**4. Application**

## *4.1. Results of Transit E*ffi*ciency*

With the results of the NSBM DEA of this research, the overall efficiency score was obtained by the average sum or weighted multiplication of each stage output from the design and efficiency stages. Table 4 and Figure 3 show the efficiency evaluation results for the station areas in Seoul. Based on the efficiency evaluation results in Table 4, the average overall efficiency score is 0.349. The transit design and efficiency evaluation results average 0.453 and 0.245, respectively. Since the overall efficiency score is calculated using the transit design and transit efficiency scores, multiplying by a weight of 0.5 implies that the overall efficiency score is affected more by the design score rather than the efficiency score. Regarding the 0.349 score and 0.207 standard deviation (S.D.), there is clearly a large gap between the efficient and inefficient station areas in Seoul. The 10 station areas were determined to be efficient, i.e., Euljiro 1ga, Shindorim, Gupabal, Dongjak, Yeongdeungpo, Digital media city, Gasan digital, Magok, Bokjung, and Gaehwa station. The top 10 efficient stations are the DMUs with the highest scores of 1.000 in both the transit design and efficiency stages is served as benchmarks for the other 342 inefficient stations. As the means of the input variables for the top 10 efficient station areas, the population density, land value, number of households, and number of companies are 21,581 (person/ km2), 4478 (1000 won/m2), 2450 households, and 110 companies, respectively. The means of the intermediate variables for the top 10 efficient station areas are 1.8 subway lines, 55 bus lines, 91 stations, and 212 m distance between stations, respectively. The means of the output variables for the efficient station areas, i.e., the number of subway trips, bus trips, transfer trips, and energy consumption, are 182,420, 97,491, 11,747 trips and 439,666 Mcal/trip, respectively.



**Figure 3.** Visualization of the efficiency evaluation results for Seoul: (**a**) efficiency result of the 352 stations; (**b**) top 10 efficient stations; and (**c**) bottom 10 inefficient stations.

A comparison of the efficient station area scores for the 352 station areas shows that all the input variable scores of the efficient station areas are lower than the mean score of all the station areas. In particular, the population density and number of households in the efficient station areas are about 37% and 16% lower than the average value of all the station areas, respectively. From the statistics for the intermediate variables, the number of subway lines and bus stations are about 66% and 54% of the means for all the station areas, and the number of bus lines is almost the same as the mean for all the station areas. These efficient station areas have relatively low population densities and small household. Although the population densities and number of households are smaller than the average values for all the station areas, the transit infrastructures are also well constructed. This is the main reason that transit design is estimated to be high. From the statistics of the output variables, the number of subway trips, bus trips, and transfer trips, and the energy consumption are about 4.98, 1.01, 1.91, and 1.16 times higher, respectively, than the mean values for all the station areas. The transit efficiency score was estimated to be 1.000, since it has a small population density, number of households, and the number of transit infrastructures compared to the output variables. This is because the inputs are lower and outputs are higher than the means for all the station areas, respectively. Since both the design and efficiency scores were 1.000, the overall efficiency score was also estimated to be 1.000.

The balance of each stage is crucial to achieve an efficient station in the concept of transit efficiency. For instance, if the transit infrastructure is well established with low transit trips, the design stage scores will be high but the efficiency stage scores will be relatively low. On the contrary, if the transit infrastructure is not well established with high transit trips, the design stage scores will be low and the efficiency stage scores will be relatively high. Hence, the overall efficient stations have balance in both design and e fficiency stage scores. From the results of the evaluation, overall e fficient stations are usually located in areas that have a relatively low population density but have well-built transit infrastructures with high transit trips.

The bottom 10 ine fficient station areas include Donrimcheon, Hakyoeul, Eungbong, Guryong, Gaepodong, Dongdaemoon, Olympic Park, Geoyeo, Macheon, and Dokbawi stations. These bottom 10 ine fficient station areas are relatively estimated to be the lowest overall e fficiency scores among the 352 station areas. These 10 relatively ine fficient station areas are DMUs in which the TOD should take top priority in improvement. The overall average score of the bottom 10 ine fficient station areas was 0.090, which indicates that an improvement of 99.1% is required. The averages of design and e fficiency scores were 0.157 and 0.023, respectively, which means that about 84.3% of these areas are poorly designed and 98.8% are ine fficient. When targeting the e fficient station, transit infrastructure-related variables should be improved by 84.3% compared to the socio-economic related variables. The transit trip-related variables should also be increased by 98.8% compared to the infrastructure-related variables. For the input variables of the bottom 10 ine fficient station areas, the population density, land price, number of households, and number of companies totaled 38,853 (persons/km2), 7300 (1000 won/m2), 5329 households, and 140 companies, respectively. The intermediate variables, which are the outputs of the design stage, i.e., numbers of subways, bus lines, and bus stations and the distance between bus stops and subway stations averaged 1.2 subway lines, 11 bus lines, 20 stations, and 258 m, respectively. For the outputs, the numbers of subway trips, bus trips, and transfer trips, and the energy consumption totaled 37,676, 57,391, and 511 trips, and 210,026 Mcal/trip, respectively.

A comparison of the statistics for the bottom 10 ine fficient station areas with all the station areas was conducted. It shows that all the input values of the bottom 10 ine fficient station areas are higher than the mean values of all the stations. With respect to the statistics for the intermediate variables, the numbers of subway lines, bus lines, and bus stations are about 1.3, 1.8, and 1.5 times more, respectively. The distance between bus stops and subway stations is about 13% less than the mean distance for all the station areas. However, the output variable values were lower than the mean values of the 352 station areas. Although the number of subway trips was similar to the mean value for all the station areas, the numbers of bus trips and transfer trips and the energy consumption were about 40%, 92%, 44% less, respectively, than the mean values for all the station areas. Since the population and number of households are higher than the mean values for all the station areas, transit infrastructures must be better equipped than the mean required for all 352 station areas. The transit e fficiency scores are also very low since the output values do not meet the required number of transit infrastructures.
