*3.2. Study Areas and Data Collection*

In addition to the research model presented above, this study conducted a survey of residents and visitors in two cities, which have often been cited as prime examples of modern smart city projects in related studies, so as to evaluate their smart city technologies, facilities and services in a comprehensive manner.

Barcelona (A) is where the TFS projects have been gradually conducted in traditional urban spaces by local governments, while Songdo (B) is where a large-scale new city development project was implemented along with the TFS introduction at the same time. However, as this study is mainly aimed at examining and verifying the applicability of the research model based upon the analysis of citizens' responses, the name of the cities and the detailed services, advantages and disadvantages of each smart city project, and their satisfaction level will not be described here. Instead, the description will be based on the figures needed to analyze the research model and survey results. Specific research sites in the two cities were chosen through discussions with experts and civil servants in each city in order to secure valid samples. The spatial scope for the survey was set within a 2-km radius of where major TFS projects were carried out. The center of City A is where support facilities for knowledge-based industries were established as part of smart city projects, and the center of City B is where a center for the smart city operation, management and experience is located.

This study also fixed the number of respondents through stratified sampling based on the population of each administrative district and their age. However, the population setting and sampling were limited, to involve those aged between 15 and 65 in Barcelona, and those aged between 20 and 65 for Songdo. Both represent around 65.8 percent of the total population on average. The questionnaire was translated into English and the language used in each nation. The pilot survey was carried out including 52 officials in related fields in Barcelona and 73 in Songdo, in May 2018. Following the revision and review of the questionnaire, the main and additional polls were conducted from July 2018 until January 2019. Responses from 211 out of 421 people (Barcelona) and 197 out of 522 (Songdo) people were used for analysis.

#### *3.3. Methodology*

This study adopted the Partial Least Square (PLS) to analyze the SEM. As a data modeling technique for spectral data, PLS performs ordinary least squares (OLS) regressions with the least square algorithm, and it is achieved by extracting from the predictors the latent variables. As it is meant to improve the rationale of the model based on the coefficient measurement that maximizes R2, the PLS is suitable for the development and the verification of a theoretical model and is widely used in such fields as behavioral science, marketing, and organization [61].

Given that "User Characteristic" in Figure 2 is not a reflective indicator but a formal indicator, the PLS-SEM is deemed suitable for this research. As an analysis program, SmartPLS ver.3.2 was employed, and the internal consistency, reliability, validity, and the discriminant validity, among others, were applied in a comprehensive manner for its assessment. However, the composition reliability (*CR*) was mainly used in evaluating the internal consistency reliability of measurement variables, and the equations are as follows.

$$CR(P\_{\mathcal{C}}) = \frac{\left(\sum\_{i=1}^{M} L\_i\right)^2}{\left(\sum\_{i=1}^{M} L\_i\right)^2 + \sum\_{i=1}^{M} var(e\_i)}, \frac{\left(\sum\_{i=1}^{M} L\_i\right)^2}{\left(\sum\_{i=1}^{M} L\_i\right)^2 + \sum\_{i=1}^{M} \left(1 - L\_i^2\right)}\tag{1}$$

where *Li* is the standardized outer loadings of the latent variables, *ei* is the error of measurement variable I, var(*ei*) is the variance (1 − *<sup>L</sup>*<sup>2</sup> *<sup>i</sup>* ) of measurement error, and *M* is the number of variables.

As the PLS adopted here is mainly used for non-parametric statistics, the t-value was calculated through bootstrapping to verify the significance of each path coefficient. The evaluation criteria to verify the SEM and each variable are shown in Table 3 below.

**Table 3.** List of Criteria for Measuring SEM's Reliability and Validity.

