*4.1. Di*ff*erentiating Influencing Factors*

The results of the binary logistic regression analysis are provided in Table 2. For all significant influencing factors not excluded via the stepwise regression, the odds ratios (Exp(B)), the regression coefficients (β) and *p*-values are presented. A significant Exp(B) value (*p*-value ≤ 0.05) greater (less) than 1.0 indicates that, as the predictor increases, the odds of the outcome occurring (being a "Future-Refurbisher") increase (decrease). Statistically insignificant results are printed in non-bold letters.




**Table 2.** *Cont.*

Reference-category: "Non-Refurbishers"; Source: own calculation.

Regarding *Behavioral beliefs,* the factors *Indoor comfort*, *Reasonable for environment, Energy bills* and *Doubts about desired e*ff*ects* were significant and thus included in the regression model. The affiliated Exp(B) values for these factors suggest that it is more likely to be in the group of "Future-Refurbishers," given house owners assume that EERM enhance the housing comfort and that EERM lead to a reduction of the energy bill. Belonging to this group is also more likely for individuals, who are not or less skeptical regarding the doubts about the pursued effects of EERM. The Exp(B) value associated with the factor *Reasonable for environment* is smaller than 1.0. This indicates that house owners who think that EERM are good for the environment are less likely to be part of the group of "Future-Refurbishers."

When considering *Normative beliefs*, only the factor *Esteem friends*/*family* is significant and included in the regression model. The calculated Exp(B) value suggests that belonging to the group of "Non-Refurbishers" is more likely with a decreased appreciation of EERM by house owners' friends and family.

A look at the *Control beliefs* shows that there are ten significant factors identified as being relevant for predicting group membership. Among these factors the most differentiating factors with Exp(B) values bigger than 1.0 are *Financing problems, Consulting during conduction*, *Legal requirements and Consulting during planning.* These results indicate that being a member of the group of "Future-Refurbishers" is more likely the lower house owners perceive financial problems to be associated with EERM. Implementing EERM is, furthermore, more likely for those house owners who have a lower demand of consultation. Further, the results of the factors *Legal requirements* but also *Complex case* imply that the uptake of EERM is more probable if complying with legal requirements in connection with EERM is perceived as less complex and also when the building does not appear to be hard to treat.

Significant Exp(B) values smaller than 1.0 are calculated for the factors *Appropriate craftsmen*, *Time conduction* as well as *Complex promotion*, *Own capabilities* and *Time planning*. With respect to the wording of our statements and the answer scales used in the questionnaire, this can be interpreted as follows: the more positive a house owner's expectation is to find appropriate craftsmen for conducting EERM, the more likely this person belongs to the group of "Non-Refurbishers." This also applies to the perceived difficulty connected to governmental promotion—if individuals perceive governmental promotion as not complex, their belonging to the group of "Non-Refurbishers" is more likely. In contrast to the latter two rather surprising results, we also find that limited time for conducting and planning EERM and low *Own capabilities* of the house owners in this area increase the odds of belonging to the group of "Non-Refurbishers."

Along with factors associated with the TPB we included *Environmental awareness* factors and *Building conditions* as additional contextual aspects in our study. When considering the results related to the *Environmental awareness* aspects, there is only one significant factor, which is *Impacts of decisions.* The associated Exp(B) value suggests that belonging to the group of "Non-Refurbishers" is more probable if house owners do not or hardly consider the potential environmental impacts of their actions when making decisions.

With respect to *Building conditions,* the factors *Energy e*ffi*ciency* and *Comfort and appearance* are included, significant and connected to strong Exp(B) values greater than 1.0. These results suggest that being a "Future-Refurbisher" is more likely if house owners perceive an immediate need or a need within the next few years to take actions to improve the energy efficiency, the appearance or the comfort of their buildings. A perceived need to take care of a buildings structural condition in the foreseeable future (*Building fabric*) was also identified as increasing the likeliness of being part of the "Future-Refurbishers."

The factors which have been excluded via stepwise backward algorithm are presented in Table 3. In total 10 factors were excluded by the stepwise regression algorithm in SPSS.


**Table 3.** Excluded factors/statements.

Source: own figure.

#### *4.2. Overall Classification*

This section is intended to allow a better evaluation of the quality of the analysis results. Next to the actual and predicted group membership via binary logistic regression in Table 4 we furthermore provide additional information on specific quality indicators.


**Table 4.** Classification table showing the predicted and actual groups from the sample.

Source: own figure.

From 734 respondents who stated their willingness to conduct relevant refurbishment measures in the future, 91.4% were assigned to the correct group. In the case of the "Non-Refurbishers," this value is 73.5%. In total, the binary logistic regression function assigned 85.6% of the sample participants correctly. By comparing this proportion of correctly classified observations with the proportion expected by chance, known as proportional chance criterion (56.2%) [43], our model improves this indicator by almost 30%.

The pseudo R2 value that determines the amount of variance in the dependent variable explained by the independent variables of 0.649 (Nagelkerke) also indicates a very good quality of the analysis [44]. The finally computed significance levels associated with the model chi-square value (678.8) of *p* = 0.000 and the Hosmer and Lemeshow test with *p* = 0.960 (>0.05) also suggest a good model fit. For the assessment of multicollinearity among the considered variables, the variance inflation factors (VIF; details on VIFs can be derived from [44]) were calculated. None of these VIFs was higher than 3, which leads to the conclusion that multicollinearity does not negatively affect the quality of our results.
