*3.4. Analysis*

After explaining the characteristics of the respondents via descriptive statistics, we analyzed the survey using the recently introduced PLS-SEM approach. Notably, PLS-SEM is a powerful tool that has minimum requirements for estimation parameters, and it is effective in modeling latent parameters in a non-normal distribution [68]. PLS-SEM is a suitable research method for path analysis with variables that are indirectly measured through other variables. Indirectly measured variables are common latent variables, and this approach uses latent variables for path coefficient analysis [59,69].

In PLS-SEM, we substantiated the validity of the model and implemented a nonrecursive least squares method to retrieve the external weights and structural model relations. Finally, we used bootstrap resampling to evaluate the statistical significance. The collected data were programmed in SPSS 20 before PLS-SEM. To verify the hypotheses we used SMART PLS 3.0, an SEM tool. Using SmartPLS 3.0, this study tested the model with a path weight scheme. We evaluated model fit and reliability, and the heterotrait/monotrait ratio of correlations (HTMT) to confirm discriminant validity. Finally, we were able to provide the results of the structural model.


**Table 1.** Questionnaire source and number of items.

## **4. Results**

The model developed in this study is a tool for analyzing customers using PropTech services. Between 11 October 2021 and 15 November 2021, the mobile survey application registered 992 responses in total. After thoroughly examining the survey, we screened 524 valid and usable samples and calculated a 58.94% response rate. Table 2 lists the demographic information of the 524 respondents.


**Table 2.** Demographic information.

To test the model, we used SmartPLS 3.0 with a path weight scheme. The bootstrap procedure drew 524 cases and 5000 samples using the unsigned option. When evaluating and reporting results [64,78], the measurement model was evaluated before the structural model.

SmartPLS uses SRMR and GOF to evaluate model fit. The GOF is obtained by multiplying the average value of R2 by the average value of the average variance extracted (AVE) and taking the square root again. The GOF value of this research model was 0.694, which constitutes a good goodness of fit [68,79]. The SRMR value is calculated based on standardized residuals [80]. When the model's goodness of fit is complete, SRMR becomes 0, and if it is less than 0.08, it is judged that the model's goodness of fit is good. It can be judged that the SRMR of this research model had a high goodness of fit of 0.051. In addition, an RMS\_theta value of 0.116 indicates that the model is appropriate, with higher values indicating lower levels of appropriateness [81].

Table 3 shows the results of the reliability and definitive factor analysis. In general, an item can be considered valid if its standard loading value is 0.5 or greater. If the mean AVE value is also greater than or equal to 0.5, the grouping factor can be considered as a reliability valid [78] composite, as was the case for the five reflectively measured constructs in our study ranging from 0.93 to 0.96, as these exceeded the minimum requirement of 0.70.

In this study, the variance inflation factor (VIF) was identified as a potential factor proposed by Knock [82] to investigate the common method variance (CMV) that may occur in PLS-SEM. As a result of checking for multicollinearity in the path between latent variables, the VIF did not exceed the threshold of 5, with minimum and maximum values of 1.442 and 3.456, respectively. The CMV was not an issue in the present study. In addition, the possibility of the CMV was low because the correlation coefficient between the variables was not high [69].

The Fornell and Larcker [83] criterion showed that all the AVE values for the specular construct were higher than the squared cross-construct correlation, indicating discriminant validity. Similarly, all the indicator loadings were higher than their respective cross-loadings, thus providing further evidence of discriminant validity. Table 4 shows the diagonal AVE values and the diagonal squared cross-composition correlations.


**Table 3.** Validity and reliability of measures.

**Table 4.** Discriminant validity results.


To confirm discriminant validity, the heterotrait/monotrait ratio of correlations (HTMT) was evaluated, as suggested by Henseler et al. [84] (Table 5). Discriminant validity was established if the HTMT value was less than 0.90. In this study, the HTMT value was found to be between 0.144 and 0.891, thereby confirming the safety of the discriminant validity.

**Table 5.** Heterotrait/monotrait ratio of correlations.


The structural model of the results is shown in Figure 2. R-squares were also used to judge the path coefficients of the endogenous latent variables. Most of the path coefficients with significance were found to be related at a level of *p* ≤ 0.01. The path coefficient of *p* ≤ 0.05 (ease of use -> user satisfaction and information quality -> intention to use) and the path coefficient of *p* ≤ 0.10 (system quality -> intention to use and service quality -> intention to use) showed a statistical relationship and indicated that meaningful analysis was possible. Table 6 lists all of the calculated values.

**Figure 2.** Structural equation model. Notes: \*\*\* *p* ≤ 0.01; \*\* *p* ≤ 0.05; and \* *p* ≤ 0.10. Dashed lines represent non-significant relationships.

In Smart PLS, one can substantiate the effect of specific individual effects; the resulting analysis is as follows.

As shown in Table 7, "Security -> Innovation resistance" describes the situation wherein system quality enhancement calls for resistance. Security-related aspects entail not only product quality but also social quality; thus, quality enhancement before ensuring perfect security might undermine a consumer's trust in service quality. Information sharing has a negative effect on IR and a positive effect on PER. When asking questions about information sharing with security, personal information security is excluded, and only the effect of information sharing is evaluated. Therefore, to enhance convenience, information sharing reduces IR, but could positively contribute to PER.




**Table 7.** *Cont.*


PropTech innovation is addressed before IT is implemented across the traditional services. Therefore, IR is enhanced, and users are required to adapt to the new service, which has a positive effect on PER. In traditional real estate-related services, the valuation of properties and the provision of information on the surrounding areas are the primary activities. However, PropTech can supply personalized information about the surroundings and provide a personalized service experience.

Efficiency is a key PropTech service feature that provides a new interface for data searching and transactions. Therefore, IR becomes more important when one needs to accept a new IT service; however, PER is positively affected.

Innovation resistance (efficiency) acts as a partial parameter in information sharing (perceived usefulness and perceived ease of use). It can be concluded that information sharing and efficiency contribute positively towards enhancing perceived usefulness and perceived ease of use by reducing innovation resistance. However, it has been shown that, by increasing IR, complementation and innovation may negatively affect perceived usefulness and perceived ease of use. Innovation has a direct positive effect on perceived usefulness and perceived ease of use; however, it has a negative effect in relation to some factors.

For PER, all independent variables except security showed positive partial factor effects between utility and accessibility.

Regarding the analysis of the parameters, the obtained results are as follows: (1) security takes IR as a parameter and reduces customer use intention through perceived usefulness and perceived ease of use; (2) information sharing takes PER as a factor and increases the users' use intentions despite IR; however, its effects are limited; (3) innovation resistance takes PER as a factor and positively contributes towards enhancing the continuous intention to use. This does not exhibit a negative effect on IR. In particular, PER takes perceived usefulness and perceived ease of use as parameters; (4) innovation resistance affects the customers' continuous intention to use the product without perceived usefulness, and a relationship exists between IR and perceived ease of use, but disruption does not accompany perceived ease of use; (5) efficiency shows a general positive effect via PER, as well as a negative effect via IR; and finally, in general (6), IR unfolds the most efficient process for aligned positive effects and does not exhibit a negative effect. Even with information sharing, the IR effect is limited. Any procedure to enhance security has been shown to have a negative effect on CIU (continuous intention to use), which is the ultimate goal of this study.
