A Cohort Analysis of Housing Choices in Taiwan. Following the Cohort of Female

A Cohort Analysis of Housing Choices in Taiwan Following the Cohort of Female (first draft, please do not quote) Li-Min Hsueh and Chih-Lung Yen + Paper Submitted to the 12 th Asian Real Estate Society Annual Conference, Macau Li-Min Hsueh is a Professor in the Department of International Business, China University of Technology. Mailing address: 56, Section 3, Tsin-Lun Road, Taipei, Taiwan, 116. e-mail: lmhsueh@cute.edu.tw. + Chin-Lung Yen is a graduate student in the Department of Economics, National Taiwan University. 1

Abstract The home ownership rate in Taiwan gradually increased as time goes by. It has reached a very high level, at 82.2% in 2000 census. Housing behavior, such as tenure choice, housing demand, etc. has been researched by many scholars in Taiwan. However, the effect of birth cohort on housing decisions has been neglected. Age has two roles in a cross-sectional household data. It reflects the life-cycle needs and also reflects the different macro-environment of different cohorts at the time of birth. The research of Hsueh and Yen (2006) was the first to explore this area in Taiwan. It designates male head of a household to represent the household and follow the male cohort. Using the same method and same data of Hsueh and Yen (2006), this research changes an angle to follow female cohort. As time changes, the status of female in family decisions has significantly raised. At the same time, the percentage of households with female head is also significantly increased. All these developments justify a research to follow the female in a cohort analysis of housing choices and compare its result with that of male cohort. Hsueh and Yen (2006) found that earlier cohorts in Taiwan have significant advantage in becoming homeowners. This effect caused the average age of homeowners to increase over time, which was mistakenly thought to be caused by the aging of population in the past. This phenomenon is also found in following female cohort. This research also finds that the age effect and birth cohort effect are very different for female head and female non-head. Although on average, the female heads have longer education years and higher job participation rate, their housing welfare seems to be in disadvantage, which is mainly caused by the fact that female head households have only one earner. Nevertheless, the gap between head and non-head is narrower when the birth cohort is younger. Key words: cohort effect, tenure choice, housing demand, female cohort 2

Introduction The home ownership rate in Taiwan gradually increased as time goes by. It has reached a very high level, at 82.2% in 2000 census. Housing behavior, such as tenure choice, housing demand, etc. has been researched by many scholars in Taiwan. However, the effect of age cohort on housing decisions has been neglected. Age has two roles in a cross-sectional household data. It reflects the life-cycle needs and also reflects the different macro-environment of different cohorts at the time of birth. Hsueh and Yen (2006) was the first to explore this area in Taiwan. They found that earlier cohorts in Taiwan have significant advantage in becoming homeowners. This effect caused the average age of homeowners to increase over time, which was mistakenly thought to be caused by the aging of population in the past. Using the same method and same data of Hsueh and Yen (2006), this research will change a perspective to follow female cohort. Housing choice is typically a household behavior, which needs to take care of the needs of every household member. At a male dominated Chinese society, it is reasonable to believe that in most of cases, male household head makes decision for all household members. Therefore, in the past, to follow male household head is a natural choice in a cohort analysis of housing choice behavior, including Hsueh and Yen (2006). However, as time changes, the status of female in family decisions has significantly raised. At the same time, households with no adult male, such as single mother households, single-female households also significantly increased. All these developments justify a research to follow the female in a cohort analysis of housing choices and compare its result with that of male cohort. More specifically, we will build a model takes account of effects of census year-age-birth cohort following female in each household in studying housing tenure choice and 3

consumption decisions. The individual household data from the Population and Housing Census of 1980, 1990 and 2000 of Taiwan will be used to estimate the model. Literature Review Dowell Myers (1999) in a survey article stated that the effect of cohort has been less discussed in the housing literature. He mentioned several misinterpretation of the past research results using one cross-sectional data. For example, past research assumes that elderly would return from the suburbs to the cities. In fact, the elderly observed in the cities had always lived in those cities, the young families in the suburbs were located there because the suburbs were newly built. He also compared the advantages and disadvantages of using cohorts from repeated cross-sectional data and panel (or retrospective) data in the longitudinal interpretation of housing career. He asserted that cohorts from repeated cross-sections are better in distinguishing between age and cohort, estimating cumulative changes, sample representativeness and sample size. Myers, Megbolugbe and Lee (1998) developed a double cohort model, incorporating both birth cohort and immigration cohort to compare the homeownership rate over time of native born and immigrant households. Crossley and Ostrovsky (2003) studied the cohort effect of Canadian housing careers by compiling a quasi-panel from repeated cross-sectional surveys over the period 1974-1999. They chose to follow cohorts of female across ages and attributing to each woman the housing of her household. This is the same as our study. However, our study go one step further to compare the result of female with that of male, and also compare the difference between the female as head and non-head. As for the housing studies in Taiwan, tenure choice and housing demand is a very crowded area. Many research have analyzed different aspects of homeownership, including this author (Hsueh and Chen, 1999). This author is the first one to incorporate birth cohort 4

effect in the housing studies (Hsueh and Yen, 2006). However, cohort effect has been analyzed in other fields of social sciences in Taiwan, such as savings behavior (Deaton and Paxon, 1993), cross period labor substitutability (Chang and Chu, 1996), poverty (Leu, Wang and Wang, 1999), etc. From these articles, we can find that income of younger generations in Taiwan has increased very fast, consumption also increased very fast, and poverty rate is decreasing. After several decades of fast economic growth in Taiwan, these favorable economic conditions are expected. These results imply the affordability of housing for younger generations is increasing. However, the housing price in Taiwan grew also very fast. In addition, the housing markets have gone through several violent cycles causing vibrations in housing price. These developments have resulted in that earlier cohorts in Taiwan have significant advantage in becoming homeowners. For the female in Taiwan, their education level and job-participation rate have improved in an increasing rate than that of male. However, the female poverty is still a phenomenon that can be observed like in other countries. What will be the effect of these different developments on the female housing decisions is the focus of this research. Data and Descriptive Statistics Data Source and Sample Selection The household data from the Household and Housing Census of 1980, 1990 and 2000 in Taiwan are used in this study. These censuses were conducted by the Directorate General of Budget, Accounting and Statistics (DGBAS) which is in the Executive branch of Taiwan Government. Housing choice is a household decision making. So, households rather than individuals are the observation unit of this study. In order to follow the cohort of female, first of all, we 5

deleted all households that do not have any adult female. Next, one female has to be designated as the object to be studied for each household. For those households, if the household head is female which is identified in the census data, then she is naturally designated; if the household head is not female, then the eldest female in the household is designated. For those households with female head, in high probability it has no adult male in the household. Therefore, the female household head is the only or at least the major decision maker in the household. For those households with male head, his wife should have been the better choice. Unfortunately, we can not distinguish the relationships between household members for the 1980 census. Consequently, it is impossible to designate wife as the object to be studied. For the consistency of definition, we designate the eldest female in the household as the object for all three censuses. The eldest female could be the mother or the wife of household head. They usually share the responsibility in housing decision. We will call all the female population who are selected to be studied as female decision-makers. We expect that female as household head and not as household head will be very different in housing choices. We choose female with age 15 to 84 as the age range of object. Age 15 is chosen because it is the lower bound to be included in the labor force survey. With the choices of age range and census years to be studied, the birth years of objects can also be found as between 1896 and 1985. After the selection, there are 2,873,451, 3,706,218 and 5,314,770 households with female decision-makers respectively for 1980, 1990 and 2000 census. Ten percent of them are randomly chosen as samples for the econometric analysis. Before going to the econometric analysis, we will first briefly describe our data using the whole population of female decision-makers. Descriptive Statistics 6

Here, in order to understand the demographic changes in the past 20 years, we will first describe the age distribution of the whole female decision-makers, and the share of female household heads in the whole female decision population in the three census years. Then, the homeownership rate and the living space per person of households by age distribution and by birth cohort of female decision-makers will be discussed 1. 1. The increase of female decision population The numbers of female decision-makers are increased greatly from 2,873,451 persons to 5,314,770 persons between 1980 and 2000. They occupied 15.11% of total population of Taiwan in 1980 and 24.91 % in 2000. The age profile of female decision makers is shown in Figure 1 2. From Figure 1, we can see that the number of female between age 35 and 55 increased sharply in the 2000 census. Figure 1 Age Distribution of Female Decision Makers 2. The increase of female household heads As mentioned in earlier section, there are two types of households in the population of female decision-makers, i.e. the oldest female in the household and female household heads. As in the male dominated Chinese culture, a female head implies that she is the only or the most important source of income in the household. With a very high probability, she is a 1 For the ease of showing data, we pick the data points in a 5 year interval. 2 The age distribution is shown at 5 year interval using the data at that point of the age. 7

single-parent or the only person in the household. Figure 2 shows the age distribution of the share of female heads in the female decision population. From Figure 2, we can find that the shape of curves is rather similar for the 1980 and 1990; in both years, the share of female heads increased with the age, from about 10 or 20 percent at age 20 to 30 or 40 percent at age 65 or 75, then it declined. However, the age distribution of the share of female heads is very different in 2000. The share is much higher than before in all ages; and it stays at about 40 percent from age 20 to 80. This shows the relative importance of female heads in the female decision population increased significantly in the last 20 years. Figure 2 The Share of Female Heads by Age Distribution of Female Decision-makers 3. Homeownership rate The age profile of homeownership rate is shown in Figure 3. From the age profile, we can find that for female decision makers between age 15 to 20, their ownership rate is decreasing, and between age 20 to 50 their ownership rate is increasing. This profile is similar to that of male. The homeownership rate of female between age 20 to 77 in the 1980 census is lower than that of other two censuses of the same age range. But this phenomenon is not found in the male sample. In addition, the highest point in ownership rate is at age 47 in 1980 census, and in age 57 in 1990 census, and in age 67 in 2000 census. This shows they are in the 8

same birth cohort. Figure 3 Homeownership Rate by Age Distribution of Female Decision Makers Figure 4 shows the ownership rate by birth cohort in the age of census year. Each birth cohort has three points indicates the ownership rate at three census year, 1980, 1990 and 2000. From Figure 4 we can find that most of the birth cohorts show the increase of homeownership rate between 1980 and 2000. The younger cohorts show a steeper increase in the home ownership rate. Four birth cohorts who are born earlier than 1935 show slight decrease in their homeownership rate between 1990 and 2000, probably due to they have entered old age. In addition, we can find that 1930 and 1935 birth cohorts show a higher homeownership rate than that of neighboring cohorts in three census years. This can also be seen in the male sample. Figure 4 Homeownership Rate by Birth Cohort of Female Decision Makers Birth Year 9

4. Living space per person The decision of homeownership of a household has two considerations, consumption and also investment. Sometimes, it is difficult to disentangle these two implications. Therefore, living space is a better measurement of housing welfare of a household. To control for the effect of household size, we decide to use living space per person as the measurement of living space. It is defined as the total floor space divided by the number of persons in the household. However, the effect of household size will still partially remain because of the effect of scale of economy in the living arrangement 3. Figure 5 shows living space per person by the age distribution of female decision-makers. It shows three rather parallel curves for three census years which indicates the living space increased census by census. It has a U-shaped age distribution between age 20 and 75. This reflects the changes of household size in life-cycle. The U-shaped curve shifted 5 years older for the male population. This shows that on average male got married 5 years older than female. The lowest point is around 40 years old in female population, and 45 years in male population in the 1980 and 1990 census. But the lowest points are about 5 years younger in the 2000 census for both male and female population. This change in interesting; it is probably caused by the fact that younger generations are raising less kids then that of older generations. Hence, they reached the lowest points of living space per person in younger age. Figure 6 shows the living space per person by birth cohort. It shows that for most of the birth cohorts, the curves are almost straight and parallel to each other. This means the increasing rates are about the same between 1980 and 1990, and between 1990 and 2000. In the same time, birth cohort does not make difference in the increases of living space over the years 1980 to 2000. However, for several younger cohorts, the increasing rate is faster 3 The larger the household size, the more people share the common living spaces. Therefore, living space per person as defined in this study will be smaller when household size is larger other things being equal. 10

between 1980 and 1990 then that of 1990 and 2000. Figure 5 Living Space per Person by the Age Distribution of Female Decision-makers Unit: Ping *1 ping = 3.3057 square meter Figure 6 Living Space per Person by Birth Cohort of Female Decision-makers Unit: Ping Birth Year *1 ping = 3.3057 square meter Modeling and Variable Definition The housing tenure choice and housing consumption which is measured by living space per person are the most important housing decisions that every household has to face with. We are interested to know to what extent age and birth cohort affect on theses two decisions. 11

These two decisions are not mutually independent. Non-home owners may want to save more money for the mortgage down payment, hence to choose a smaller living space. On the other hand, home owners may live in a smaller living space because the cost of owning a residence is much higher than that of renting in Taiwan. Therefore, these two decisions will be jointly estimated. The home tenure choice is a binary choice, so a conditional probability model will be estimated first. Then the predicted probability of homeownership of a particular household will be included in estimating the living space equation. Since living space is a continuous variable, a linear model is used. Models for tenure choice and living space per person can be shown in Equation (1) and Equation (2) as follows: ( ho y a c) f y a c X1 Pr = 1,, = (,,, )...(1) PerA g( y, a, c, X, Hˆ ) =.(2) 2 Where in Equation (1), ho=1 means the household owns its own residence. ( ho = ) Pr 1 shows the conditional probability of owning a residence. PerA stands for the living space per person. y, a, and c are vectors which stand for census year, age and birth cohort respectively. X 1 and X 2 are vectors which contain other control variables. f is a cumulative probability function, and g is a linear function. For vectors of y, a, c, they are all dummy variables. The y vector has 3 variables, y0, y1, and y3, stand for 1980, 1990 and 2000 census year. The a vector has 14 variables, a0 to a13, represent 14 age groups. Each group contains 5 years in age, starting from age 15 to 85. The c vector has 18 variables, c0 to c17, represent 18 birth cohort groups. Each group contains 5 years in birth year, starting from 1896 to 1986. Because individual household data will be used in the estimation, the differences among 12

households that might also have an effect on housing decisions have to be controlled. For Equation (1), X 1 include a vector of variables for marriage status (MARRIED, DIVORCE and WIDOW), participate in the job market or not (WORK), years of education (EDU, EDU_SQUARE), household size (MEMBER) and whether a female household head or not (HEAD). Marriage status can be a proxy for whether there is a spouse in the household to share the financial responsibility in housing decisions. WORK and education level can be proxies for the income of the female decision-makers. HEAD is important because it differentiates two different types of female decision-makers. For Equation (2), X 2 include MEMBER, WORK and HEAD. For identifying Equation (2), variables for marriage status and education are not included in Equation (2). In addition, the predicted probability of whether the household owns its residence or not ( Ĥ ) from Equation (1) is also included. Notations and definitions for all variables are shown in Table 1. The descriptive statistics of variables other than year-age-cohort variables by female head and non-female head is shown in Table 2. The descriptive statistics by homeowner or not is shown in Table 3. From Table 2, we can find that the homeownership rate is lower for the female head households, and their living space per person is also lower. This seems to show that the housing welfare of female head households is in disadvantages. However, the female heads have higher average education and higher job market participation rate. The family size is smaller for the female-head households. As for the marriage status, the female heads have higher rate in the categories of unmarried, divorced and widowed, and lower rate in married. From Table 3, we can find that female decision makers who are not homeowners have higher percentage in the job market and have higher education. The living space per person for the non-homeowners is slightly smaller than that of homeowners. The family size of homeowners is on average 0.6 persons larger than that of non-owners. As for the marriage 13

status, homeowners have higher rate in married and widowed, and lower rate in unmarried and divorced than that of non-owners. Table 1 Variable Notation and Definition Notation Definition Notation Definition Year Group Cohort Group y0 =1 if in 1980 census c0 =1 if born in 1896~1900 y1 =1 if in 1990 census(control group) c1 =1 if born in 1901~1905 y2 =1 if in 2000 census c2 =1 if born in 1906~1910 Age Group c3 =1 if born in 1911~1915 a0 =1 if age is 15~19 c4 =1 if born in 1916~1920 a1 =1 if age is 20~24(control group) c5 =1 if born in 1921~1925 a2 =1 if age is 25~29 c6 =1 if born in 1926~1930 a3 =1 if age is 30~34 c7 =1 if born in 1931~1935 a4 =1 if age is 35~39 c8 =1 if born in 1936~1940 a5 =1 if age is 40~44 c9 =1 if born in 1941~1945 a6 =1 if age is 45~49 c10 =1 if born in 1946~1950 a7 =1 if age is 50~54 c11 =1 if born in 1951~1955 a8 =1 if age is 55~59 c12 =1 if born in 1956~1960 a9 =1 if age is 60~64 c13 =1 if born in 1961~1965 a10 =1 if age is 65~69 c14 =1 if born in 1966~1970(control group) a11 =1 if age is 70~74 c15 =1 if born in 1971~1975 a12 =1 if age is 75~79 c16 =1 if born in 1976~1980 a13 =1 if age is 80~84 c17 =1 if born in 1981~1985 Ĥ Others Predicted probability of homeownership HEAD =1, if the female decision-maker is a household head MEMBER Number of household member WORK =1, if the female decision-maker is at work EDU Years of education of the female decision-maker EDU_SQUARE Square of the EDU UNMARRIED =1 if the female decision-maker is unmarried MARRIED =1 if the female decision-maker is married DIVORCE =1 if the female decision-maker is divorced WIDOW =1 if the female decision-maker is widowed 14

Table 2 Descriptive Statistics by Female Head and Non-head Households ALL ( n=1,187,987 ) Female Head ( n=364,078 ) Non-female Head ( n=823,909 ) Variables Unit Mean Std. Mean Std. Mean Std. Dependent Variables HOMEOWNERSHIP (0,1) 0.8354 0.3709 0.7979 0.4016 0.8519 0.3552 LIVING SPACE PER PERSON ping 33.5671 19.6207 32.7141 18.8295 33.9441 19.9488 Independent Variables UNMARRIED (0,1) 0.0606 0.2386 0.1100 0.3129 0.0388 0.1930 MARRIED (0,1) 0.7983 0.4013 0.5628 0.4960 0.9024 0.2968 DIVORCE (0,1) 0.0280 0.1650 0.0795 0.2705 0.0053 0.0725 WIDOW (0,1) 0.1131 0.3167 0.2477 0.4317 0.0536 0.2252 WORK (0,1) 0.3708 0.4830 0.4594 0.4983 0.3317 0.4708 MEMBER person 4.3080 2.1917 3.2707 2.0999 4.7663 2.0718 EDU year 7.4041 4.7048 8.0101 4.8362 7.1363 4.6203 * 1 ping = 3.3057 square meter Table 3 Descriptive Statistics by Homeownership Non-homeowner Homeowner ( n=195586) ( n=992401) Variables Unit Mean Std. Dev. Mean Std. Dev. UNMARRIED (0,1) 0.1009 0.3012 0.0527 0.2234 MARRIED (0,1) 0.7619 0.4259 0.8055 0.3958 DIVORCE (0,1) 0.0500 0.2180 0.0237 0.1521 WIDOW (0,1) 0.0872 0.2822 0.1182 0.3228 WORK (0,1) 0.3825 0.4860 0.3685 0.4824 MEMBER person 3.8241 1.9032 4.4033 2.2319 EDU year 8.4170 4.3707 7.2045 4.7425 LIVING SPACE PER PERSON ping 10.1980 10.6224 10.4296 9.6967 HEAD (0,1) 0.3762 0.4844 0.2927 0.4550 Model Estimation A probit model is used to estimate the Equation (1) and a two step procedure is used to estimate Equation (2), which means the estimated probability of homeownership ( Ĥ ) of each 15

household obtained from the Equation (1) is included in the Equation (2) as an independent variable. Before going to the model estimation, two problems have to be clarified. Firstly, in the literature of cohort analysis, the colinearity of age, birth year and census year is a well known problem. Deaton (1997), Fienberg and Mason (1978) have developed different approaches to solve this problem. In this research, we will solve it by constraining the discretion of choosing the control group in the dummy variable set. In the usual case, for a set of dummy variables, any variable in the set can be chosen as the control group to be omitted. Three sets of dummy variables, age, birth year and census year need three omitted variables. However, because age, cohort and census year are mutually dependent, when omitted variables from any two sets of variables, say age and census year, are chosen, then the omitted variable for the third set, i.e. birth year is automatically determined. This means that we lose the discretion of choosing any one of cohort group as the control group. For example, any female in the sample which is at age 21 in 1990 must be born in 1969. In this research, we choose a1 group (age 21-25), y1 group (census year in 1990) and hence c14 group (birth year in 1966-1970) as the omitted variables. Secondly, instead of using semi-panel data which is compiled from group mean of each year-age-cohort combination, the individual household data are used to estimate the model. This will greatly increase the sample size in estimation; especially it can avoid the problem of not enough sample size for the birth cohort in the tail. In this research, due to the restriction of selecting sample between age 15 and 84, eight birth cohorts out of eighteen cohorts were not all observed in three censuses, 1980, 1990, 2000. Four cohorts, i.e. c0, c1, c16, c17 were observed only once; and four cohorts, i.e. c2, c3, c14 and c15 were observed only twice. Discussion of the Estimation Results In order to compare the age-cohort-year effects between the female household head and 16

other female decision-makers, we separate them into two samples. The all sample is also estimated to compare with that of male sample. In the all sample model, weather the female decision-maker is household head or not (HEAD) is included to control for the effect of head on ownership decisions. The estimation results of homeownership rate model will be discussed first, then that of living space per person model. 1. Homeownership rate model The estimation result of Equation (1) is shown in Table 4 4. The census year, age and cohort effect are also drawn into Figures 7. From Table 4 5, we can find that after control for the birth cohort, number of household member (MEMBER) has a negative effect on homeownership rate which is very different than what have found in the literature (Hsueh and Chen, 1999). This is probably because the earlier birth cohorts have larger family size. The positive effect of household size found in the past literature is actually the birth cohort effect. The effect of marriage status shows that married female have higher homeownership rate than other three status, except UNMARRIED in the non-head female households. The WORK has positive effect which indicates that female decision makers who are employed have higher affordability on homeownership. From the all sample, we can find that household head has negative effect on homeownership rate. This finding is conformed with that of other research (Leu, Wang and Wang, 1999) that female heads have a higher probability of poverty. From the all sample model, it shows an increasing year effect in the past twenty years. This means with the growth of economy, the affordability of becoming a home owner is increasing. And there is not much difference with that of male sample. 4 The marginal effects of variables are shown in the Table 4. Because the constant term has no marginal effect, it is not listed. 5 To control for the effect of regional difference, the county/city dummies indicating where the household lives are also included as control variables in the equation; however, they are not reported in the table to save the space. 17

The age effect shows that the ownership rate increases from the young age and reach the highest point in mid-age and decline. This can be explained by the life-cycle theory that means the probability of becoming a home owner is corresponding with her wealth accumulated in the life cycle. The shape of the curve is also similar with that of male. However, for the male sample, it reaches the highest point at age 36-45, but at 50-54 for female. This difference may be due to that we choose the oldest female in the household. The effect of birth cohort shows that the earlier a female decision-maker was born, the higher probability she is a home owner. This can be explained by the fact that space competition has been very strong in Taiwan in the past decades. The price of land and housing in Taiwan has increased faster than the growth of income. This result can also be seen in the male sample. To compare the age-cohort-year effect between heads and non-heads, we can find firstly, the year effect for the female head is steeper than that of non-head. This means the female head households have benefited more from the economic growth. Secondly, for the age effect, we find the home ownership rate of female head households reaches the highest point at about age 30-35, much earlier than that of non-head households. Meanwhile, through out the life-cycle, the ownership rate of female head households is lower than that of non-head. This may due to that at age 30-35, high proportion of female decision-makers got married and become non-head. Those still remain as head are in high proportion single, widowed or a single parent who are economically disadvantageous. Thirdly, it shows a stronger birth cohort effect for the female head households. This means that female head households are affected by the increasing of housing price more seriously than that of non-head households. 18

Table 4 The Estimation Result of Homeownership Rate Model variable All Others Head variable All Others Head Year 1926~1930(c6) *0.0794 *0.0675 *0.1577 1980(y0) *-0.1004 *-0.0961 *-0.1255 1931~1935(c7) *0.0867 *0.0750 *0.1619 1990(y1) - - - 1936~1940(c8) *0.0814 *0.0705 *0.1487 2000(y2) *0.0278 *0.0294 *0.0497 1941~1945(c9) *0.0718 *0.0600 *0.1433 Age 1946~1950(c10) *0.0672 *0.0571 *0.1279 15~19(a0) *0.0482 *0.0448 *-0.0445 1951~1955(c11) *0.0479 *0.0385 *0.1115 20~24(a1) - - - 1956~1960(c12) *0.0307 *0.0262 *0.0697 25~29(a2) *0.0208 *0.0216 0.0087 1961~1965(c13) 0.0048 0.0007 *0.0403 30~34(a3) *0.0310 *0.0287 0.0132 1966~1970(c14) - - - 35~39(a4) *0.0390 *0.0394-0.0143 1971~1975(c15) *-0.0200 *-0.0239 *-0.0243 40~44(a5) *0.0412 *0.0395-0.0117 1976~1980(c16) *-0.0504 *-0.0297 *-0.1185 45~49(a6) *0.0512 *0.0491-0.0239 1981~1985(c17) *-0.0773 *-0.0617 *-0.0697 50~54(a7) *0.0513 *0.0468-0.0174 Marriage 55~59(a8) *0.0486 *0.0460 *-0.0466 MARRIED - - - 60~64(a9) *0.0435 *0.0393 *-0.0495 UNMARRIED *-0.00602 *0.01502 *-0.03466 65~69(a10) *0.0364 *0.0348 *-0.0898 DIVORCE *-0.08019 *-0.07822 *-0.09884 70~74(a11) *0.0264 *0.0290 *-0.1180 WIDOW *-0.00731 *-0.01132-0.00285 75~79(a12) 0.0167 0.0201 *-0.1605 MEMBER *-1.8225 *-1.2256 *-3.5193 80~84(a13) -0.0038 0.0092 *-0.2113 WORK *0.0163 *0.0172 *0.0113 Cohort EDU_SQUARE *0.0006 *0.0007 *0.0006 1896~1900(c0) *0.1214 *0.1032 *0.1840 EDU *-0.0097 *-0.0107 *-0.0074 1901~1905(c1) *0.1187 *0.0997 *0.1860 HEAD *-0.0355 - - 1906~1910(c2) *0.1159 *0.0983 *0.1831 Wald chi-square *67171.4 *46567.0 *17550.3 1911~1915(c3) *0.1061 *0.0898 *0.1809 Pseudo R-square 0.0656 0.0718 0.0479 1916~1920(c4) *0.0920 *0.0777 *0.1712 1921~1925(c5) *0.0825 *0.0687 *0.1676 Note:* means that Coefficients are statistically significant at 5% level. 19

Figure 7 Year, Age and Cohort Effect of Homeownership Decision of Female Decision Makers 20

2. Living space per person model The estimation results of the living space per person model are listed in Table 5, and the coefficients of census years, ages and cohorts are drawn into Figure 8. The coefficients of MEMBER is negative which shows when the household size is bigger, the living space per person is smaller. This is expected because of the scale effect, in the sense that larger households have more members to share public space, e.g. living room, kitchen etc. and resulting in a smaller living space per person. WORK has positive coefficients for all sample and non-head sample. This can be explained by the fact that working female decision-makers have higher income. Ĥ has negative effect on living space; this is different from the descriptive statistics in Table 3, which shows home owners have slightly larger living space per person. This result shows that to own a residence needs to sacrifice living space, other things being equal. This may be due to the cost of owning a residence is higher than that of renting in Taiwan. Although the descriptive statistics show that female head households have smaller living space per person, the coefficient of HEAD in the model of all samples is positive, indicating after control other effects, female heads have larger living space per person. This finding is opposite to the result of ownership rate model, in which we find that female heads have lower homeownership rate. Probably this is due to some unexplained effect from the smaller household size of female head households. Figure 8 shows the year, age and cohort effect. From Figure 8, we can see that the census year effect is increasing. This is the same as that of homeownership rate model. The growth of economy in the recent decades reflects in the larger living space. The age effect for both the female head and non-head is U-shaped with the lowest point at age 37, which is similar to that in the descriptive statistics (Figure 5) to reflect the changes of household size in the life cycle. 21

However, different effects for the head and non-head can still be found. Firstly, the effect is much stronger for the female-head households. Secondly, after age 50 for the non-head households, the effect start to decline; and this phenomenon can not be found in the female-head households. This is probably caused by the coming of the third generation in the family for the non-head households. As for the birth-cohort effect, the non-head households show a declining effect, which means the earlier they were born, the larger living space per person the household have. This can also be explained by the fact that housing price increased faster than that of income over time. However, the effect for the female-head is very different. It shows an increasing effect for those cohorts who were born between 1898 and 1935 and then starting to decline similarly to the non-head households. This may be caused by the disadvantages of the female in the earlier birth cohorts, and the socio-economic situation improved over time for female. 22

Table 5 The Estimation Result of Living Space per Person Model variable All Others Head variable All Others Head Year 1911~1915(c3) *6.2007 *3.1904 0.3001 1980(y0) *-3.2548 *-2.4907 *-0.9884 1916~1920(c4) *5.3501 *2.8515 0.4930 1990(y1) - - - 1921~1925(c5) *5.2783 *2.7579 1.1726 2000(y2) *3.1488 *2.2851 *2.7106 1926~1930(c6) *4.7279 *2.5817 *1.5452 Age 1931~1935(c7) *4.8047 *2.4800 *1.8401 15~19(a0) *0.4106 0.2678-0.6727 1936~1940(c8) *3.9154 *2.0672 *1.5093 20~24(a1) - - - 1941~1945(c9) *3.6170 *1.8737 *1.5409 25~29(a2) 0.0486-0.1580 *-1.3070 1946~1950(c10) *2.6423 *1.3797 *1.0857 30~34(a3) *0.5669 *-0.2362 *-1.5041 1951~1955(c11) *2.3589 *1.1744 *1.3589 35~39(a4) -0.0148 *-0.5746 *-2.1253 1956~1960(c12) *1.1671 *0.6347 *0.6720 40~44(a5) *0.3260 *-0.3918 *-1.7894 1961~1965(c13) *0.7637 *0.3891 0.5819 45~49(a6) 0.1349 *-0.3175 *-1.8489 1966~1970(c14) - - - 50~54(a7) *0.5989 0.0435 *-1.2127 1971~1975(c15) *0.5444 *0.2763 *0.7073 55~59(a8) *0.2424 0.0954-0.9375 1976~1980(c16) 0.0917 *0.3333 0.2591 60~64(a9) 0.3296 0.0937-0.5195 1981~1985(c17) *0.5832 *0.9654 *2.1853 65~69(a10) -0.1665-0.1076-0.2741 MEMBER *-1.8225 *-1.2256 *-3.5193 70~74(a11) -0.0634 *-0.3977 0.3598 Ĥ *-0.1737 *-0.0640-0.0069 75~79(a12) -0.4691 *-0.9726 1.0928 WORK *0.4113 *0.4244 *-0.2244 80~84(a13) -0.1593 *-1.2278 *2.5332 HEAD *2.3175 - - Cohort INTERCEPT *26.4308 *16.6898 *23.7947 1896~1900(c0) *7.3793 *4.5975 *-3.1209 F *3461.63 *3386.69 *1383.65 1901~1905(c1) *7.4924 *4.1866-1.2285 R-square 0.1403 0.1270 0.3292 1906~1910(c2) *6.4224 *3.5045-0.7946 Note:* means that Coefficients are statistically significant at 5% level. 23

Figure 8 Year, Age and Cohort Effect of Living Space per Person Decision of Female Decision Makers 24

Conclusion In this research, we follow the female birth cohort to analyze its effect on housing tenure choice and housing consumption of households. We find that in general the age effect and the birth cohort effect are similar to that of male in both homeownership and living space. The age effects on homeownership and on living space are opposite to each other. People, male or female, in their mid-age have the highest ownership rate and the lowest living space per person. This is due to persons in mid-age has the highest accumulated wealth and also the largest household size in the life-cycle. The birth cohort effect on homeownership shows that the earlier a person, male or female, was born, the higher probability he/she is a home owner. This is due to the price of land and housing has increased faster than the growth of income in the past decades in Taiwan. The birth cohort effect on living space per person is rather different between male and female. For male, it shows almost no cohort effect, but for female, it shows a declining effect, in the sense that the earlier cohorts have larger living space. The age effect and birth cohort effect are very different for female head and female non-head. Although on average, the female heads have longer education years and higher job participation rate, their housing welfare seems to be in disadvantage, which is mainly caused by the fact that female head households have only one earner. Nevertheless, the gap between head and non-head is narrower when the birth cohort is younger. 25

Reference Chang, Hong-chi and Cyrus, C.Y. Chu (1996), Using Pseudo Panel Data to Estimate the Male Cross Period Substitution Elasticity of Male Labor in Taiwan, Taiwan Economic Review, Vol. 24, No. 3, pp. 313-37. (in Chinese) Deaton, Angus (1985), Panel Data from Time Series of Cross Sections, Journal of Econometrics, Vol. 30, pp.109-126. Deaton, Angus S. and Christina H. Paxson (1993), Saving, Growth and Aging in Taiwan, NBER, Working Paper Series, No. 4330. Deaton, Angus (1997), The Analysis Of Household Surveys: A Microeconometric Approach To Development Policy, Baltimore, MD: Published for the World Bank [by] Johns Hopkins University Press, pp.123-127. Fienberg, Stephen E. and William M. Mason (1978), Identification and Estimation of Age-Period-Cohort Models in the Analysis of Discrete Archival Data, in K.F. Schuessler edited, Sociological Methodology, Vol. 8, pp. 1-67, San Francisco: Jossey-Bass. Hsueh, Li-Min and Hsiu-Li Chen (1999), An Analysis of Home Ownership Rate Changes in Taiwan in the 1980s in Asian Economic Journal, Vol. 13, No.4, Dec. (NSC-85-2415-H-170-002) Hsueh, Li-Min and Chin-Lung Yen (2006), The Cohort Analysis of Housing Tenure Choice and Housing Demand in Taiwan, Proceedings of 16 th Annual Conference of Chinese Society of Housing Studies, December, Taipei, Taiwan. (in Chinese) Leu, Chao-hsien, Te-Mu Wang, Shu-Twu Wang (1999), Age, Period, Cohort and the Poverty Rates in Taiwan, Journal of Population Studies, Vol. 20, December, pp. 125-138. (in Chinese) Myers, Dowell (1999), Cohort Longitudinal Estimation of Housing Careers, Housing Studies, Vol. 14, No. 4, July, pp.473-491. Myers, D., Megbolugbe, I & Lee, S.W. (1998), Cohort Estimation of Homeownership Attainment among Native-born and Immigrant Populations, Journal of Housing Research, Vol.9, No. 2, pp.237-269. 26

Thomas F. Crossly and Yuri Ostrovsky (2003), A Synthetic Cohort Analysis of Canadian Housing Career, SEDAP Research Paper, No. 107, August 2003. Verbeek, M. (1992), Pseudo Panel Data, in L. Matyas and P. Sevestre, eds., The Econometrics of Panel Data, Kluwer Academic Publishers, pp. 303-315. 27