the most important factor affecting the GFR, it is a statistically significant factor.
Other feasible pro-natal policies are also proposed by this study. An increase in household disposable income can raise the GFR, according to empirical results. Therefore, if the government can adopt policies that increase household disposal income, such as mitigating the family tax burden, providing subsidies to families, or stimulating economic growth, people will be encouraged to have more children, further increasing the GFR. In addition, the government must promote employment opportunities by implementing labor policies. When the unemployment rate decreases, it will cause an increase in the GFR. Finally, it is found that the female ratio has a positive effect on the GFR. To in-crease the female ratio, the government can consider re-leasing restrictions on the immigration of females from foreign countries, given that cross-country marriage is becoming common and popular in Taiwan.
Acknowledgments The authors acknowledge the financial support provided for this research by the National Science Council, Taiwan (NSC 101-2410-H-004-206-). The authors would like to thank all participants of the 2014 Biennial Conference of the Asian Consumer and Family Economics Association (ACFEA) at National Chengchi University, Taipei, on July 8–11, 2014, particularly Dr. Elizabeth M. Dolan, the editor-in-chief of Journal of Family and Economic Issues, Dr. Clifford Robb, the associate editor of Journal of Consumer Af-fairs, and Dr. Jing Jang Xiao, the editor of Journal of Financial Counseling and Planning, for their helpful and constructive com-ments. The authors also deeply appreciate two anonymous reviewers. All views and errors are solely those of the authors.
Anselin, L., Bera, A. K., Florax, R., & Yoon, M. J. (1996). Simple diagnostic tests for spatial dependence. Regional Science and Urban Economics, 26(1), 77–104. doi:10.1016/0166-0462(95) 02111-6.
Basu, A. M. (2002). Why does education lead to lower fertility? A critical review of some of the possibilities. World Development, 30(10), 1779–1790. doi:10.1016/s0305-750x(02)00072-4. Caldwell, J. C. (1980). Mass education as a determinant of the timing
of fertility decline. Population and Development Review, 6(2), 225–255. doi:10.2307/1972729.
Chen, Y.-H. (2012). Trends in low fertility and policy responses in Taiwan. The Japanese Journal of Population, 10(1), 78–88. Retrieved from http://www.ipss.go.jp/webj-ad/webjournal…/ Web%20Journal_Vol.10_04.pdf.
Chen, W.-Y. (2013). Does housing cost affect birth rates in Taiwan? The ADL test for threshold co-integration. Romanian Journal of Economic Forecasting, 16(3), 90–103. Retrieved from http:// www.ipe.ro/rjef/rjef3_13/rjef3_2013p90-103.pdf.
Ekert-Jaffe, O., & Stier, H. (2009). Normative or economic behavior? Fertility and women’s employment in Israel. Social Science Research, 38, 644–655. doi:10.1016/j.ssresearch.2009.02.005. Elhorst, P. (2010). Spatial panel data models. In M. M. Fischer & A.
Getis (Eds.), Handbook of applied spatial analysis: Software tools, methods and applications. Berlin: Springer.
Fenge, R., & Meier, V. (2005). Pensions and fertility incentives. Canadian Journal of Economics, 38(1), 28–48. doi:10.1111/j. 0008-4085.2005.00267.x.
Freedman, R., Chang, M.-C., & Sun, T.-H. (1994). Taiwan’s transition from high fertility to below-replacement levels. Studies in Family Planning, 25(6), 317–331. doi:10.2307/ 2137876.
Hagan, J., & Wheaton, B. (1993). The search for adolescent role exits and the transition to adulthood. Social Forces, 71(4), 955–980. doi:10.2307/2580126.
Hakim, C. (2003). A new approach to explaining fertility patterns: Preference theory. Population and Development Review, 29(3), 349–374. doi:10.1111/j.1728-4457.2003.00349.x.
Hausman, J. A. (1978). Specification tests in econometrics. Econo-metrica, 46(6), 1251–1271. Retrieved fromhttp://www.jstor.org/ discover/10.2307/1913827?uid=2&uid=4&sid=21105342207 423.
Huang, J.-T. (2002). Personal tax exemption: The effect on fertility in Taiwan. The Developing Economies, 40(1), 32–48. doi:10.1111/ j.1746-1049.2002.tb00909.x.
Huang, J.-T. (2003). Unemployment and family behavior in Taiwan. Journal of Family and Economic Issues, 24(1), 27–48. doi:10. 1023/A:1022431003513.
Huang, J.-T., Kao, A.-P., & Hung, W.-C. (2006). The influence of college tuition and fees on the fertility rate in Taiwan. Journal of Family and Economic Issues, 27(4), 626–642. doi:10.1007/ s10834-006-9037-4.
Ku, Y.-C., Bloomfield, P., & Ghosh, S. (2014). A flexible observed factor model with separate dynamics for the factor volatilities and their correlation matrix. Statistical Modelling, 14, 1–20. doi:10.1177/1471082x13490016.
LeSage, J., & Pace, K. (2009). Introduction to spatial econometrics. Boca Raton: Chapman and Hall/CRCm.
Lichter, D., LeClere, F., & McLaughlin, D. (1991). Local marriage market conditions and the marital behavior of black and white women. American Journal of Sociology, 96(4), 843–867. doi:10. 1086/229610.
Lo, K.-T. (2012). The crowding-out effect of homeownership on fertility. Journal of Family and Economic Issues, 33(1), 108–117. doi:10.1007/s10834-011-9256-1.
Moran, P. A. P. (1950). Notes on continuous stochastic phenomena. Biometrika, 31(1), 17–23. doi:10.2307/2332142.
Poston, D. L. (2000). Social and economic development and the fertility transitions in Mainland China and Taiwan. Population and Development Review, 26, 40–60. Retrieved from http:// www.jstor.org/stable/3115211.
Sampson, R. J., & Laub, J. H. (2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 319–339. doi:10.1111/j.1745-9125.2003. tb00997.x.
Shreffler, K. M., & Johnson, D. R. (2013). Fertility intentions, career considerations and subsequent births: The moderating effect of women’s work hours. Journal of Family and Economic Issues, 34(3), 285–295. doi:10.1007/s10834-012-9331-2.
Sobotka, T., Skirbekk, V., & Philipov, D. (2011). Economic recession and fertility in the developed world. Population and Develop-ment Review, 37(2), 267–306. doi:10.1111/j.1728-4457.2011. 00411.x.
South, S., & Messner, S. (2000). Crime and demography: Multiple linkages, reciprocal relations. Annual Review of Sociology, 26, 83–106. doi:10.1146/annurev.soc.26.1.83.
Sui, C. (2011, August 15). Taiwanese birth rate plummets despite measures. BBC News. Retrieved from http://www.bbc.co.uk/ news/world-asia-pacific-14525525.
Tso, N. (2009, December 7). Why has Taiwan’s birthrate dropped so low? Time. Retrieved fromhttp://content.time.com/time/world/ article/0,8599,1945937,00.html.
Tucker, M. B., & Mitchell-Kiernan, C. (1995). African American marital trends in context: toward a synthesis.). In M. B. Tucker & C. Mitchell-Kiernan (Eds.), The Decline in marriage among African Americans. New York: Russell Sage Foundation.
van Groezen, B., Leers, T., & Meijdam, L. (2003). Social security and endogenous fertility: Pensions and child allowances as Siamese twins. Journal of Public Economics, 87(2), 233–251. doi:10. 1016/s0047-2727(01)00134-7.
Whittington, L., Alm, L., & Peters, E. (1990). Fertility and the personal exemption: Implicit pronatalist policy in the United States. American Economic Review, 80(3), 545–556. Retrieved fromhttp://www.jstor.org/stable/2006683.
Wilson, M., & Daly, M. (1997). Life expectancy, economic inequal-ity, homicide, and reproductive timing in Chicago neighbour-hoods. British Medical Journal, 314(26), 1271–1274. doi:10. 1136/bmj.314.7089.1271.
Yamaguchi, K., & Kandel, D. (1987). Drug use and other determi-nants of premarital pregnancy and its outcome: A dynamic analysis of competing life events. Journal of Marriage and Family, 49(2), 257–270. doi:10.2307/352298.
Yang, W.-S., & Liu, Y.-Y. (2014). Gender imbalances and the twisted marriage market in Taiwan. In D. L. Poston, W.-S. Yang, & D. N. Farris (Eds.), The family and social change in Chinese societies. Netherlands: Springer.
Tsung Huang Jr. is a Distinguished Professor and Chair of Public Finance at National Chengchi University, Taiwan. He received his Ph.D. in Economics from University of Washington, Seattle, USA. His research interests are Family Economics, China Economy, Taxation, and Public Finance. His researches have been published in Journal of Family and Economic Issues, China Economic Review, Public Finance Review, Singapore Economic Review, Contemporary Economic Policy, Defense and Peace Economics, Developing Economy, etc.
Tsun-Feng Chiangis an Assistant Professor at School of Economics, Henan University, China. He received his Ph.D. in Economics from North Carolina State University. His research interests are Household Finance, Microeconometrics and China Economy. His research has been published in China Economic Review.
Jiun-Nan Panis an Assistant Professor at General Education Center, Kainan University, Taiwan. He received his Ph.D. in Public Finance from National Chengchi University, Taiwan. His research interests are Family Economics, China Economy, Taxation, and Public Finance. His researches have been published in Journal of Family and Economic Issues, and China Economic Review.