Self-Sufficiency of New Administrative Capitals (NACs) Based on Types and Commuting Characteristics of Citizens: Case Study of Sejong

Abstract

In recent decades, new administrative capitals (NACs) are being developed in Asia and developing countries due to the overcrowding of capitals and large cities. The self-sufficiency of a planned city is considered important for balanced national development. However, no study has specifically analyzed the degree of self-sufficiency of NACs. Therefore, focusing on the city of Sejong (NAC, South Korea) as an example, this study evaluated self-sufficiency using data regarding household composition and travel characteristics. The results of the three-step analysis are as follows: First, the commuting distance of the NAC was longer than that of traditionally developed cities, with relatively little internal commuter traffic in the NAC. Second, commuting to and from the NAC was primarily to large cities nearby. Third, regarding the characteristics of households living in the NAC, the ratio of second-generation households was higher and that of single-person households was relatively small compared with traditional cities. In addition, a spatial correlation in the form of a longer commuting distance in the second generation and shorter commuting distance in single-person households was confirmed. The findings of this study hold important implications for policymakers and urban planning bodies when developing an NAC.

1. Research Background and Objectives

New administrative capitals (NACs) are being developed worldwide to allay the side-effects of overpopulation in big cities and generally pursue balanced national development.

Sejong is one of South Korea’s NACs. Sejong City was selected as a candidate for the new administrative capital as it received a high evaluation in five categories: balanced national development effect, domestic and foreign accessibility, impact on the natural environment, natural conditions as a living ground, and urban development cost/economic feasibility. On 11 August 2004, Sejong was confirmed as the target area for the new administrative capital after collecting public opinion through public hearings and various media in 13 cities, including Seoul [1].

As part of the Policy on Balanced National Development, Sejong was designed to reduce overpopulation in metropolitan areas. Initially, the objective was to create an administrative capital by relocating government offices and administrative agencies to Sejong. Sejong was, however, labeled an administrative city since it was considered unconstitutional to build a new administrative capital [2]. Although Sejong is referred to as an administrative city instead of an administrative capital, it functions as a de facto administrative capital, and most administrative agencies have relocated there.

The self-sufficiency of cities is a widely known and acknowledged concept, with several studies conducted on the topic [3,4]. Self-sufficiency is an essential function of cities in terms of their maintenance and construction. The self-sufficiency of cities is critical for an NAC designed for balanced national development [5].

Although several studies have been conducted on policy-level and social aspects of NACs [6,7,8], there exists insufficient research evaluating whether NACs are self-sufficient. Accordingly, this study aimed to analyze the self-sufficiency of Sejong as an NAC.

2. Literature Review

2.1. Literature on NACs

In pursuit of balanced national development, many NACs have recently been built worldwide by relocating various government departments and administrative agencies and investing massive resources. An administrative capital refers to the central city responsible for the main administrative functions of a country.

South Korea has built Sejong, an NAC, to allay the overcrowding of Seoul, the capital city, and pursue the national policy of balanced national development [9].

Egypt has built an NAC approximately 45 km east of Cairo to resolve issues emerging from overpopulation in its capital. Key government organs, including parliament, have been relocated to the administrative capital. Several studies have centered on this development [7,10,11].

In the Philippines, a state-led urban development plan to build an administrative capital (New Clark City) has emerged to decentralize the population in the metropolitan areas. A study has focused on sustainable environmental city research on the topic [12].

In Malaysia, Putrajaya was built as the administrative capital as a remedy to combat overpopulation and traffic congestion in its capital, Kuala Lumpur [13]. Various studies on urban growth and ecosystem protection in Putrajaya have been conducted [14].

Indonesia is pursuing a state-led urban planning project to relocate the capital from Jakarta to the eastern part of Kalimantan. The Indonesian parliament passed the Capital City Bill on 18 January 2022, leading to many studies on urban planning in relation to capital relocation [8,15].

Therefore, NAC construction is currently a global trend. Countries that have completed or are building and discussing NACs are shown in

Table 1. Cases of construction and discussion of NAC construction.

Classification	Brazil	Australia	Japan	Malaysia	Republic of Korea	Egypt	Philippines	Indonesia
Type	NAC Construction	NAC Construction	NAC Construction	NAC Construction	NAC Construction	NAC Construction	NAC Discussion	NAC Discussion
Purpose/Background	Inland area development	Federal national symbol project	Balanced national development	Balanced national development	Balanced national development	Balanced national development	Balanced national development	Balanced national development, resolving environmental issues
Period	1955–1970	1908–1980	1992–Currently invalid	1993–2010	2007–Currently in progress	2015–Currently in progress	Currently under discussion	Currently under discussion
Special note	Security and national development center	Academic/research/art function integration	Reflecting the national characteristics of decentralization	Except for the Royal Family and the Bundestag	Changed to a multi-functional administrative city due to unconstitutional issues	5 million metropolis targets	Considered as a way to spread the growth of large cities	President’s strong push

. In addition to building new administrative capitals, many countries have pursued state-led planned city development, such as capital relocation and large-scale new town development for various reasons, including addressing environmental problems and efficiently utilizing urban space. Such forms of urban development are commonly found in Asia and developing nations.

The Xiong’an New Area near Beijing in China is a state-led project designed to disseminate key functions of the Chinese capital. As one of President Xi Jinping’s core projects, the project was designed and developed as a national-level special zone, receiving an investment of over USD 300 billion. Various related urban planning studies have been conducted [16,17].

State-led planned city construction is being promoted in many countries in addition to China, and the construction of new capitals has been completed in various countries, including Pakistan (Islamabad) and Kazakhstan (Astana) [9].

2.2. Literature on Self-Sufficiency

Several studies have been conducted on the self-sufficiency of cities, especially new cities [3,4]. Self-sufficiency is an important factor in the functionality of cities. In particular, the self-sufficiency of cities has become a key factor for NACs, which are created with the goal of balanced national development and the relocation of capital cities.

Urban self-sufficiency is an important function that allows new cities to become more sustainable [18]. A study identified self-sufficiency as a key criterion for sustainability indices [19]. Some studies have focused on the “land” aspect in evaluating self-sufficiency. One study focused on establishing self-sufficiency through housing site/lot development projects [20], and another stressed establishing self-sufficiency through the allocation of land to enhance the self-functionality of a city [21].

Commutes are used as a primary indicator to evaluate urban self-sufficiency. A study revealed that commuting is an important indicator when evaluating the economic self-sufficiency of a city [22]. Another study conducted a comparative analysis of the level of economic self-sufficiency in new cities by focusing on commutes [23]. While evaluating the self-sufficiency of new cities (Phase 1) in metropolitan areas, commutes were confirmed as a key indicator [24]. A study derived self-sufficiency variables and used the analytic hierarchy process (AHP) to conduct a time-series analysis of self-sufficiency in metropolitan areas [25]. The authors concluded that economic self-sufficiency can be determined through the efficiency of commutes. Another study argued that commuting to work reflects the economic function of a city and that the clarity of origin-destination location and regular travel can be sufficient indicators for measuring the evaluation index of economic self-sufficiency [26].

Reducing commuting distances is a prerequisite for cities to become self-sufficient and furthermore to build sustainable cities. For urban self-sufficiency, a work—worker balance must be sought. Some studies found that better work—worker balance resulted in shorter average commuting distances and times [27,28,29]. A study claimed that shorter commutes, both in terms of time taken and distance, are desirable for greater sustainability [30]. Another study emphasized that commuting distance can be reduced depending on the land use type and that reducing the commuting distance will lead to the expansion of a sustainable city [31].

The results of the above studies have confirmed that urban self-sufficiency is critical for the optimal development of cities. In addition, commutes have been identified as a key indicator to evaluate the self-sufficiency of cities. Accordingly, this study used commutes as a key indicator to evaluate the self-sufficiency of Sejong.

While several studies have been conducted on the policies and development of new cities [32,33], there is insufficient research on the self-sufficiency of cities, not-withstanding the fact that NACs have been built in line with comprehensive balanced national development plans. Therefore, this study assessed the self-sufficiency of Sejong using indicators used in previous studies. The study is unique in that it evaluated self-sufficiency using correlation analysis between travel behavior and household composition.

3. Research Scope and Methods

3.1. Research Scope

The administrative districts of South Korea are composed of one special city, six metropolitan cities, eight provinces, one special self-governing province, and one special self-governing city. The subject of this study is Sejong, an NAC in South Korea (Figure 1). Sejong was initially designed as an administrative capital following the relocation of government offices and administrative agencies [2]. It was, however, labeled an administrative city based on constitutional requirements, as stated earlier.

The full-scale development of Sejong began in 2007 with the goal of building a self-sufficient city with a population of 500,000 by 2030 aimed at pursuing balanced national development and enhancing national competitiveness [34]. The development of Sejong has been conducted in three phases. Phase 1 (initial phase), from 2007 to 2015, involved the relocation of the central administrative agency and the building of urban infrastructure. Phase 2 (maturity phase), from 2016 to 2020, involved fostering self-sufficiency and developing urban infrastructure. Phase 3 (completion phase), from 2021 to 2030, involves the completion of a self-sufficient city.

This study evaluated the self-sufficiency of Sejong, which has recently completed its Phase 2 development. The cities analyzed in comparison with Sejong were selected among those geographically close and demographically comparable to Sejong (Daejeon, Cheongju, and Cheonan) (Figure 1). The regional scope of research was narrowed to administrative dongs, or neighborhoods in which the new administrative capital is built. Accordingly, the administrative dong was also set as the research scope for the cities used in the comparison.

3.2. Research Method

Similar to previous studies, this study set commutes used in self-sufficiency indices as the key indicator of the self-sufficiency of Sejong. Regarding the data sources, to analyze commutes, we used the Korea Transport Database for data on commutes from 2019, when the Phase 2 development of Sejong was near completion. For geographic information system (GIS)-related analysis, we used the Statistical Regional Boundary (2019) data of the Statistical Geographic Information Service (SGIS). For analysis by household type, we used statistical data from SGIS (2020).

The specific data analysis process was divided into three stages. First, to compare and analyze basic economic self-sufficiency, we extracted the travel distance for each city as well as the ratio of intra-city and inter-city travel and confirmed the same using analysis of variance (ANOVA). Second, we verified the spatial dispersion of the destination of commutes in each region using GIS to analyze the commuting behavior of travelers by city. Third, we verified the correlation between household type and self-sufficiency in cities using bivariate Moran’s I, the LISA cluster map, ANOVA, and a spatial economic model.

4. Data Analysis and Results

4.1. Commuting Distance and Ratio of Intra-City and Inter-City Travel

To analyze the commuting distance by city, we obtained commuting data and the SGIS of Sejong and the cities for comparison (Daejeon, Cheongju, and Cheonan). We then calculated the distance between dongs using GIS. When the point of departure and destination are the same, the distance is considered zero. In this case, the dong area with the same origin and destination is set as S. The area (S) of this region was assumed to be a circle, and the r value was set as the distance value. The value of r was obtained using $r=\sqrt{\frac{S}{\pi }}$. The following Equation (1) was used to obtain the average commuting distance by city. The results are shown in Figure 2.

$$A{D}_C^P=\frac{{{\displaystyle \sum }}_i^n\frac{T_{c_{ij}}^p\times D_{ij}}{{{\displaystyle \sum }}_i^n{T}_{c_i}^p}}{n}$$

(1)

$A{D}_C^P$ is the average travel distance of $P$ in city $C$; $T_{c_{ij}}^p$ is the total time for $P$ to travel from the starting point $i$ to the ending point $j$ of city $C$; $D_{ij}$ is the distance from the starting point $i$ to $j$; $C$ is the city analyzed (Sejong, Daejeon, Cheonan, or Cheongju); $P$ is the purpose of travel (commute to work, commute to school, commute to a private academy, or commute to shopping areas); $n$ is the number of administrative dongs in city $C$.

We identified the commutes in dongs of each city to analyze intra-city and inter-city commutes. The destinations of the commutes were identified from all areas, including the dongs of Sejong and the other cities compared. We calculated the ratio of intra-city travel by processing the data examined.

We conducted ANOVA to assess whether the differences in the average travel distance among the cities and the ratios of intra-city travel were statistically significant. In addition, we conducted multiple comparisons using the Tukey test. The results are shown in

Table 2. Multiple comparisons of average commutes by city using the Tukey test.

Multiple Comparisons
Tukey’s HSD
Dependent variable	(i) Regional code	(j) Regional code	Mean difference (i–j)	Standard error	p value	95% confidence interval
						Lower bound	Upper bound
Average commuting distance	Sejong	Daejeon	2621.0023636 *	588.0476634	0.000	1090.866767	4151.137960
		Cheongju	3472.1191656 *	635.2671735	0.000	1819.115547	5125.122784
		Cheonan	3634.0101259 *	682.3864079	0.000	1858.399409	5409.620843
Average distance to school	Sejong	Daejeon	−983.7157282	852.7424699	0.657	−3202.603225	1235.171768
		Cheongju	−2415.3300036 *	921.2166501	0.048	−4812.391460	−18.268548
		Cheonan	−1373.0647563	989.5454180	0.509	−3947.921800	1201.792288
Average distance to private academy	Sejong	Daejeon	−624.1560903	486.3043008	0.575	−1889.549302	641.237121
		Cheongju	−594.5800376	525.3539429	0.671	−1961.582780	772.422705
		Cheonan	−1255.5413656	564.3206590	0.122	−2723.937861	212.855130
Average distance to shopping areas	Sejong	Daejeon	514.5920831	494.6446715	0.726	−772.503278	1801.687444
		Cheongju	637.8508508	534.3640352	0.632	−752.596697	2028.298399
		Cheonan	820.1581663	573.9990507	0.484	−673.422090	2313.738422

* The mean difference is significant at the level of 0.05.

and

Table 3. Multiple comparisons of average internal ratio by city using the Tukey test.

Multiple Comparisons
Tukey’s HSD
Dependent variable	(I) CODE	(J) CODE	Mean difference (I–J)	Standard error	p value	95% confidence interval
						Lower bound	Upper bound
Internal commuting ratio	Sejong	Daejeon	−0.1948380 *	0.0208854	0.000	−0.249183	−0.140493
		Cheongju	−0.2060066 *	0.0225625	0.000	−0.264716	−0.147298
		Cheonan	−0.2036130 *	0.0242360	0.000	−0.266677	−0.140549
Internal school commuting ratio	Sejong	Daejeon	−0.0341300	0.0189290	0.277	−0.083384	0.015125
		Cheongju	−0.0057187	0.0204490	0.992	−0.058928	0.047491
		Cheonan	−0.0324174	0.0219658	0.455	−0.089574	0.024739
Internal private academy commuting ratio	Sejong	Daejeon	−0.0008355	0.0297607	1.000	−0.078275	0.076604
		Cheongju	−0.0005510	0.0321504	1.000	−0.084208	0.083106
		Cheonan	0.0546093	0.0345351	0.393	−0.035253	0.144472
Internal shopping commuting ratio	Sejong	Daejeon	−0.0545728 *	0.0067486	0.000	−0.072133	−0.037013
		Cheongju	−0.0561269 *	0.0072905	0.000	−0.075097	−0.037157
		Cheonan	−0.0569028 *	0.0078312	0.000	−0.077280	−0.036525

* The mean difference is significant at the level of 0.05.

.

The data analysis in Table 2 shows that the difference in the average commuting distance between Sejong and the compared cities was 2621 for Daejeon, 3472 for Cheongju, and 3634 for Cheonan, with a p value of p < 0.001. Therefore, the average distance of commutes in Sejong was statistically significantly greater than that of the compared cities.

The results of the analysis in Table 3 show that the ratio of intra-city commutes and the ratio of trips made for shopping within the city had negative values (−) compared to those of the other cities. The mean difference between the ratio of intra-city commutes and the ratio of trips made for shopping within the city showed a p value of p < 0.001. This indicates that the ratio of intra-city commutes and the ratio of trips made for shopping within the city of Sejong were lower than those of the other cities. Therefore, Sejong has a higher ratio of inter-city travel for commuting to work and shopping.

The above multiple comparisons indicate that, first, travelers in Sejong, on average, commute for a longer distance than those in other cities. Second, travelers in Sejong engage in a greater ratio of inter-city travel for commuting to work and for shopping than those in other cities. Although self-sufficiency in a new administrative capital is an important urban function, according to the results of the analysis, Sejong does not yet meet the standards of a self-sufficient city.

4.2. Spatial Variance Analysis of Commuting Destinations by Region Using GIS

To illustrate the commutes of Sejong and the other cities with spatial dispersion and analyze commuting behavior, we used the standard deviational ellipse (SDE) function of GIS. This analysis method has been used in various studies to analyze spatial distribution [35,36].

First, we set the dong of each city as the X point and the destination as the Y point and calculated the volume of commutes based on X and Y. Then, we created an ellipsoid using the deviation representing the distribution of this value. The shape of the SDE is shown in Figure 3.

According to the analysis, the SDE of the compared cities is formed within the area and the ellipsoid is relatively circular. However, the SDE of Sejong is illustrated as an ellipsoid that is wider than that of the other cities. In addition, the ellipsoid moves into that of Daejeon, intersecting it. These results indicate that numerous travelers in Sejong commute to and from the neighboring city Daejeon.

4.3. Analysis of Household Type by Region

Households in South Korea are classified into four types: single-person, one-generational, two-generational, and three-generational households. This classification has been commonly used in previous studies when analyzing household composition [37,38]. We first examined the household member data of Sejong and the other cities to derive the ratio of each household type. The results are shown in Figure 4. Sejong indicated a relatively higher ratio of two-generational households and a lower ratio of single-person households compared to the other cities.

To ensure that the values represent meaningful results, we assessed whether the ratios of two-generational households and single-person households in Sejong were statistically significant compared to those of the other cities. We conducted ANOVA on the ratio of household data by city, and multiple comparisons and verification using Tukey’s test. The results are shown in

Table 4. Multiple comparisons of ratio of household type by city using Tukey’s test.

Multiple Comparisons
Tukey HSD
Dependent variable	(I) Local_code	(J) Local_code	Mean difference (I–J)	Standard error	p value	95% confidence interval
						Lower bound	Upper bound
One-generational household ratio	Sejong	Daejeon	−0.0237611 *	0.0030318	0.000	−0.031552	−0.015970
		Cheongju	−0.0113181 *	0.0032984	0.003	−0.019794	−0.002842
		Cheonan	0.0096786 *	0.0034025	0.023	0.000935	0.018422
Two-generational household ratio	Sejong	Daejeon	0.1267152 *	0.0099961	0.000	0.101028	0.152403
		Cheongju	0.0959205 *	0.0108751	0.000	0.067974	0.123867
		Cheonan	0.0727344 *	0.0112183	0.000	0.043906	0.101563
Three-generational household ratio	Sejong	Daejeon	−0.0025234	0.0012602	0.187	−0.005762	0.000715
		Cheongju	−0.0069762 *	0.0013711	0.000	−0.010499	−0.003453
		Cheonan	−0.0021211	0.0014143	0.438	−0.005756	0.001513
Single-person household ratio	Sejong	Daejeon	−0.1246426 *	0.0107597	0.000	−0.152292	−0.096993
		Cheongju	−0.1049691 *	0.0117058	0.000	−0.135050	−0.074888
		Cheonan	−0.0990003 *	0.0120752	0.000	−0.130031	−0.067970

* The mean difference is significant at the level of 0.05.

.

The results indicated that the mean difference in the ratio of two-generational households between Sejong and the other cities was positive (+), whereas the mean difference in single-person households was negative (−). The mean difference of the ratios of both two-generational and single-person households had a p value of p < 0.001. This finding indicates that Sejong comprised more two-generational households and fewer single-person households than the other cities.

To analyze whether the calculated ratio of household composition by city was related to the self-sufficiency of the cities, we evaluated the spatial autocorrelation of the commuting distance and household type ratio by city. Spatial autocorrelation measures the extent to which a variable at a specific location is related to other values at nearby locations. Spatial autocorrelation is positive when the level of interaction exceeds the expected level and the surrounding locations have comparable values. In contrast, it is negative when the high value of one variable is close to the low value of the variable at a nearby location. Spatial autocorrelation is 0 when there is no relationship between close values [39].

We applied bivariate Moran’s I and the LISA cluster map, which are the most common techniques to analyze the existence of spatial autocorrelation. Bivariate Moran’s I and the LISA cluster map explain the spatial patterns formed by two different variables [40]. Bivariate Moran’s I derives the spatial scatter plot of the first variable on the vertical axis and that of the second variable on the horizontal axis. The two variables are internally standardized. Spatial delayed operation is applied to the standardized variables. The slope of the regression line represents the linear correlation between the variable on the horizontal axis and the variable on the vertical axis of a nearby location [41]. Based on Moran’s I, the bivariate LISA cluster map provides a feasible method to characterize spatial correlations between the spatial distributions of several variables [42]. Bivariate Moran’s I (Ikl) can be presented as shown in Equation (2) below:

$$I_{kl}=Z_k^i{\displaystyle \sum }_{j=1}^n{W}_{ij}{Z}_l^j$$

(2)

where $Z_k^i=\left[x_k^i-\overline{x_k}\right]/{\sigma }_k, Z_l^j=\left[x_l^i-\overline{x_l}\right]/{\sigma }_l; x_k^i$ is the value of variable $k$ at location $i$; $x_l^j$ is the value of variable $l$ at location $j$; $\overline{x_k}$ and $\overline{x_l}$ are the mean values of the variables $k$ and l, respectively; ${\sigma }_k$ and ${\sigma }_l$ are the variance of $x$ for variables $k$ and $l$, respectively; and $W_{ij}$ is the spatial weight matrix, which can be represented based on the distance weighting between locations $i$ and $j$ [39].

First, a bivariate Moran’s I scatter plot was derived from the results of the spatial correlation between commuting distance by city and the ratio of household type. The values are shown in Figure 5. The results of the bivariate Moran’s I analysis indicate that the correlation between commuting distance by city and the ratio of two-generational and single-person households was higher. The bivariate Moran’s I of two-generational households was approximately 0.203, indicating that the regression line was an upward slope. This finding implies that a higher ratio of two-generational households makes it more likely that the commuting distance will be longer. The bivariate Moran’s I of single-person households was approximately −0.185, indicating that the regression line was a downward slope. This result implies that a higher ratio of single-person households makes it more likely that the commuting distance will be shorter.

Second, the bivariate LISA cluster map, which illustrates the interaction (spatial correlation) between commuting distance by city and ratio of household type, was classified into six categories: high-high (HH), high-low (HL), low-high (LH), low-low (LL), insignificant variables, and neighborless. The bivariate LISA cluster map between commuting distance by city and the ratio of household type is shown in Figure 6. In particular, in the new administrative capital Sejong, five HH-type regions were derived from the cluster map. The HH-type regions were found only in Sejong. Considering that there are a total of nine dongs in Sejong, approximately 56% of the regions in Sejong were determined to be of the HH type. This finding indicates that Sejong had more two-generational households than the other cities, which suggests a longer average commuting distance, thereby demonstrating a spatial correlation. In addition, in the cluster map of the ratio of single-person households and commuting distance, Sejong indicated a pattern different from the other cities. Four HL-type regions were found only in Sejong. Considering that there are a total of nine dongs in Sejong, approximately 44% of the regions in Sejong were determined to be of the HL type. This finding indicates that fewer single-person households lead to a greater average commuting distance, thereby confirming a spatial correlation.

Therefore, the results of the above analysis indicate that Sejong has more two-generational and fewer single-person households than the other cities. Household type was correlated with commuting distance. The presence of more two-generational households suggests a longer commuting distance on average. In addition, fewer single-person households also indicated a longer commuting distance on average. The new administrative capital was found to comprise more two-generational households than the other cities, which suggested a correlation with longer average commuting distance. Furthermore, the new administrative capital had fewer single-person households than the other cities, also indicating a correlation with longer average commuting distance.

We further analyzed the relationship between household type and commuting distance. An appropriate analysis model was selected to analyze the relationship between the two variables with special autocorrelation. Earlier studies have confirmed that the spatial economic model is suitable to analyze the relationship between variables while controlling autocorrelation [43].

The spatial economic model consists of the following steps. First, the researcher conducts an ordinary least squares (OLS) regression model. If the Lagrange multiplier (LM) statistics of both models as calculated by the LM diagnoses are not significant, OLS is the most appropriate model for the analysis [39]. Secondly, if any of the LM diagnoses from the spatial lag model and spatial error model are meaningful, the corresponding model is determined as an appropriate model. Thirdly, if the LM diagnoses for both models are significant, the robust LM diagnosis is examined, and the model with the relatively higher significance is chosen as the final model.

The appropriate analysis model was chosen using OLS in this study. In OLS, the commuting distance was set as the dependent variable. The independent variable was defined as one-generational, three-generational, single-person, and non-family households, excluding the ratio of two-generational households. Two-generational households were defined as the reference value of the model. In addition, a spatial weight matrix was constructed for LM diagnoses. The standard for the inter-individual adjacency of the spatial weight matrix was applied with the queen-based contiguity standard. Queen-based contiguity is a method of defining the adjacency of a specific spatial entity with all neighboring spatial entities with common edges or corners. Based on the set values, OLS was performed and LM diagnoses were identified. The LM lag value was significant at 0.02083, whereas the LM error value was insignificant at 0.11992. The results of the analysis are shown in

Table 5. Results of the spatial dependence diagnoses.

Classification			p-Value
Commuting distance	Lagrange multiplier	Lag **	0.02083
		Error	0.11992
	Robust LM	Lag **	0.03289
		Error	0.20188

** The mean difference is significant at the level of 0.05.

.

Based on these findings, the spatial lag model was selected as the appropriate model for analysis. The derived R-squared value was 0.141762, indicating a somewhat low explanatory power. The results of the analysis are shown in

Table 6. Analysis of spatial lag model based on commuting distance.

Dependent Variable	Independent Variable	Coefficient	p-Value
Commuting distance	Spatial lag Coef **	0.1871784	0.02763
	One-generational household ratio	5007.54	0.41118
	Three-generational household ratio	−37328.51	0.08294
	Single-person household ratio **	−3786.229	0.04884
	Non-family household ratio	−12986.07	0.49267
R-squared			0.141762

** The mean difference is significant at the level of 0.05.

. The results indicate that the commuting distance decreases when the ratio of single-person households exceeds that of two-generational households. This finding was significant in the standard of significance probability of 5% and is consistent with the analysis results performed previously in this study. Regarding other household types, the commuting distance increased or decreased as the ratio increased. However, because the derived results were insignificant, we did not interpret them further.

5. Conclusions

The construction of NACs has emerged as an important developmental goal worldwide [9]. NACs are a new form of urban development without precedents and, therefore, are characteristically different from existing cities. It is important that such new cities are economically self-sufficient to realize balanced development [23]. However, there are a lack of studies that analyze the self-sufficiency of NACs. Regarding previous research, various studies have been conducted on urban self-sufficiency by using commuting as an important indicator. Therefore, this study evaluated the urban self-sufficiency of Sejong, an NAC, using commuting as the main indicator. The analysis indicated the following findings.

First, Sejong lacks self-sufficiency compared to existing cities in terms of commuting distance, which was longer in Sejong. In addition, a higher ratio of inter-city commutes compared to that of intra-city commutes was observed in Sejong. This finding shows that many residents of Sejong commute to other cities, suggesting that the city is not self-sufficient.

Second, Sejong appears to be greatly influenced by neighboring metropolitan cities. The SDE drawn using GIS displayed the travel behavior of people living in Sejong commuting frequently to big cities nearby. It seems that despite the better residential environment of Sejong, its residents commute to neighboring big cities since there are not enough jobs (excluding those at government departments and research institutes that were involuntarily relocated) in Sejong.

Third, Sejong has a different household composition ratio compared with that of other cities. It is presumed that household composition is related to the self-sufficiency of cities. The results of Moran’s I analysis showed a spatial correlation in which a higher proportion of two-generational households is linked to greater commuting distance, and fewer single-person households are linked to a longer commuting distance. The results of the LISA cluster map analysis of Sejong also indicated a spatial correlation between two-generational households and long commuting distance, and a spatial correlation between fewer single-person households and long commuting distance. The spatial economic model indicated that the commuting distance decreased as the ratio of single-person households exceeded that of second-generational households. A study found that single-person households are more heavily affected by their work than multi-person households [44]. Based on this finding, it can be assumed that NACs with a low ratio of single-person households that commute to work within a short distance lack a variety of jobs. NACs are unique in that administrative departments and public offices are mostly involuntarily relocated. This may explain the high ratio of two-generational households in the NAC. The fact that household type is related to commuting distance indicates that commuting distance is also related to self-sufficiency.

Based on the above results, the policy implications of NACs are as follows. First, Sejong is not as self-sufficient as it was designed and expected to be. To build a more economically self-sufficient city, policies must attract and relocate firms to the city when the state decides to involuntarily relocate its government departments and research institutes. By gradually and effectively providing government-level benefits such as tax exemptions and subsidies to firms relocating to the NAC, it would be possible to create new jobs within the city. This will encourage single-person households to reduce their commuting distance and help the city become self-sufficient.

Second, the NAC is heavily impacted by the big cities nearby. This raises the need to consider turning the NAC into a megacity and expanding the characteristics of the city. Expanding the metropolitan scale of urban functions to that of a megacity would lead to the development of various industries, the creation of better infrastructure, and the improvement of the residential environment. In this way, the NAC will grow into a city that better meets the objectives of balanced national development. Along with various ongoing studies on megacities [5], the current findings are expected to be effectively used in establishing related policies on NACs.

This study is significant in that it analyzed and verified the self-sufficiency and travel behavior of an NAC and deduced pertinent policy implications. In particular, although many studies have evaluated self-sufficiency using commuting data, there is insufficient research on the correlation between household type and the self-sufficiency of an NAC. Accordingly, this study is novel in that it analyzed the types of household composition, identified their spatial correlation with commuting behavior, and established a correlation between household composition and the self-sufficiency of the NAC.

Although the present study reveals important findings, it has several limitations. The findings of this study are limited in that the self-sufficiency analysis was restricted to Sejong. Further research to evaluate the self-sufficiency of NACs in other countries needs to be conducted to determine whether the results are consistent with those of this study. Moreover, the correlation analysis between household composition and self-sufficiency was also limited to Sejong. Since household composition can be affected by various factors such as gender and age, it is necessary to not only analyze the NACs of other countries but also classify household types using more detailed criteria. Such limitations are hoped to be addressed in future research on the analysis of the correlation between household composition and the commuting characteristics of NACs.