Impact of Highway Construction on Internal Migration: A Korea Perspective

Abstract

Transportation facilities in the classical migration model expand the scale of population movement by reducing the physical distance and costs of travel between regions. This study empirically analyzes the effect of population inflow into a region near a highway construction using a migration model. The purpose was to examine whether large-scale transportation infrastructure such as highways contributes to the growth and balanced development of the connected regions. Four metropolitan transport-type and commuter transport-type highway routes constructed in South Korea after 2000 were used as study subjects. The results show that population inflow occurred in the region where commuter transport-type highways were constructed. However, no effect was observed in regions where metropolitan transport-type highways were constructed. The reason for these unique results is that factors affecting migration include not only transit facilities, but also parameters such as economic factors, housing, and amenities. Moreover, the findings show that the impact of population inflow is affected by the geographical characteristics of the highway construction areas and city size.

1. Introduction

Rapid industrialization in Korea during the 1970s triggered internal migration, a phenomenon known as “rural–urban migration”, in which a large number of people moved from rural areas to cities and from non-metropolitan to metropolitan areas in search of job opportunities paying higher wages, similar to Western society. Against this backdrop, the pace of development differed among regions as the central government’s economic development strategy led to the development of industries and the expansion of social overhead capital (SOC) facilities centered in specific regions. In particular, the issue of territorial imbalance, in which both population and capital are concentrated around large cities and adjacent areas, such as the capital city and metropolitan areas, had a negative impact on the country’s overall economic development. Population decrease in a specific area leads to a host of complex problems, such as a decrease in the working-age population in the region, urban finance issues, weakening social and cultural conditions, delays in the expansion of social infrastructure in the region, and urban decline. Therefore, it is necessary to deal with this population imbalance at the national level and handle the construction of large-scale transport infrastructure in terms of balanced national development [1].

Roads are important in the new network city development model because they connect cities. They improve spatial accessibility, reduce costs, improve productivity, maximize profit, and activate human and material resource exchanges. Some assert that such developments may cause straw effects, referring to the mobility of people and resources from regions of low to high populations and economic activity, or the subordination of small cities to larger cities. Others differentiate economic infrastructure (Economic SOC) and social infrastructure (Social SOC), or point-type infrastructure and network-type infrastructure. Network-type infrastructure, such as transportation infrastructure, is believed to moderate interregional disparities [2].

The national policy for balanced development as a long-term core governmental initiative includes diverse governmental projects and budgets, and aims to alleviate differences between metropolitan and non-metropolitan regions. The “National 5-Year Balanced Development Plan” was established for a systematic and sustainable promotion of this policy. Currently, the 4th Plan (2018–2022) is in progress. A USD 80 billion budget (KRW 112 billion) was originally planned for the 4th 5-year National Balanced Development Plan, among which the budget for metropolitan transportation, including roads and railroads, was USD 11 billion (KRW 15 billion) (13.7% of the total budget) [3]. The South Korean government massively invests in moderating regional disparities while ensuring fair spatial access to markets by expanding transportation infrastructure [4].

Since the 2000s, in the construction of transportation infrastructure such as highways, railroads, and airports, which are social overhead capital facilities, demand and economic feasibility as well as contribution to balanced development became significant factors. The West Coast Highway connected in 2001 was constructed to promote growth, reinforce vulnerable industrial structures, and improve productivity and underdeveloped infrastructure in the west coast region. Highways constructed after the 2000s are classified into long- and short-range commuting and traffic roads. Further, roads constructed in the 2000s, apart from highways, are divided into metropolitan transportation, in which the route connects metropolitan and non-metropolitan areas to improve underdeveloped regions, and short-scale commuter traffic-type transportation for commuting and tourism near metropolitan areas.

This study aims to examine the impact of South Korea’s highway construction on migration, and the contribution of large-scale transportation infrastructure to the balanced growth of developed regions. For this purpose, the impact of the four major highways constructed after 2000 was empirically analyzed based on interregional migration. The empirical analysis was conducted using a panel regression model based on national statistical data. The target roads for this study were two metropolitan transportation sections (west coast and central inland) and two tourism transportation sections (Seoul–Chuncheon and secondary center). The characteristics of migration according to the purpose of highway construction were examined through road type-dependent analysis.

2. Theoretical Considerations and Literature Review

2.1. Theoretical Considerations

A typical theory that transportation facilities, such as highways, have a direct impact on internal migration is the gravity model in the macroscopic migration model. The macroscopic migration model proposes that the population moves from a low-income region to a high-income one due to regional income disparities, and physical distance acts as the limiting factor. In other words, this model focuses on the distance that restricts internal migration and regional economic disparity, which is a key factor inducing migration.

In the gravity model, population size and distance play a crucial role in migration flows. The size of migration is proportional to the product of the population in two regions, and inversely proportional to the distance between the two regions. That is, in the gravity model, the migration distance is regarded as an obstacle to migration, and internal migration is claimed to minimize the migration efforts. Here, the population of the origin area represents the size of migration, and the population of the destination represents the level of available employment opportunities. Moreover, physical distance refers to travel cost. As the migration distance increases, the cost (direct cost) required for migration increases, and the existing social relationship (psychological cost) weakens. A longer distance weakens the migration motive due to uncertain information about available opportunities. This is attributable to the difference in information acquisition depending on the level of human capital; accordingly, there is a difference in migration distance between individuals. The distance parameter (b) refers to the influence of distance on migration. A higher distance parameter undermines migration, and a lower distance parameter accelerates migration. However, the distance parameter is not the same across regions, as it varies depending on spatial characteristics such as means of transport, highways, and topography. For example, as the economy develops, the transportation network improves, and this reduces the friction coefficient for distance, and it varies across regions.

$${\mathrm{M}}_{\mathrm{ij}}=\mathrm{k}\times \frac{{\mathrm{P}}_{\mathrm{i}}\times {\mathrm{P}}_{\mathrm{j}}}{{\mathrm{D}}_{\mathrm{ij}}{}^{\mathrm{b}}}$$

(1)

• M_ij: Migration flow inensity between region i and j;
• P: Population in region i or j;
• D_ij: Physical (road,railway,etc.) distance;
• k: Constant;
• b: Distance parameter

The modified gravity model (Lowry, 1966) explains that the benefits accrued from regional economic disparities are the underlying cause of migration, and specifically overcomes the limitations of the classical gravity model, which explains that the regional characteristics inducing migration are only population size [5]. Due to regional wage differences and employment opportunity gaps, including wages and employment rates, aside from the labor force size between the two regions, it is assumed that migration flows occur from the region where wages are low and unemployment rates are high (oversupply of labor) to the region where wages are high and unemployment rates are lower (labor shortage). In this case, distance between the two regions also serves as a key variable, and physical distance changes according to the available transport, such as roads and railways.

$${\mathrm{M}}_{\mathrm{ij}}=\mathrm{K}\times \frac{{\mathrm{U}}_{\mathrm{i}}}{{\mathrm{U}}_{\mathrm{j}}}\times \frac{{\mathrm{W}}_{\mathrm{i}}}{{\mathrm{W}}_{\mathrm{j}}}\times {\mathrm{L}}_{\mathrm{i}}\times {\mathrm{L}}_{\mathrm{j}}\times \frac{1}{{\mathrm{D}}_{\mathrm{ij}}{}^{\mathrm{b}}}$$

(2)

• M_ij: Migration from region i to region j;
• U_i, U_j: Unemployment rate in region i and region j;
• W_i, W_j: Manufacturing wage per hour in region i and region j;
• L_i, L_j: Labor force in non-agricultural sector per hour in region i and region j;
• D_ij^b: Distance between region i and region j

However, the gravity model does not explain why some people migrate and others continue to live in the same social conditions. In other words, the model has a limitation in that it excludes the personal motives, circumstances, and social determinants of migration.

In a typical example of the neoclassical migration model used to overcome this limitation, Todaro (1971, 1976) views the expected wage (or expected income) that can be earned in a city as the underlying cause of migration [6,7]. People who move from rural to urban areas are employed in the city’s informal economy and earn less than in their rural heartland, but over time they get the opportunity to work in the city’s formal sector. In other words, notwithstanding the high unemployment rate, migrants move from rural to urban areas because the income increase effect of migrants includes the expected wages for employment in the formal sector.

However, Todaro’s expected income hypothesis does not explain why in developing countries, most migrants have low levels of education, but they have permanent jobs rather than temporary jobs in the informal sectors of the city. Accordingly, a modified model, including the amenities variable, has been presented through the hypothesis that individuals decide to move as they are affected by not only regional wage disparities, but also differences in the level of local public amenities.

The local amenities theory (Tiebout, 1956) is based on the premise that public goods are distinguished from private goods [8]. Unlike private goods, public goods are not completely consumed, and have the characteristic of being able to offer services without purchase. Tiebout stated that public goods provided by local governments, such as local schools, hospitals, parks, and welfare facilities, are a key driving migration. In other words, local public goods are differentiated according to the financial capabilities of local governments, the benefits of public goods and services are different, and local taxes, which are costs, also vary among regions. Here, people migrate to jurisdictions where benefits from local public goods are maximized.

This study analyzes the various factors affecting migration using a migration model that reflects the Todaro model and the Tiebout hypothesis based on the theoretical concept of the gravity model, which proposes that internal migration is affected by the creation of transportation networks, such as highways.

2.2. Literature Review

Previous studies have been divided into empirical research investigating the impact of social overhead capital facilities on improving transportation accessibility, such as highways on regional economies, and research on social and economic amenities that impact migration.

First, various studies examine the impact of investment in transportation facilities, such as highways or local roads, on the local population, economy, and industry. Some studies show that investment in transportation facilities had a positive effect, while others claim that it had no effect or a negative outcome [9].

Kim (2002) estimated the ripple effect in the logistics, tourism, and trade sectors based on statistical data and current status data between areas involving highway construction and major cities instead of a quantitative analysis model to examine the industrial ripple effect in Daejeon and Chungnam as the outcome of highway construction [10]. The analysis results show that the opening of the Seohaean Expressway reduced logistical costs due to shortened distance, the improvement of traffic congestion on highways including the Gyeongbu Expressway, the development of large-scale industrial complexes in the Seohaean area, accessibility to Incheon International Airport, and the development of the Seohaean tourism complex.

Shim and Yoon (2002) analyzed the outcome of highway construction in Korea, specifically production increases in all manufacturing businesses and 36 manufacturing sectors using the growth accounting method, or total factor productivity analysis [11]. This method uses a production function that includes the input of intermediate goods, an aspect ignored in the Solow model. Using production functions such as labor and capital input—excluding capital accumulation in highway construction—intermediate input, capital investment in highway construction, and technological progress, the impact of highway construction investment from 1968 to 1996 was analyzed based on the increase in manufacturing productivity and output. The results show that intermediate goods had the highest impact on the increase in manufacturing output, followed by capital, labor, total factor productivity, and highways. Moreover, the impact of highways on manufacturing output was not large, and there was no effect in the present compared to the past. The contribution rate of highways accounted for only 0.0002% of the overall productivity increase in manufacturing, while it was only 0.0004% in 1968–1980, and declined to 0% in 1981–1993.

Jeon and Kang (2007) analyzed the economic effect of the Seohaean Expressway on the development of the local economy and regional growth in terms of population and employment using a multiple regression model, that is, dividing employment changes by industry into construction and non-construction areas between 2001 and 2005 [12]. The dependent variables were population change rate, changes in manufacturing business workers, and changes in service workers, and the explanatory variables were net migration rate, number of workers in the manufacturing business (log value), change rates of workers in the manufacturing business, number of workers in the service sector (log value), change rates of workers in the service business, and dummy variables (Seoul, Gyeonggi, Gangwon, Chungcheong, Jeolla, Gyeongsang, and the Seohaean Expressway construction area). The key analysis result shows that the Seohaean Expressway contributed to the increase in employment in the service sector, and population in the area with the highway exit decreased.

Hong and Kim (2008) analyzed the impact of the Seohaean Expressway by comparatively analyzing changes across 13 cities and counties along the Seohaean Expressway construction area, and 13 cities and counties in the comparison group, focusing on changes in population and industrial structure [13]. Cities and counties with a small mean difference were selected as the comparison group through variance analysis using data on the rate of population change, the number of manufacturing establishments, the rate of change in the number of workers, and the number of accommodation and food service companies, (1999–2005). The findings suggest that the impact of opening a highway on the surrounding areas is not shown consistently according to region and sector, but emerges differentially, according to the region’s strategic geopolitical attributes and regional development policy [13,14].

Ahn and Han, Kim (2008) empirically analyzed if there is any difference in the supply or investment of road and transportation facilities based on regional characteristics [15]. For the analysis, 160 cities and counties were divided into regions with low, medium, and high GRDP per capita, and regression analysis was conducted with a fixed effects model using the Cobb–Douglas production function. Similarly, GRDP per capita and population were used as the dependent variables. For explanatory variables, those related to transportation facilities (road ratio, national land coefficient) were set as the proxy variables for estimating the stock of public capital. The number of local employees was considered as the variable representing labor, and the number of highly educated workers and the ratio of elderly population were set as the variables for human capital. Dummy variables were used for each region, such as metropolitan area, Gangwon, Chungcheong, Gyeongsang, and Jeolla. Based on the findings, it was concluded that concentrating investment in cities and counties with relatively low road ratios or national land coefficients, and medium GRDP per capita, increases investment efficiency. In the group with high GRDP per capita, both proxy variables showed negative (−) values for road transportation facilities and GRDP, and groups with medium and low GRDP per capita did not show a significant value in the road ratio.

Kim (2009) used the consumer utility function and demand function to theoretically explain the “straw effect” of highways [16]. First, through the indifference curve, it is suggested that consumption in external regions increases if an individual’s indifference curve tends to prefer external goods and high-priced goods. Second, through the demand function, it was determined that the occurrence of the straw effect in adjacent and remote areas differed depending on product characteristics. Arguably, when new transport facilities, such as highways and high-speed railways, are opened, they have both a positive and negative impact on the local economy, and consumption growth varies depending on product characteristics.

Examining the Daejeon–Tongyeong Expressway, Choi, Jeong, and Kim (2011) analyzed how the opening of a highway has affected the surrounding areas [17]. The analysis was conducted through a t-test on changes between key indicators such as rate of population change, net population increase, factory site increase/decrease rate, farmland increase/decrease rate, number of workers in the manufacturing business, wholesale and retail business, transportation business, and accommodation and food service business. The findings suggest that the impact of highway construction does not appear consistently, but differs slightly depending on regional characteristics. In other words, although the highway has improved accessibility, it does not have a positive effect across all sectors.

Choi and Kim, Yoon (2014) examined the impact of highway construction between Seoul and Chuncheon on local industries in 114 eup, myeon, and dong (76 city areas and 38 county areas) including cities and counties along the Seoul–Chuncheon Expressway and neighboring areas using a multiple regression model based on data during the period 2000–2010 [18]. The dependent variables are the number of workers and number of establishments, and the independent variables are population, area of administrative districts, distance from IC (interchange) to the center of eup, myeon, and dong, and whether or not IC is included. The regression analysis showed that the highway construction between Seoul and Chuncheon fell within 10 km from the nearest IC, and the size of impact varied among industries. For example, the number of employees in the “construction” sector had the greatest impact on highway access during the construction period, but an insignificant impact on “manufacturing”, “accommodation and food service”, and the “wholesale and retail business”.

Major overseas studies argue that highway or road construction has a positive effect on the surrounding area’s population and economy, while others argue that it has a differential effect or a negative effect on the construction areas.

Carino and Voith (1992) analyzed if there was any difference in the productivity change of the manufacturing business for public investment, consistent with the characteristics of the industrial structure, human capital, and infrastructure by state. According to the analysis results, highway construction has a positive effect on the increase in local employees and productivity [19].

Rephann and Isserman (1994) stated that investment in large-scale infrastructure such as highways had a differential ripple effect depending on the distance from the metropolitan area [20]. Highway construction increases income across industries such as manufacturing, retail, and distribution, specifically in areas that are adjacent to large cities. It is found that income in the retail industry increases in small and medium-sized cities that are far from large cities and with a population of 25,000 or more. In addition, highway construction did not show any positive effect on small towns and rural areas separated from large cities.

Chandra and Thomson (2000) analyzed the impact of highway construction in the United States using a spatial competition model [21]. The findings showed that some industries along the highway construction areas expanded due to reduced logistics costs, but industrial sectors deteriorated due to improved traffic conditions.

Bunea (2012) analyzed internal migration in Romania using an improved gravity model that reflected economic and social factors. For the analysis, a fixed effects model and a GMM (Generalized Method of Moments) model based on panel data were used [22].

He supplemented the study of Philbrick (1973) [23], which interpreted transportation infrastructure and population movement as a gravity model. The analysis results show that population size, GRDP per capita, road density, amenity index, and crime rate were affected. In particular, if road density increases by 1%, the migration rate increases by 0.44%. In other words, it can be concluded that an effective transportation system facilitates migration.

Poot et al. (2016) reviewed the gravity model of migration to help in understanding population projections by analyzing the stability of the distance parameter using the case of New Zealand [24]. The analysis of New Zealand’s domestic migration showed that the transport system, technology, and highways did not increase migration. This indicates that other factors, such as changes in commuting patterns, could induce migration.

The following studies were conducted on regional social and economic amenities for understanding the cause of migration.

Kim and Yang (2013) distinguished large-scale, mid-scale, and non-city regions to examine the impact of public infrastructure on population increase [25]. The increase in businesses positively impacted the migration into regions of all scales, while cultural services negatively impacted mid-scale cities, and housing prices had a positive impact only in non-cities. Other variables were not statistically significant. Hong and Yoo (2012) analyzed the factors of migration by age group [26]. Regional migration is inclined towards regions of higher expected income, lower population density, higher terrain, and lower social welfare budget. In migration between metropolitan and non-metropolitan regions, the young generation migrated towards regions with higher expected income and rates of marriage. While regions with lower population densities were preferred within the metropolitan region, regions with high population densities were preferred in migration between metropolitan and non-metropolitan regions. This implies that while distinct cultural opportunities are an important factor in interregional migration, movement within metropolitan areas tends towards regions of lower chaos as interregional cultural opportunities are similar, and interregional interactions are available.

Kim and Seo (2014) distinguished metropolitan and non-metropolitan regions and selected specific variables that significantly impact population change [27]. In metropolitan regions, the number of newly born infants or economic factors affected the population change. Non-metropolitan regions were significantly affected by the number of houses, local taxes, parking area, and number of doctors, in addition to economic factors.

Kwon (2005) examined the migration characteristics between metropolitan and non-metropolitan regions through personal and regional factors [28]. The common factors influencing migration from metropolitan to non-metropolitan regions were age(−), education (+), marital status, and new address in rural areas (+). For the rate of manufacturing industry, the coefficient was negative (−) for the metropolitan area between 1995 and 2000 and positive (+) for the non-metropolitan area. This implies that the rate of manufacturing industry was insignificant for migrations into metropolitan regions.

Previous research shows that improving transportation accessibility through establishing social overhead capital facilities, including highways, positively impacts regional economies, population increase, and underdeveloped regions. However, in regional economies and populations, the rich may become richer and the poor may become poorer due to lock-in.

The following was hypothesized in response to the research question of the impact of highway type and regional characteristics on changes in regional populations. First, metropolitan transportation highways connect regions with significantly different economic scales, and the probability of the occurrence of the straw effect is high, as shown in the gravity model. Second, commuter-type highways increase accessibility between regions of similar economic scale and allow short-distance commuting. Therefore, movements to the suburbs where housing prices are lower and a population increase in regions where highways are constructed are expected. To validate these hypotheses, we empirically analyzed two of each of the two highway types.

3. Analytical Model and Results

3.1. Research Method and Analysis Model

The results of previous studies establish that socioeconomic factors, environmental issues, and transportation facilities reduce the physical distance between regions. This study examined if the impact of highway construction accelerated migration into the construction area due to improved transportation facilities and accessibility. This study analyzed four representative highway sections among the 51 highway routes in Korea, as of 2019: Seohaean Expressway, Jungbu Naeryuk Expressway, Seoul–Chuncheon Expressway, and Second Jungbu Expressway(

Table 1. Opening year and highway regions analyzed in this study.

Expressway	Opening Year	Road Length	Region
Seohaean Expressway	2001	336.1 km	(Seoul) Geumcheon-gu, (Gyeonggi) Siheung-si, Ansan-si, Pyeongtaek-si, Hwaseong-si, Gunpo-si, Gwangmyeong-si, Anyang-si, (Chungnam) Dangjin-si, Seosan-si, Hongseong-gun, Boryeong-si, Seocheon-gun, (Jeonbuk) Gunsan-si, Gimje-si, Buan-gun, Gochang-gun, (Jeonnam) Hampyeong-gun, Yeonggwang-gun, Muan-gun, Mokpo-si
Jungbu Naeryuk Expressway	2012	302.3 km	(Daegu) Dalseong-gun, (Gyeonggi) Yeoju-si, (Chungbuk) Chungju-si, Goesan-gun, Eumseong-gun, (Gyeongbuk) Gimcheon-si, Gumi-si, Sangju-si, Mungyeong-si, Seongju-gun, Goryeong-gun, (Gyeongnam) Changwon-si, Haman-gun, Changnyeong-gun
Seoul–Chuncheon Expressway	2009	150.2 km	(Seoul) Gangdong-gu, (Gyeonggi) Namyangju-si, Hanam-si, Gapyeong-gun, Yangpyeong-gun, (Gangwon) Chuncheon-si, Hongcheon-gun
Second Jungbu Expressway	2001	31.1 km	(Gyeonggi) Icheon-si, Gwangju-si, Hanam-si

Source: Among the opening sections of the Seoul–Chuncheon Expressway, Inje-gun and Yangyang-gun were excluded from the analysis as they were opened in 2017.

). The highways selected were completed after 2000 for the ease of obtaining local data, avoiding the redundancy of route locations and reflecting the diversity of the construction areas and route length. The Gyeongbu Expressway, the most representative expressway in Korea that was completed in 1970, was excluded from the analysis because it was difficult to determine the exact outcome due to limited regional data (Figure 1).

The temporal scope of this study is from 2006 to 2018, and the spatial scope includes 229 cities/counties/districts. Statistical data from Yeongi-gun, Chungnam were used prior to the construction of Sejong City. For our analysis, we accessed data on internal migration statistics, financial independence, GRDP, number of establishments, and number of workers in establishments from Statistics Korea’s Korean Statistical Information Service and Census on Establishments.

This study aims to identify the dynamic characteristics of internal migration as the outcome of highway construction. We constructed a theoretical model by combining the Todaro model, found in the private economy sector, and the Tiebout model, found in the public sector, for factors of internal migration, and empirically analyzed the factors affecting internal migration through panel data regression analysis.

Panel data analysis is classified into pooled ordinary least squares (OLS) regression, the fixed effects model, and the random effects model. Pooled OLS is estimated by ignoring the fact that panel data have a panel structure. The fixed effects model assumes that the error term of individual characteristics of the panel is fixed, whereas the random effects model assumes that the error term of individual characteristics of the panel is a random variable.

The model of pooled OLS ignoring the panel structure and estimating with OLS is as follows. ${\mathrm{y}}_{\mathrm{i},\mathrm{t}}$ is the net migration rate for region i, which is the dependent variable, ${\mathrm{X}}_{\mathrm{i},\mathrm{t}}$ is the vector of explanatory variables (i.e., GRDP increase rate, employee increase rate, metropolitan area dummy), ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$ is the error term, and $\mathrm{a}$ is the intercept. However, if OLS is applied without considering the characteristics of panel data, it cannot be deemed an unbiased estimator, as it does not have the characteristics of a consistent estimator, even if the number of observations increases.

$$\begin{array}{l}{\mathrm{y}}_{\mathrm{i},\mathrm{t}}={\mathrm{a}}_{\mathrm{i}}+{\mathsf{\beta }}_{}{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}\\ \mathrm{i}=1,2,\dots ,\mathrm{n}\left(16 \mathrm{cities} \mathrm{and} \mathrm{provinces}\right) \mathrm{and} \mathrm{t}=1,2,\dots ,\mathrm{T}\left(\mathrm{year}\right)\\ \mathrm{Yet},{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}={\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\nu }}_{\mathrm{i},\mathrm{t}}\\ {\mathsf{\mu }}_{\mathrm{i}}=\mathrm{unobserved} \mathrm{regional} \mathrm{characteristic} \mathrm{effect}\\ {\mathsf{\lambda }}_{\mathrm{t}}=\mathrm{unobserved} \mathrm{time} \mathrm{effect}, {\mathsf{\nu }}_{\mathrm{i},\mathrm{t}}=\mathrm{stochastic} \mathrm{random} \mathrm{term}\end{array}$$

(3)

In the fixed effects model, ${\mathrm{u}}_{\mathrm{i}}$ refers to the constant term difference according to the characteristics of each panel group, but it is not recognized as a random variable; instead it is considered a specific fixed value. In other words, as it can be viewed as a model without an error term, the consistency of the estimation coefficient can be guaranteed even if there is a correlation between the error term and the explanatory variable, thereby resolving the endogeneity problem. However, as the metropolitan area dummy is included in this study, a general fixed effects model is not suitable, and the least squares dummy variable (LSDV) model, a fixed effects type of model, was used. In this method, the OLS estimation method is applied after treating it as a dummy variable with the characteristics of the panel group.

$${\mathrm{y}}_{\mathrm{i},\mathrm{t}}=\mathrm{a}+{\mathsf{\beta }\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(4)

The random effects model assumes that regional difference is a random variable and analyzes regional difference as part of the error term. Therefore, it is assumed that the intrinsic effect of each region is not related to the explanatory variable. ${\mathrm{v}}_{\mathrm{i},\mathrm{t}}$ is a random variable, and $\mathrm{cov}\left({\mathrm{X}}_{\mathrm{i},\mathrm{t}},{\mathrm{v}}_{\mathrm{i}}\right)=0$ is assumed, as it is deemed more suitable than the fixed effects model.

$$\begin{array}{l}{\mathrm{y}}_{\mathrm{i},\mathrm{t}}={\mathrm{a}}_{\mathrm{i}}+{\mathsf{\beta }\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}\\ \mathrm{However},{\mathrm{v}}_{\mathrm{i},\mathrm{t}}={\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}\end{array}$$

(5)

3.2. Variable Selection and Data Analysis

The proposed internal migration model in this study explains that the underlying causes of migration are regional differences in economic, cultural, and social factors. Through this, we intend to identify the internal migration factors of highway construction and non-construction areas and derive the net migration rate of the construction area after the highway is constructed.

The method for measuring the flow of internal migration is divided into the number of migrants and the net migration rate. For the number of migrants, the underlying advantage is focusing on analyzing the impact of the destination characteristics on the population inflow; however, there may be huge regional disparities depending on the region’s population size. Conversely, the net migration rate has the advantage that internal migration differs depending on the combination of origin and destination and reflects asymmetric characteristics, but as it divides the total population of the region, the meaning of the migrant size can be reduced by the region’s population size. This study selected net migration rate as the dependent variable because it is convenient to compare it with the results of previous studies and to reflect the asymmetry due to regional size.

The explanatory variables comprised economic factors and local amenities that affect internal migration, based on the analytical model of previous studies related to migration, and the highway construction section was processed as a dummy variable(

Table 2. Key variable definition.

Division	Variable	Description	Unit
Dependent variable	move	(Number of in-migrants—number of out-migrants in each city, county or district)/resident registration central population	%
Explanatory variable	grdp	GRDP per capita	KRW 1 million
	r.grdp	Year-on-year increase in GRDP per capita	%
	r.firm	Year-on-year increase in total number of establishments	%
	r.labor	Year-on-year increase in total number of workers	%
	finc	(Local tax + non-tax receipt) × 100/general accounts budget size	%
	cult	Number of cultural facilities per 100,000 population	Number of facilities
	welf	Number of social welfare facilities per 100,000 population	Number of facilities
	med	Number of beds in medical institutions per 1000 population	Number of beds
	seohae_dummy	Seohaean Expressway construction area = 1, non-construction area = 0	-
	jungbu_dummy	Jungbu Naeryuk Expressway construction area = 1, non-construction area = 0	-
	seochun_dummy	Seoul–Chuncheon Expressway construction area = 1, non-construction area = 0	-
	jungbu2_dummy	Second Jungbu Expressway construction area = 1, non-construction area = 0	-

Note: Reflecting the year of opening; Source: Statistics Korea, Korean Statistical Information Service (http://kosis.kr/index/index.do (20 November 2021)), each year [30].

). Economic factors comprised GRDP per capita, rate of increase in GRDP per capita, rate of increase in establishments, and rate of increase in jobs. Additionally, GRDP per capita is an indicator that represents regional production level or income level, and was used for reflecting the disparity in regional income levels. The rate of increase in GRDP per capita is an indicator for denoting the variation speed of companies in the region, reflecting the local economic development activity level, and the rate of increase in jobs is a leading indicator for job variation speed in the region.

Factors influencing local amenities comprised financial independence, the number of cultural and social welfare facilities per 100,000 population, and the number of beds in medical institutions per 1000 population. Financial independence is an indicator for projecting a region’s level of economic autonomy. If the region has sound financial conditions, the level of service and life can be improved through the benefits of public goods. The number of cultural and social welfare facilities per 100,000 population and number of beds in medical institutions per 1000 population are indicators that are used to measure the cultural, social and welfare service levels, and medical and health care, in each region.

For the highway construction sections, we examined the causal relationship between the construction area and the net migration rate by designating the construction section as 1 (even if it is a construction area, the point when the highway was not open is 0) and the unopened area as 0 (the point of opening in each construction area (city, county, district)) of the four highways.

Basic statistics before log transforming the variables constructed in this analysis are shown in

Table 3. Basic statistics.

Variable	Observed Value	Mean	Standard Deviation	Minimum Value	Maximum Value
move	2977	−0.03	2.18	−11.58	25.15
grdp	2977	29.41	30.55	5.23	431.69
r.grdp	2977	5.57	9.20	−41.86	73.13
r.firm	2977	1.90	2.95	−8.43	26.04
r.labor	2977	2.96	4.17	−19.80	56.63
finc	2977	29.20	16.69	6.40	94.30
cult	2977	8.29	8.11	0.40	74.90
welf	2977	14.63	10.69	0.00	72.60
med	2977	12.56	8.23	0.00	70.10
seohae_dummy	2977	0.09	0.29	0.00	1.00
jungbu_dummy	2977	0.06	0.24	0.00	1.00
seochun_dummy	2977	0.03	0.17	0.00	1.00
jungbu2_dummy	2977	0.01	0.11	0.00	1.00

. The mean net migration rate is −0.03%, the minimum value is −11.58%, and the maximum value is 25.15%. The net migration rate between regions at the national level is 0% because population inflow and outflow remain the same; however, it indicates the mean of 229 cities, counties, and districts. The mean GRDP per capita and average annual increase were KRW 29.41 million and 5.57%, and the maximum and minimum values of each variable have huge differences, thereby indicating significant regional economic disparities. In the case of the rate of increase in establishments and the rate of increase in workers, there was a huge gap between regions with negative growth and those with positive growth. The mean of financial independence was 29.2%, with 94.3% at the highest and 6.4% at the lowest, thereby indicating wide regional disparities in financial levels. For cultural and social welfare facilities and number of beds in medical institutions, the minimum value is sometimes close to 0, indicating that some regions do not have cultural, social, and medical infrastructure.

The correlations between the variables constructed in this analysis are as follows (

Table 4. Correlation analysis by variable.

Division	move	grdp	r.grdp	r.firm	r.labor	finc	cult	welf	med
move	1.000
grdp	0.071 *	1.000
r.grdp	−0.124 *	0.011	1.000
r.firm	0.464 *	0.010 *	−0.015	1.000
r.labor	0.313 *	0.033	0.139 *	0.489 *	1.000
finc	0.068 *	0.238 *	0.002	0.042 *	0.030	1.000
cult	−0.017	0.133 *	−0.003	−0.013	−0.019	−0.258 *	1.000
welf	0.089 *	−0.021	−0.002	0.147 *	0.064 *	−0.246 *	0.442 *	1.000
med	−0.129 *	0.039 *	−0.027	−0.044 *	−0.058 *	−0.056 *	−0.023	0.163 *	1.000

Note: * shows statistical significance at the 5% significance level.

). The correlation between the net migration rate (dependent variable) and the explanatory variables was significantly positive with GRDP per capita, rate of increase in establishments, rate of increase in workers, financial independence, and number of social welfare facilities. However, the rate of increase in GRDP per capita and number of beds in medical institutions showed significant negative values, and the per capita income growth rate and local medical infrastructure were contrary to the net migration rate in the region.

3.3. Analysis Results

3.3.1. Seohaean Expressway

To understand the overall migration effect due to highway construction, we used three models for analysis: a pooled OLS model, a fixed effects model, and a random effects model.

Table 5. Panel data regression analysis of the Seohaean Expressway *.

Variable	Pooled OLS	LSDV (Fixed Effect)	Random Effect
log grdp	0.174 *** (0.066)	0.301 *** (0.087)	0.222 *** (0.083)
r.grdp	−0.034 *** (0.004)	−0.040 *** (0.004)	−0.033 *** (0.004)
r.firm	0.274 *** (0.014)	0.284 *** (0.015)	0.217 *** (0.014)
r.labor	0.072 *** (0.010)	0.049 *** (0.009)	0.062 *** (0.009)
log finc	0.220 *** (0.070)	0.203 ** (0.089)	0.197 ** (0.090)
log cult	−0.037 (0.050)	−0.045 (0.061)	−0.059 (0.061)
log welf	0.091 ** (0.037)	0.108 *** (0.040)	0.032 (0.038)
log med	−0.153 *** (0.035)	−0.082 ** (0.039)	−0.111 *** (0.039)
seohae_dummy	−0.041 (0.121)	−0.067 (0.159)	−0.019 (0.161)
year_dummy	No	Yes	No
cons	−1.630 *** (0.302)	−1.870 *** (0.387)	−1.497 *** (0.372)
$\mathsf{\rho }$	-	0.077	0.073
$\mathrm{adj}-{\mathrm{R}}^2$	0.252	0.296	0.252
F-test(OLS)/${\mathsf{\chi }}^2$(fixed/random)	112.44 ***	817.96 ***	658.92 ***
Hausman test	100.06 ***
N	2977

Note: (1) *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. (2) The number in parentheses is a standard error.

shows the results of analysis. To find the most suitable method of analysis among the three models, the F-test and Hausman test were conducted; as a result, the fixed effects model was found to be the most suitable. To verify the appropriate method between the pooled OLS model and the fixed effects model, the F-test value was 112.44, thereby rejecting the null hypothesis, and indicating that the fixed effects model is more suitable. The Hausman test conducted to verify the adequacy of the fixed effects model and the random effects model showed the Hausman statistic at 100.06, thereby rejecting the null hypothesis that the correlation between the individual characteristics of the panel (${\mathrm{u}}_{\mathrm{i}}$) and the explanatory variable was 0, and proving that the fixed effects model was more suitable.

Therefore, according to the analysis focused on the results of the fixed effects model, both economic factors and local amenities were found to have an important impact on internal migration; however, the Seohaean Expressway construction did not affect migration.

Among the economic factors, GRDP per capita, rate of increase in establishments, and rate of increase in workers showed significant positive (+) values, while the rate of increase in GRDP per capita showed a significant negative (−) value. Specifically, it appears that population flows into regions with high income levels and rapid growth in the number of companies and job creation, indicating that individual income level and jobs have a significant impact on population agglomeration. On the other hand, the rate of increase in GRDP per capita has a negative (−) value because the size of income is considered more important than the speed at which income grows in internal migration. This can be attributed to the fact that in cities and counties with low GRDP per capita, the increase in GRDP per capita is significant even if only some positive factors impact the local economy, such as industrial complexes or company relocation.

Among the factors influencing local amenities, financial independence and the number of social welfare facilities had significant positive (+) values, whereas the number of beds in medical institutions had a significant negative (−) value, and the number of cultural facilities was not statistically significant. It was found that huge population flows are witnessed into regions that offer strong local financial systems, efficient public goods systems, and protective good social welfare programs. This serves as a key factor in migration, as residents are concerned about the benefits of social welfare policies and perceive the importance of public goods provided by both central and local governments. Regions with more beds in medical institutions, indicating the level of medical services, have lower net population inflows because the number of beds in medical institutions is not high when demand is high (local population), and outpaces supply even if the region has very high levels of medical and healthcare services, such as metropolitan areas or large cities. Conversely, it seems that the population in cities and counties is small, and the quantitative parameters of medical services are high due to the recent establishment of large-scale nursing hospitals. A greater number of cultural facilities per population, representing the quantitative level of cultural services, is considered attractive because it leads to greater population flows, even though there may be no significant effect on internal migration.

After the Seohaean Expressway was constructed, the dummy variable to determine the causal relationship with the net migration rate of the construction area was not statistically significant. Therefore, it can be concluded that construction of the Seohaean Expressway did not affect net population inflows into the construction area.

3.3.2. Jungbu Naeryuk Expressway

Our analysis of the model to understand the causal relationship between constructing the Jungbu Naeryuk Expressway and internal migration shows that the economic factors and local amenities driving internal migration were almost similar(

Table 6. Panel data regression analysis of Jungbu Naeryuk Expressway *.

Variable	Pooled OLS	LSDV (Fixed Effect)	Random Effect
log grdp	0.165 ** (0.066)	0.291 *** (0.087)	0.214 *** (0.082)
r.grdp	−0.034 *** (0.004)	−0.040 *** (0.004)	−0.033 *** (0.004)
r.firm	0.273 *** (0.014)	0.283 *** (0.015)	0.217 *** (0.014)
r.labor	0.072 *** (0.010)	0.049 *** (0.009)	0.062 *** (0.009)
log finc	0.228 *** (0.070)	0.211 ** (0.089)	0.206 ** (0.090)
log cult	−0.034 (0.049)	−0.042 (0.061)	−0.058 (0.061)
log welf	0.088 *** (0.037)	0.105 *** (0.040)	0.030 (0.038)
log med	−0.156 *** (0.035)	−0.084 ** (0.039)	−0.113 *** (0.039)
jungbu_dummy	0.183 (0.146)	0.144 (0.0192)	0.232 (0.193)
year_dummy	No	Yes	No
cons	−1.635 *** (0.302)	−1.877 *** (0.388)	−1.509 *** (0.372)
$\mathsf{\rho }$	-	0.077	0.074
$\mathrm{adj}-{\mathrm{R}}^2$	0.252	0.296	0.252
F-test(OLS)/${\mathsf{\chi }}^2$(fixed/random)	112.66 ***	817.1 ***	659.55 ***
Hausman test	96.43 ***
N	2977

Note: (1) *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. (2) The number in parentheses is a standard error.

).

Among the economic factors, GRDP per capita, rate of increase in establishments, and wage rate increases for workers showed significant positive (+) values, and rate of increase in GRDP per capita showed a significant negative (−) value.

Among the factors influencing local amenities, financial independence and the number of social welfare facilities showed significant positive (+) values, whereas number of beds in medical institutions showed a negative (−) value.

The impact of constructing the Jungbu Naeryuk Expressway on the region’s net migration rate was statistically insignificant, and highway construction did not affect local migration, unlike the Seohaean Expressway.

3.3.3. Seoul–Chuncheon Expressway

Analysis of the Seoul–Chuncheon Expressway shows that economic factors and availability of local amenities affecting internal migration were almost similar to the other aforementioned assessments; however, the dummy variable for the highway construction area showed a statistically significant positive (+) value, unlike the Seohaean and Jungbu Naeryuk Expressways (

Table 7. Panel data regression analysis of Seoul–Chuncheon Expressway.

Variable	Pooled OLS	LSDV (Fixed Effect)	Random Effect
log grdp	0.215 *** (0.066)	0.338 *** (0.087)	0.265 *** (0.082)
r.grdp	−0.034 *** (0.004)	−0.040 *** (0.004)	−0.033 *** (0.004)
r.firm	0.271 *** (0.014)	0.281 *** (0.015)	0.215 *** (0.014)
r.labor	0.070 *** (0.010)	0.049 *** (0.009)	0.061 *** (0.009)
log finc	0.180 ** (0.070)	0.166 * (0.089)	0.155 * (0.089)
log cult	−0.046 (0.049)	−0.054 (0.060)	−0.071 (0.061)
log welf	0.076 ** (0.037)	0.095 ** (0.040)	0.020 (0.038)
log med	−0.153 *** (0.034)	−0.082 ** (0.039)	−0.111 *** (0.039)
seochun_dummy	1.042 *** (0.203)	1.088 *** (0.266)	1.162 *** (0.269)
year_dummy	No	Yes	No
cons	−1.608 *** (0.300)	−1.865 *** (0.386)	−1.482 *** (0.371)
$\mathsf{\rho }$	-	0.076	0.073
$\mathrm{adj}-{\mathrm{R}}^2$	0.259	0.301	0.258
F-test(OLS)/${\mathsf{\chi }}^2$(fixed/random)	116.34 ***	839.91 ***	682.45 ***
Hausman test	94.87 ***
N	2977

Note: (1) *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. (2) The number in parentheses is a standard error.

).

Among the economic factors, GRDP per capita, rate of increase in establishments, and rate of increase in number of workers showed significant positive (+) values, whereas the rate of increase in GRDP per capita showed a significant negative (−) value.

Among the factors influencing local amenities, financial independence and number of social welfare facilities showed significant positive (+) values, whereas the number of beds in medical institutions showed a negative (−) value.

After the construction of the Seoul–Chuncheon Expressway, the construction area showed an increase in the net migration rate, and net population inflows increased due to highway construction. This indicates that improved accessibility due to highway construction had a positive effect on the region’s population growth.

3.3.4. Second Jungbu Expressway

Analysis of the Second Jungbu Expressway shows that economic factors and local amenities affecting internal migration were almost similar to other expressways, and the dummy variable for highway construction area showed a statistically significant positive (+) value, as in the Seoul–Chuncheon Expressway (

Table 8. Panel data regression analysis of Second Jungbu Expressway.

Variable	Pooled OLS	LSDV (Fixed Effect)	Random Effect
log grdp	0.178 *** (0.065)	0.298 *** (0.086)	0.225 *** (0.081)
r.grdp	−0.034 *** (0.004)	−0.040 *** (0.004)	−0.033 *** (0.004)
r.firm	0.268 *** (0.014)	0.281 *** (0.015)	0.215 *** (0.014)
r.labor	0.071 *** (0.010)	0.049 *** (0.009)	0.062 *** (0.009)
log finc	0.171 ** (0.070)	0.156 * (0.089)	0.144 (0.089)
log cult	−0.040 (0.049)	−0.045 (0.060)	−0.061 (0.061)
log welf	0.088 ** (0.037)	0.105 *** (0.040)	0.030 (0.038)
log med	−0.142 *** (0.034)	−0.074 * (0.039)	−0.102 *** (0.039)
jungbu2_dummy	1.593 *** (0.308)	1.652 *** (0.402)	1.778 *** (0.406)
year_dummy	No	Yes	No
cons	−1.512 *** (0.301)	−1.747 *** (0.386)	−1.368 *** (0.371)
$\mathsf{\rho }$	-	0.075	0.072
$\mathrm{adj}-{\mathrm{R}}^2$	0.259	0.301	0.258
F-test(OLS)/${\mathsf{\chi }}^2$(fixed/random)	116.42 ***	844.36 ***	686.59 ***
Hausman test	95.71 ***
N	2977

Note: (1) *, **, and *** indicate that statistical significance at the 10%, 5%, and 1% levels, respectively. (2) The number in parentheses is a standard error.

).

Among economic factors, GRDP per capita, rate of increase in establishments, and rate of increase in workers showed significant positive (+) values, whereas the rate of increase in GRDP per capita showed a significant negative (−) value.

Among factors influencing local amenities, financial independence and the number of social welfare facilities showed significant positive (+) values, whereas the number of beds in medical institutions showed a negative (−) value.

After the construction of the Second Jungbu Expressway, net migration rate in the construction area increased, and accessibility improved due to highway construction, and this had a positive effect on the region’s population growth. Moreover, compared with the Seoul–Chuncheon Expressway, the coefficient value of the dummy variable is large, and the impact of increasing the net population inflow in the construction area was greater following the construction of the Second Jungbu Expressway.

4. Conclusions

This research examined the effect of large-scale transportation infrastructure that connects regions by analyzing the impact of highway construction on interregional migration. In particular, the SOC construction effect under the National Balanced Development Policy promoted after 2000 was examined through the population change in the highway construction area. Therefore, we established two hypotheses, that is, the metropolitan-type highways (Seohaean and Jungbu Naeryuk) have a high probability of experiencing the straw effect, and commuter-type highways (Seoul–Chuncheon and second Jungbu) are likely to experience population growth in the construction region. These hypotheses were validated through an empirical analysis.

Empirical analysis showed that the regional migration effect differed for each of the four highways. Thus, metropolitan-type highways did not experience a migration effect, while commuter-type highways did. All four highways showed migration into regions where the gross regional domestic product, the rate of business, labor, and financial independence increased. Therefore, similar to previous studies, our results show that people migrated towards regions that experienced local economic growth.

In summary, the first hypothesis of migration into large cities from underdeveloped areas, as observed in the gravity model, was rejected in the metropolitan-type highways. However, the second hypothesis, that all population increased in the construction region, was accepted in the commuter-type highway. The reason for such inconsistent results among the examined highways was the complex interaction of the economy, housing, and amenities in interregional migration, in addition to accessibility. Therefore, while highway construction contributes to the expansion of interregional accessibility, not all highway constructions result in consistent migration into the construction regions.

In particular, roads passing through Seoul, Gyeonggi-do, and the Seoul metropolitan area were mainly built for regional commuting. Therefore, strengthened regional accessibility owing to highway construction is a significant factor for determining residences in large metropolitan areas, which motivates migration into highway construction areas.

The Seohaean Expressway and the Jungbu Naeryuk Expressway, which encompass small and medium-sized cities of non-metropolitan areas along their routes, do not appear to have any effect on internal migration. Thus, in terms of internal migration, the analysis results did not satisfy the objective that highway construction, connecting metropolitan areas and undeveloped areas (non-metropolitan area), leads to balanced national development.

This research examined the impact of highway construction on interregional migration through modeling. Previous research analyzed the economic, societal, and population impact of one route of a specific highway. However, the characteristics and purpose of the highway construction area have only been secondarily examined. This study found that different effects of highway construction are observed as the types of users vary according to the purpose of the highway.

Population serves as an important variable in relation to local production, labor, consumption, and infrastructure, but it cannot be deemed as the direct purpose and outcome of highway construction. Basically, the main purpose of building transport infrastructure is to revitalize human and material exchanges by reducing transportation costs, and the expansion of trade volume and production between regions can be seen as a direct effect. Therefore, we must be cautious about claiming the effectiveness of highway construction with population inflow and outflow. However, it has been confirmed that the impact of migration occurs differentially depending on the highway construction section in the analysis result, and this study can be used to give basic data to prepare a preemptive measure for local population redistribution during highway construction.

In future research, additional control variables that reflect the features of each highway must be reviewed, and highways must be systematically categorized based on these features. Additionally, research that establishes the causative relationship between population increase in the highway construction areas and data relevant to the purpose of highway use are required.