Migration Decision and Residential Location Choice: Empirical Models of Science-based Industrial Park in Taiwan Chao-Hong Lu * Yen-Jong Chen ** Abstract Migration decision is one of the important factors that affecting the variation of housing demand. Variables of household s socio-economic characteristics such as job opportunity, wage rates, the working environment, as well as the living preference are significantly affect the decision of migration. A detailed observation recently taken in Taiwan shows that different sites of Science-based Industrial Parks(SIP) located in different cities results in different choices of residential location, and hence generate different growth of population in regional development. More specifically, the migration caused by Hsinchu Science-based Industrial Park (HSIP, in north Taiwan) leads to a significant population growth in Hsinchu city. However, different from the phenomenon expected in HSIP, the migration attracts the employment in Tainan Science-based Park (TSIP, in south Taiwan) but does not increase the population in the city where the science park is located. Instead, the increasing population tend to reside in the nearby large city. In this paper, we constructed a Logit-type location choice model based on the data of Panel Study of Family Dynamics (PSFD), attempting to depict the migration in different urban hierarchies. Independent variables were grouped into three categories :(1)personal characters; (2)economic properties; and(3)variables of regional development. Tentative results show that people with higher degree of education, and with occupation of the commercial and service industry would have higher probability of career migration due to the establishment SIP. Also, households with long-distance migration tend to move into the cities with higher hierarchy. Keyword:Migration, Science-based Industrial Park(SIP), Urban hierarchy, Multinomial logit model, Taiwan. * Ph. D. Student, Department of Urban Planning,National Cheng-kung University, Tainan, Taiwan. p2893105@mail.ncku.edu.tw ** Professor, Department of Urban Planning, National Cheng-kung University, Tainan, Taiwan. yj_chen@mail.ncku.edu.tw
1. Introduction Migration, urban size, and natural population increment are three respective important essentials of urban population. In most developing countries, the migration and urban size are the factors which affect 40 percent of urban growth, and the natural population increment affects 60 percent. In some high-speed growing cities, the migration affects 65 percent of urban growth and even more(liao,1985). Therefore, all the countries always pay much attention to the policy or research of migration. There are no countries which don t devote to balancing regional development based on migration between city and county. Furthermore, migration also results in a huge change of social economy, and redistributes the manpower. As a result, there is interaction between migration and economic development ( Kuznets,1960& Thomas,1964). A simple definition of migration is that an individual or household change the usual dwelling. Migration usually divided into short distance migration and long distance migration in research(stillwell & Congdon,1991), the former is also called mobility which is related to the adjustment of housing demand ( Clark & Dieleman,1996: Xue&Zeng,2000). The latter is caused by occupation factors. Many researches point out long distance migration is related to industrial structure, occupation conditions, employed rate and wage level(harkman,1989;wheaton & Sofrenko,1979). Migration decision is one of the important factors that explain the variation for housing demand. Variables such as job opportunity, wage rates, the working environment, as well as the living quality and the socio-economic characteristics of household all significantly affect the decision of migration. In the end of 20 century, in order to develop high-tech industry, Taiwan government built science-based industrial park in Hsinchu and Tainan one after another. HSIP was located in Hsinchu city of north Taiwan and TSIP was located in Tainan county of south Taiwan(see Figure1). The different urban hierarchy location caused different appearance of migration. HSIP brought population both into Hsinchu city and county. However, TSIP brought population into Tainan city, but not into Tainan county which TSIP was located(see Figure2). Hence, why similar industrial construction located in different urban hierarchy result in different flows of migration? This issue is worth discussing. In this study, we try to quote present database(psfd) and construct migration choice model to examine the phenomena of Taiwan.
Hsinchu Science-based Industrial Park(HSIP) Tainan Science-based Industrial Park(TSIP) Figure1:Locations of HSIP and TSIP social migration rate 12 Social migration rate of Hsinchu and Tainan Hsinchu city Hsinchu county Tainan city Tainan county 10 8 6 4 2 year 0 1990 1992 1994 1996 1998 2000 2002 2004 2006-2 -4-6 -8 Figure 2: Social migration rate of Hsinchu and Tainan Urban Areas Source: http://www.moi.gov.tw/stat/ 2. Brief introduce of HSIP & TSIP Taiwan government started to promote high-tech industry due to the slow development of traditional industry during 1970s. At the same time, the concept of high-tech science park was brought up, and started to attract international capital investment. The purpose of high-tech science park was to push forward the research and innovation of industrial technique in Taiwan, and speeded up the economical growth. The science-based industrial park(sip) was quiet different from other
industrial districts of Taiwan which was managed by economic department. The SIP was conducted by National Science Council(NSC) of Taiwan, including operation affairs and factory service of science park. 2.1 Hsinchu Science-based Industrial Park(HSIP) Hsinchu Science-based Industrial Park built in 1980 was the first one in Taiwan. It was located in Hsinchu city. The major industries in the park were Numbers of the Employment in science park HSIP semiconductor, computer, TSIP PEOPLE communication, pfotonics, etc. Due to the efforts for several years, HSIP had become one of 140,000 120,000 100,000 80,000 the successful science park in the 60,000 world. The science park offered occupation opportunity from 40,000 20,000 8,275 in 1986 to 117,851 in 2006 0 (see Figure 3), and the main education level of employees was senior high to master(see Figure 3: Numbers of the Employment in science park Source: HSIP and TSIP bureau website. Figure 4). YEAR 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2.2 Tainan Science-based Industrial Park(HSIP) Tainan Science-based Industrial Park was built in 1997 because of National Economic Development Program(1993). It was located at Xinshi, Shanhua, and Anding village of Tainan county. The location was 10km far from Tainan city. HSIP people Education of the employment in science park The major industries in the park TSIP were semiconductor and 35,000 TFT-LCD. The park offered 30,000 25,000 occupation opportunity from 20,000 9,489 in 2001 to 47,371 in 2006 15,000 (see Figure 3), and the main education level of employees was also senior high to master(see 10,000 5,000 0 Figure 4). doctor master bachelor training senior others education Figure 4: Education of the employment in science park Source: HSIP and TSIP bureau website.
3. Assumptions and the choice models For observing the migration of differential location of science park. We constructed a Logit-type location choice model based on the data of Panel Study of Family Dynamics (PSFD), also attempting to depict the migration in different urban hierarchies. Our model include two parts, first one is to confer the character of the migrator whether migration, and second is to confer the direction of the migration further more. Independent variables were grouped into three categories :(1)personal characters : male, job position, industrial classes, etc ;(2)economic properties : employment rate, wage level, etc ; and(3)regional development : birth place, migration distance, etc. 3.1 Assumptions and Empirical data Empirical analysis data was from Panel Study of Family Dynamics(PSFD) of Taiwan. The migration is defined as that if someone s birth place 1 was different from the current residence place 2, according to PSFD. The sample, deduct missing data, from PSFD includes 3,110 observations who was born in 1933 to 1976. Under our assumption, there are 1,946 migration and 1,164 non-migration in this data set (see Figure5). Then, we group the migration data into four urban hierarchies 3, include metropolis, satellite, suburban and town. Urban hierarchies represent the observation s present place where located, and also different from birth place. The migration with sample size are metropolis(174), satellite(765), suburban(748), town(259)(see Figure6). Total sample from PSFD (3110) Migration(1946) Present place was different from birth place Non-migration(1164) Present place was the same with birth place Figure5 :migration choice data 1 Birth place : residential place of birth which was discriminated from administrative division. 2 Current place : residential place presently which was discriminated from administrative division. 3 Urban hierarchies : was classified by Directorate-General of Budget, Accounting and Statistics, Executive Yuan, R.O.C(Taiwan).
Migration(1946) Present place was different from birth place Migrate to metropolis Migrate to satellite Migrate to suburban Migrate to town (174) (765) (748) (259) Figure6: flow of migration 3.2 Independent variable The independent variables which we want to survey whether migration and directions in this study were grouped into three categories: personal characters, economic properties, regional development. (1)Personal characters Male: In literature, female would migrate easily than male, due to marriage (Ravenstein,1885:Fields,1982). In this study, we set variable Male as a dummy with value, 1 for male and 0 for female. Education: Many research showed that people with higher education would migrate easily and less limited in distance. It might be due to the reason of receiving wage information from other regions more easily. ( Levy & Wadycki,1974:Fields,1982:Schultz,1982) Job Position: There are more White collars and profession workers migrate in advanced country, and migrators with lower economic ability in developing country(liao,1985). We set this variable as capitalist or professional, white collars worker, small property, worker 4. Industrial Classes: Agriculture would hardly migrate than industry and business (McCormick & Wahba,2003:Xu,1992:Liao,1985). In this study, we set this variable into three categories industries: the first class, the second class, and the third class.. (2)Economic properties Employed increasing rate: If employment rate increasing locally, it would attract people migrating in(hung,1991:wu,1993:jiang,1994). We set two variables here, one is employed increasing rate in the current place and the other is employed rate in birth place. 4 Classifying principles were refer to Goldthorpe and Wu(1997).
Wage level: Regional wage level and migration had a significant relation, and higher wage level attract migration in(schultz,1982:fields,1982:xu,1992 ). Because of many people living in town and working in city, we set two variables in this study. One is the wage level of town and the other is the wage level of city. (3)Regional development : Birth place: The variable is to confer the flow between birth place and living place. We set this variable into four categories which metropolis, satellite, suburban,town respectively. Migration distance: Migration distance over 20km, represents cross-regions migration or long distance migration(xue,2000).it is usually caused by occupation factors. We set this variable into two parts over 20km(long distance migration)or less then. 3.3 The Choice Models Under rational and economic choice, we assume that choosing alternatives caused of its utility. Higher utility brought higher opportunity of chosen. If we put utility into measurable(represent as V )and unmeasuerable(represent as ij ε )part, and further ij more assumed that ε ij appeared independent and identical (TypeⅠextreme value distribution). Then the Multinomial Logit model (MNL)would be stand for chosen probability of alternative(see formula(1)). exp( Vij ) Pij = (1) exp( V ) j ij Measurable utility function shows as formula (2). In which X i represent vectors of personal character and social economic character, β j represent vectors of marginal utility. V = X β (2) ij i j Then, combine formula(1)and(2), it would show the ratio logarithm probability of two alternatives(j v.s. m).(see formula(3)) Pij ln( ) = ( β j βm) X (3) P im In this paper, we define y=1 represent migration, and y=0 represent non-migration. If the utility cumulative distribution function appeared Logistic distribution. It would be
infer the ratio of probability when y=1 and y=0, as formula(4) P y = 1 ln( ) Py = 0 = β 1X (4) In parameter estimate method, we introduce Maximum Likelihood Method to estimate β. In the other hand, we also calculate the odds-ratio 5 of the model to express the results. 4. Empirical results Two models were constructed to explain migration decision and residential location choice between HSIP and TSIP in this study. The first model is Migration Model(see table1), explain the decision of migration or not based on personal characters, economic properties, and regional development. The second model is so called Urban Hierarchy Model(see table2), explaining the flow of migration based on variables of personal characters and regional development. 4.1 The Migration Model (1) Male: As the result, we find male gets lower migration probability than female. It s because that male is still playing an important role in Taiwan family, male is easier migrating out the birth place by the way of marriage. (2) Education: Higher education will increase probability of migration. Because education stands for competition ability, and easily migrate to other place. (3) Job Position: Capitalist and professional have highest migration probability, the next are worker and white collars worker in ModelⅡ. Moreover, ModelⅠ shows small property is harder to migrate, it maybe caused of restricting by it s customer. (4) Industrial Classes: The people belong to second industry and third industry are both easier to migrate, and belong to third industry is easier more migrating than second industry especially. (5) Employed increasing rate: Employed increasing rate of current place has positive relation to migration, and employed increasing rate of birth place has negative relation to migration. (6) Wage level: Wage level of town and city have no significant influence to migration. But wage level of city is slightly higher than town. (7) Urban hierarchy of birth place: metropolis and satellite of birth place have lower migration probability than suburban. Higher urban hierarchy of birth place will lead to lower migration probability. 5 Odds-ratio = e β It represent the ratio of chosen alternative to corresponding alternative when independent variable increasing one unit.
Table1: Migration Model ModelⅠ ModelⅡ ModelⅢ Independent variable parameter Odds- ratio P- value parameter Odds- ratio P- value parameter Oddsratio P- value Constant -3.933* 0.02 0-4.052* 0.02 0-4.064* 0.02 0 Male -1.039* 0.35 0-1.044* 0.35 0-1.028* 0.36 0 Education 0.026* 1.03 0.02 0.032* 1.03 0 0.033* 1.03 0 Job Position capitalist or professional 0.021-0.93 0.623* 1.86 0 white collars worker -0.372-0.13 0.253* 1.29 0.09 small property -0.593* 0.55 0.02 0.051-0.77 worker -0.273-0.21 0.299* 1.35 0.01 Industrial Classes the second classes 0.62* 1.86 0 0.378* 1.46 0 the third classes 0.748* 2.11 0 0.46* 1.58 0 employed rate (current place) 0.332* 1.39 0 0.336* 1.4 0 0.335* 1.4 0 employment rate (birth place) -0.087* 0.92 0-0.087* 0.92 0-0.084* 0.92 0 wage level of town 0.003* 1 0.04 0.003* 1 0.03 0.003* 1 0.03 wage level of city 0.009* 1.01 0 0.009* 1.01 0 0.009* 1.01 0 birth place metropolis -0.484* 0.62 0-0.491* 0.61 0-0.486* 0.62 0 satellite -0.392* 0.68 0-0.402* 0.67 0-0.396* 0.67 0 suburban 0.355* 1.43 0 0.364* 1.44 0 0.36* 1.43 0 sample 3110 3110 3110 L( ˆ α k ) L(0) 2 χ 2 ρ -1739.12-1745.02-1746.3-2056.31-2056.31-2056.31 634.39* 622.57* 620.01* 0.15 0.15 0.15 Predict percent(%) 71.19 71.09 71.9
4.2 The Urban Hierarchy Model (1) Male: Migration into metropolis and satellite have higher probability than town of male. But male is not significant to migrate into suburban. (2) Education: Higher education is easier to migrate into regions of higher urban hierarchy. (3) Job Position: White collars worker have higher probability to migrate into metropolis, the next are capitalist,professional, and worker. But white collars worker have highest probability to migrate into satellite, the next are worker, capitalist, professional. (4) Urban Hierarchy of Birth Place: Higher urban hierarchy of birth place is easier to migrate into higher urban hierarchy region, and so forth as lower urban hierarchy of birth place. (5) Migration Distance: Long distance migration is easier to migrate into satellite and metropolis, and it s not significant to migrate into suburban. Table2: Urban Hierarchy Model Migrate to metropolis Migrate to satellite Migrate to suburban Independent y=1 vs. y=0 y=2 vs. y=0 y=3 vs. y=0 variable Oddsparameter P- parameter Odds- P- parameter Odds- P- ratio value ratio value ratio value Constant -4.46* 0.01 0-1.892* 0.15 0.01-2.705* 0.07 0 Male 0.421* 1.52 0.06 0.507* 1.66 0.02 0.272-0.27 Education 0.149* 1.16 0 0.049* 1.05 0.09 0.116* 1.12 0 Age 0.044* 1.05 0 0.006-0.59 0.022* 1.02 0.06 Job Position capitalist or professional 0.884* 2.42 0.01 1.149* 3.15 0 0.632-0.11 white collars worker 1.315* 3.73 0 1.39* 4.01 0 0.427-0.29 small property 0.604-0.1 0.373-0.32 0.52-0.19 worker 0.553* 1.74 0.02 1.217* 3.38 0 0.736* 2.09 0 Birth Place metropolis 2.167* 8.73 0 1.845* 6.32 0 0.188-0.6 satellite 1.796* 6.03 0 1.769* 5.87 0 0.713* 2.04 0.03 suburban 0.622* 1.86 0.01 0.752* 2.12 0 0.677* 1.97 0 Migration Distance 1.423* 4.15 0 1.556* 4.74 0 0.332-0.13 sample 1946
L( ˆ α k ) L(0) 2 χ 2 ρ -2157.3-2371.9 429.18 0.09 Predict percent(%) 50.77 5. Tentative Conclusions In this study, we employed migration character of Taiwanese to explain residential choice and location of HSIP and TSIP. Conclusions are in the next context. 5.1 Character of migration in Taiwan The empirical analysis shows higher education, job position, and industrial classes will result in higher migration. Female is easier migrating than male by marriage. Higher employed rate and wage level of migrating destination also increases the probability of migration. Besides, people born in metropolis and satellite are less likely migrating than born in suburban and town, because the former were provided with more complete living conditions. 5.2 Urban hierarchy of residential choice To interpret the direction of migration, we divide the odds-ration(migrate to metropolis /Migrate to satellite)(see table3). The results show, higher education and birth place where in metropolis and satellite are easier migrating into metropolis than satellite. And the other variables like male, higher job position, birth place of suburban, and long distance migration are easier migrating into satellite. In addition, from the urban hierarchy of birth place, it also shows migration by stages ( Ravenstein,1985) in Taiwan. Table3: odds-ratio of migration(metropolis to satellite) Migrate to metropolis Independent variable /Migrate to satellite (Odds-ratio) Constant 0.0667 Male 0.9157 Education 1.1048 age - Job Position Capitalist or professional 0.7683
White collars worker 0.9302 Small property - worker 0.5148 Birth Place metropolis 1.3813 satellite 1.0273 suburban 0.8774 Migration Distance 0.8755 5.3 Effect of science park to migration Set up a science park will provide job opportunity in local region and attract workers from other regions. But whether attract residence to migrate in and bring about the aggregating economy, it will depend on its location conditions of science parks. In the migration tendency of HSIP and TSIP, it shows the migration caused by HSIP resulted in a significant population growth in Hsinchu city and county, but the migration caused by TSIP did not increase the population in the satellite where the science park was located. The reasons maybe are: First, the education of science park workers is higher than other regions, so people tend to migrate into metropolis(see table1). Second, the distribution of migration follows migration by stages, so people also tend to migrate into metropolis in the long run. Third, long distance migration shows in TSIP, it may because the population are dispersed in south Taiwan. Fourth, compare HSIP with TSIP, the former has well living function, and the latter doesn t, so the householders are tending to migrate into nearby metropolis which are with well living function. Finally, examine the locations of science-based industrial park, if we want to guide the regional balance and bring about the aggregating economy in local region, besides deliberating the factors of production, also need to examine the local living function and nearby cities which are attracting population. 6 Acknowledgment The authors thank to Mr. Wu, Ying-De of the Department of Urban Planning, National Cheng Kung University, for his data collection and computer works. Reference 1. Chen, C. N. (1991). Migration Selectivity and Its Consequences on the Occupational Structures in the Taipei Metropolis, Journal of Population Studies,
14: 29-58. 2. Fields G. S. (1979). Lifetime Migration in Colombia: Tests of the Expected Income Hypothesis, Population and Development Review, 5(2): 247-265. 3. Goss, E. P. (1985). General Skills, specific skills, and the migration decision, Regional Science Perspective, 15: 17-26. 4. Harris, J. R., M. P. Todaro (1970). Migration, Unemploymant and Development: A Two-Sector Analysis, The American Economic Review, 60(1): 126-142. 5. Hicks, J. (1932). The Theory of Wages, London: Macmillan. 6. John Stillwell & Peter Congdon(1991). Migration Models: Macro and Micro Approaches. London and New York: Belhaven Press. 7. Kuznets, S., R., Miller, A., Easterlin (1960) Population Redistribution and Economic Growth- United States, 1870-1950, Vol. 2. 8. Levy, M., W. Wadycki (1974). Education and the decision to migrate: an econometric analysis of migration in Venezuela, Econometrica, 42: 358-377. 9. Li-Min Hsueh, Hsi-Peng Tseng, Yea-Jiun Chen(2002). Locational Choice in Cross-regional Migration: An Empirical Study of the Taipei Metropolitan Area Journal of Housing Studies, Vol. 11. 10. Maddala, G. S. (1989). Limited Dependent and Qualitative Variables in Econometrics, Cambridge University Press, Cambridge, Massachusetts, 22-46. 11. Nivalainen S. (2004). Determinants of family migration: short moves vs. long moves, Journal of Population Economics, 17: 157 175. 12. Ravenstein, E. G. (1885). The Laws of Migration, Journal of the Royal Statistical Society, June: 167-227. 13. Schultz, T. W. (1962). Reflection on Investment in Man, Journal of Political Economy, 70(5): 1-9. 14. Schultz, T. P. (1982). Lifetime Migration within Educational Strata in Venezuela: Estimates of a Logistic Model, Economic Development and Cultural Change, 30(3): 559-93. 15. Sjaastad, L. (1962). The costs and returns in human migration, Journal of Political Economy, Supplement, 70, 80-93. 16. Stouffer, S. A. (1940). Intervening Opportunities: A Theory Relating Mobility and Distance, American Sociological Review, 5, 845-867. 17. Stouffer, S. A. (1960). Intervening Opportunities and Competing Migrants, Journal of Regional Science, 2, 1-26.