M-insurance Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies

Mobile Insurance Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Disclaimer: The copyright of this document rests with the author The views expressed in this thesis are those of the author and not necessarily express the views of the Delft University of Technology or Deloitte consulting ii M-insurance Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Master thesis December, 2014 S.A.J.P (Sebastian) Derikx S.A.J.P.Derikx@gmail.com Delft University of Technology MSc System, Engineering, Policy Analysis & Management (SEPAM) Faculty of Technology, Policy and Management Section Information and Communication Technology SPM5910 Master Thesis Project The Netherlands Graduation Committee Chair: Prof Dr Yao-Hua Tan – ICT section Officious Chair: Prof Dr W.A.G.A Bouwman – ICT section First Supervisor: Dr ir G.A Reuver MSc – ICT section Second Supervisor: Dr ir M Kroesen TLO section External Supervisor: Drs A Beers – Deloitte Consulting External Supervisor: A Mahawat Khan MSc – Deloitte Consulting Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies “Science and technology revolutionize our lives, but memory, tradition and myth frame our response” – Arthur M Schlesinger Jr iv |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Acknowledgements This thesis marks the final step for the completion of my Master System Engineering, Policy Analysis and Management (SEPAM) at the Delft University of Technology This graduation project started in May 2014 intern at Deloitte Consulting and since that moment a significant part of my life has been spent on this thesis Therefore, I am proud to present the final version of my thesis First, I would like to express my gratitude to my graduation committee for their great support last seven months I would like to thank Assistant Professor Mark de Reuver, my first supervisor, for his general support in both theoretical and practical field I would like to thank Assistant Professor Maarten Kroesen, my second supervisor, for his sharpening insights in the statistical methodologies and Professor Harry Bouwman for chairing my graduation committee Second, I would like to thank Deloitte Consulting for giving me the great opportunity to conduct my master thesis intern Your innovative and progressive mindset helped me to finalize this thesis I would like to express special thanks to Arjen Beers and Amira Mahawat Khan for their feedback, enjoyable way of working and involving me in the business of mobile insurance Third, I would like to thank my friends and family for their support during my graduation Special thanks to my friends, siblings and (former) roommates for lending their listening ears and encouragement, my parents for bringing me to this point in life and my girlfriend for giving me the freedom and loving support required for this accomplishment I hope you all enjoy reading the result, Amsterdam, December 2014, Sebastian Derikx i Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies List of Acronyms AV B1/B2 BMT BR CB CSMIS CSMS DOI FA FSI GR IID IoT KR LU MB MI M-Insurance MNL-model MOE PAYD PC PU RW SD SEM TAM UBI WTP ii Additional insurance offerings Relative consumer saving Business Model Transformations (innovation) Behavioral rewarding Choice-based conjoint Context Sensitive Mobile Insurance Services Context Sensitive Mobile Services Diffusion of innovation Factor Analysis Financial Service Industry Third party advertisement Independently and Identically Distributed Internet of Things Kilometer Registration Likelihood of use Expected monetary benefit Mobile insurance Mobile insurance Multinomial logit model Margin of error Pay-as-you-drive Privacy concerns Perceived usefulness Registration road behavior Standard deviation Structural Equation modeling Technology acceptance model Usage based insurance Willingness to pay |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Management summary: Research problem Ongoing digitalization results in both threats as opportunities for the insurance sector Increased transparency stimulates switching behavior and shifts the insurance market to a more price based competition Together with recent developments such as the ban on intermediary commissions and the separation of banking and insurance activities, the traditional business model is put under pressure By fully reaping the benefits of mobile technologies, such as portability, social interactivity, context sensitivity, connectivity and individuality, a variety of opportunities for innovative insurance services arises A more differentiated product portfolio can shift the price based competition to a more quality focus which enables insurers to operate in more niche markets focusing on higher margins In the last few years, privacy concerns associated with the consumer use of mobile technologies, have been the subject of many research papers A number of privacy studies empirically verified the negative effect of perceived privacy concerns on the intention of use online and mobile services As the disclosure of personal information is often necessary in obtaining online and mobile services, privacy concerns could inhibit people’s intention to use them as well This could have major implications for the adoption of mobile insurance since privacy concerns regarding the insurance industry are already relatively high in general Therefore, it is essential, in the development of future mobile insurance services, to understand the role of associated privacy concerns Accordingly, this study aims to increase understanding of mobile insurance related privacy concerns, its relation on consumer’s ‘likelihood of use’ and potential compensating factors as perceived usefulness and expected monetary benefits Therefore, the objective of this research is to further develop understanding towards the mitigating effect of perceived usefulness and monetary rewards on privacy concerns regarding the likelihood of use for mobile insurance services In line with this objective the following main research question is developed: RQ In what way can privacy concerns, affecting the likelihood of use mobile insurance services, be mitigated by expected monetary benefits and perceived usefulness? Domain on Mobile Insurance For a clear and consistent understanding of this research question the definition of mobile insurance for this study is defined as “insurance products and services based on context sensitive mobile technologies” Hereby insurance products and services involve all direct customer focused activities of an insurer Thus, both the insurance policy itself and supportive services are involved Context sensitivity of mobile technologies involves the ability to both gather and respond to real or simulated data unique to current location, environment, and time Mobile insurance covers a broad field of insurance services In order to get a better understanding on the scope of mobile insurance a categorization is made This categorization is based on an explorative scan to all worldwide mobile insurance services These worldwide mobile insurance services are subsequently categorized on its consumer functionalities and validated with insurance industry and technology experts The final categorization, with a brief elaboration per category is listed below: Usage based insurance; Behavioral rewarding; Up-to-date insurance package; With a usage-based insurance premium, consumers pay only premium for actual use of their insurance By rewarding customers for less risky behavior, the insurer is trying to reduce the risk of accidents By using personal (context sensitive) information of consumers, relevant personalized insurance products could be provided iii Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Preventative information services; Accident detection & prevention; Mobile accessibility; Personal dashboards; Additional informative services; Consumer context information offers insurers the opportunity to provide consumers with relevant context related preventative information By detecting (potential) accidents as early as possible, damages could be prevented and minimized Mobile technologies facilitate a communication channel for sales and services By measuring individual behavior, insight could be provided in risk profiles of consumers to increase risk awareness Context sensitive information offers opportunities for several semiinsurance services Theoretical background on the concept of privacy Within literature a variety of definitions and interpretations for privacy is present, however a unified account of privacy has yet to emerge This study interprets the definition of privacy as a tradable interest; “an interest that individuals have in sustaining a ‘personal space’ free from interference by other people and organizations” Subsequently, this definition is operationalized to facilitate the measurement of privacy A commonly used (reverse) operationalization of privacy in literature is the measurement of privacy concerns Therefore, privacy is measured in this study by privacy concerns Due to its plurality and inconsistency, a unified account for privacy is still absent in literature Some scholars used another approach and instead of searching for an inclusive definition of privacy, they developed a typology for privacy Recent literature defined seven types of privacy of which three are relevant for the application of (current) mobile insurance: "The right to move about in public or semi-public space without being Privacy of location and space identified, tracked, or monitored." Privacy of behavior and action "The ability to behave in public, semi-public or one’s private space without having actions monitored or controlled by others." Privacy of data and image "Concerns about making sure that individuals’ data is not automatically available to other individuals and organizations and that people can exercise a substantial degree of control over that data and its use.” A majority of consumers considers the disclosure of personal information as essential in modern life The disclosure of personal information is however contrary with the definition of privacy; sustaining a ‘personal space’ Consequently, numerous studies consistently concluded that people are very concerned about their online privacy Aforementioned contradiction imply that individuals consider a utilitarian trade-off between perceived benefits and sacrifices of disclosing personal information Hereby privacy concerns have to be considered as a sacrifice Previous literature states that providers can mitigate the negative effect of privacy concerns on the ‘likelihood of use’ in two ways; (1) by offering privacy policies regarding the handling and use of personal information and (2) by offering benefits such as monetary rewards or convenience These compensating are further operationalized as expected monetary benefits and perceived usefulness No existence of a direct relation between the construct of privacy concerns, perceived usefulness and expected monetary benefits is found in literature However, several IT adoption studies in literature suggest an indirect relation through the construct of likelihood of use Hereby, the likelihood of use is positively affected by the perceived usefulness and expected monetary benefits and negatively affected by privacy concerns These findings are combined in a conceptual model which is validated for the case of mobile insurance by the explorative assessment iv |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Analysis and results In order to provide an answer on the main research question, two quantitative assessments are conducted By means of a consumer survey and multiple regression, an explorative assessment is conducted to the relations between the constructs of likelihood of use, privacy concerns, perceived usefulness and expected monetary benefit Hereby, the conceptual model is validated By means of a conjoint survey, a more in-depth assessment to the buy-off value of privacy is conducted for all relevant types of privacy, for the case of Pay-As-You-Drive (PAYD) insurance An overview of both assessments is provided in Table 0.1 Table 0.1: Overview assessments Explorative assessment: Conjoint assessment Method Descriptive statistics & multiple linear regression MI Category All identified categories of MI services Conjoint analysis (statedchoice) One MI service (PAYD) Goal output: Collection technique: Consumer attitude on Likelihood of use & relation to PC, PU & MB Hard-copy & electronic survey Buy-off value for privacy concerns Hard-copy survey Number of respondents 137 55 Explorative assessment The construct of perceived usefulness appears to be in general the strongest predictor for the likelihood of use mobile insurance The relation between these two constructs is significant for all categories of mobile insurance Mobile insurance services with a higher perceived usefulness are likely to raise more interest of consumers for future use The relation between the construct of expected monetary benefits and the likelihood of use shows to be positive as well, however not significant for all categories of mobile insurance Expected monetary benefits appear not to be a significant predictor for the use of mobile accessibility Overall it can be concluded that mobile insurance services with a higher expected monetary benefit for the consumer are likely to raise more interest of consumers for future use In contrast to previous constructs, the relationship between the construct of privacy concerns and the likelihood of use appears to be negative, however not significant for all categories of mobile insurance Privacy concerns appear not to be a significant predictor for the use of Accident detection and prevention and Mobile accessibility Overall it can be concluded that mobile insurance service with raised privacy concerns are likely to have a negative impact on the likelihood of use mobile insurance Altogether, it can be concluded that the likelihood of use mobile insurance services is primarily driven by its perceived usefulness Thereafter, consumers’ likelihood of use mobile insurance services is driven by raised expectations on accompanied monetary benefits and inhibited by increased privacy concerns However, not for every category of mobile insurance the predictors have a significant relation with the likelihood of use, no significant contra relations are found These findings seem to support the relations as found in literature Conjoint assessment Although the explorative assessment shows us that monetary benefits are not the strongest predictor for consumers’ likelihood of use mobile insurance services, the conjoint assessment is used for a more in-depth analysis to the buy-off value of privacy For this analysis, the buy-off value of privacy is determined for all individual relevant types of privacy for the case of pay-as-you-drive (PAYD) insurance PAYD insurance is an automobile insurance whereby the premium is dependent on the actual car-use Most common used indicators for car-use are mileages, and driving behavior v Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Respondents are willing to sell their privacy of location and space through continuously disclosing the GPS-location of their car for a financial compensation of €2,27 per month Privacy of behavior and actions appears to have slightly higher buy-off value since respondents are willing to continuously provide insight in their car-acceleration, car-deceleration and steering behavior, for a financial compensation of €2,98 per month Regarding the privacy of data and image two buy-off values are determined related to the internal and external (secondary) use of personal information Hereby, secondary use is operationalized as the unauthorized use of personal information for personalized advertisement Respondents are willing to sell their privacy of data image for third party advertisement for a financial compensation of €2,77 per month In contrast to the external use of personal information, respondents are willing to pay a monthly contribution of €2,91 for internal (insurance related) personalized advertisement However, these outcomes cannot blind be generalized to the entire population, it can be concluded that respondents derive more disutility from external use of privacy related information than internal use Discussion and conclusion In conclusion, we can say that privacy concern are likely to rise with the use of mobile insurance services However these concerns can be compensated by both perceived usefulness of the service and an expected monetary benefits The compensation by the expectation for financial benefits appears to have a smaller effect than compensation by elevated perceptions on the usefulness of a mobile insurance service However when the expectation on monetary benefits is amplified with a financial compensation, the buy-off values for different types of privacy appear to be rather small Hereby, consumers perceive their privacy of behavior and action as more valuable than their privacy of location and space Regarding privacy of data and image, the buy-off value seems to be dependent on the one who exploits their data; the data holder or an external party While the use of consumers’ personal information for personalized advertisement by the data holder appears to be beneficial, personalized advertisement by third parties is perceived as adversely This study is the first attempt in literature in which the buy-off value for different types of privacy is determined As this study proves, is the buy-off value of privacy varying for different types of privacy, supporting its plurality A plural approach on privacy could provide a more detailed method for future technology acceptance studies Emerging trends, such as the ongoing digitalization, quantified-self, internet of things and big data require the disclosure of different sets of personal and contextual information Consequently, different types of privacy may be involved affecting consumer adoption to another extent Therefore, it is recommended to include a plural construct of privacy in future technology acceptance studies Further research is recommended to evaluation the value of privacy for other mobile (insurance) services A comparison between the values of privacy of these individual services may result in interesting insights for technology adoption and privacy literature By proving the existence of multiple types of privacy dependent on the specific characteristics of concerned (mobile) services, this study validates the findings of Nikou (2012) that IT artifact should no longer be treated as ‘Black-Box’ Further, analysis methods such as factor analysis and structural equation modeling (SEM) have not been applied in the explorative survey By applying SEM in further research on the explorative dataset to examine both the effect of individual constructs per categories of mobile insurance and a generic constructs on the likelihood of use, could result in interesting insights, in line with Nikou’s (2012) findings, that IT artifact should no longer be treated as ‘Black-Box’ vi |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Category 2a: Behavioral discounting (car) Table 7.11: Multiple regression C2a Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,752a ,566 ,562 1,1082 ,763b ,582 ,576 1,0908 ,771c ,595 ,585 1,0787 a Predictors: (Constant), C2PU1 b Predictors: (Constant), C2PU1, C2PC1 c Predictors: (Constant), C2PU1, C2PC1, C2MB1 Model Regression Residual Total Regression Residual Total Regression Residual Total Sum of Squares 215,958 ANOVAa df Mean Square 215,958 F 175,832 Sig ,000b 165,808 135 1,228 381,766 222,325 159,442 136 134 111,162 1,190 93,424 ,000c 381,766 227,010 136 75,670 65,032 ,000d 154,757 133 1,164 381,766 136 a Dependent Variable: C2LU1 b Predictors: (Constant), C2PU1 c Predictors: (Constant), C2PU1, C2PC1 d Predictors: (Constant), C2PU1, C2PC1, C2MB1 Coefficientsa Model (Constant) C2PU1 (Constant) C2PU1 C2PC1 (Constant) C2PU1 C2PC1 C2MB1 a Dependent Variable: C2LU1 Unstandardized Coefficients B Std Error ,722 ,271 ,728 ,055 1,575 ,456 ,682 ,058 -,135 ,058 1,204 ,487 ,622 ,064 -,147 ,058 ,154 ,077 Standardized Coefficients Beta ,752 ,705 -,138 ,643 -,150 ,126 t 2,659 13,260 3,458 11,851 -2,313 2,472 9,667 -2,532 2,007 Collinearity Statistics Tolerance VIF Sig ,009 ,000 ,001 ,000 ,022 ,015 ,000 ,013 ,047 1,000 1,000 ,881 ,881 1,135 1,135 ,690 ,872 ,778 1,450 1,147 1,285 133 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Category 2b: Behavioral discounting (health) Table 7.12: Multiple regression C2b Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,671a ,451 ,447 1,3696 ,686b ,471 ,463 1,3487 ,702c ,493 ,481 1,3262 a Predictors: (Constant), C2PU2 b Predictors: (Constant), C2PU2, C2PC2 c Predictors: (Constant), C2PU2, C2PC2, C2MB2 Model Regression Residual Sum of Squares 207,789 Total Regression Residual Total Regression Residual Total ANOVAa df Mean Square 207,789 F 110,780 Sig ,000b 253,219 135 1,876 461,007 217,261 243,746 136 134 108,631 1,819 59,720 ,000c 461,007 227,079 136 75,693 43,035 ,000d 233,929 133 1,759 461,007 136 a Dependent Variable: C2LU2 b Predictors: (Constant), C2PU2 c Predictors: (Constant), C2PU2, C2PC2 d Predictors: (Constant), C2PU2, C2PC2, C2MB2 Unstandardized Coefficients B Std Error Model (Constant) 1,112 ,322 C2PU2 (Constant) C2PU2 C2PC2 (Constant) ,692 1,956 ,662 -,148 1,598 ,066 ,487 ,066 ,065 ,502 ,559 -,176 ,204 ,078 ,065 ,086 C2PU2 C2PC2 C2MB2 a Dependent Variable: C2LU2 134 Coefficientsa Standardized Coefficients Beta ,671 ,642 -,146 ,542 -,174 ,176 t Sig Collinearity Statistics Tolerance VIF 3,455 ,001 10,525 4,016 10,011 -2,282 3,181 ,000 ,000 ,000 ,024 ,002 1,000 1,000 ,960 ,960 1,042 1,042 7,137 -2,719 2,363 ,000 ,007 ,020 ,662 ,926 ,688 1,511 1,079 1,453 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Category 3: Up-to-date insurance package Table 7.13: Multiple regression C3 Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,736a ,541 ,538 1,1361 ,759b ,577 ,570 1,0958 ,770c ,593 ,584 1,0780 a Predictors: (Constant), C3PU b Predictors: (Constant), C3PU, C3MB c Predictors: (Constant), C3PU, C3MB, C3PC Model Regression Residual Sum of Squares 205,720 Total Regression Residual Total Regression Residual Total ANOVAa df Mean Square 205,720 F 159,381 Sig ,000b 174,251 135 1,291 379,971 219,075 160,896 136 134 109,537 1,201 91,226 ,000c 379,971 225,408 136 75,136 64,654 ,000d 154,563 133 1,162 379,971 136 a Dependent Variable: C3LU b Predictors: (Constant), C3PU c Predictors: (Constant), C3PU, C3MB d Predictors: (Constant), C3PU, C3MB, C3PC Unstandardized Coefficients B Std Error Coefficientsa Standardized Coefficients Beta Model (Constant) ,674 ,280 C3PU C3MB (Constant) ,711 ,144 ,633 ,221 ,932 ,056 ,313 ,059 ,066 ,457 ,736 C3PU (Constant) ,584 ,223 -,127 ,062 ,065 ,054 ,605 ,207 -,138 C3PU C3MB C3PC a Dependent Variable: C3LU ,655 ,204 t Sig Collinearity Statistics Tolerance VIF 2,410 ,017 12,625 ,460 10,687 3,335 2,040 ,000 ,646 ,000 ,001 ,043 1,000 1,000 ,842 ,842 1,187 1,187 9,461 3,426 -2,335 ,000 ,001 ,021 ,748 ,842 ,874 1,336 1,188 1,144 135 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Category 4: Preventative information services Table 7.14: Multiple regression C4 Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,736a ,542 ,538 1,1163 ,780b ,608 ,602 1,0361 ,799c ,638 ,629 ,9998 a Predictors: (Constant), C4PU b Predictors: (Constant), C4PU, C4MB c Predictors: (Constant), C4PU, C4MB, C4PC Model Regression Residual Sum of Squares 198,711 Total Regression Residual Total Regression Residual Total ANOVAa df Mean Square 198,711 F 159,453 Sig ,000b 168,238 135 1,246 366,949 223,098 143,851 136 134 111,549 1,074 103,910 ,000c 366,949 233,989 136 77,996 78,020 ,000d 132,960 133 1,000 366,949 136 a Dependent Variable: C4LU b Predictors: (Constant), C4PU c Predictors: (Constant), C4PU, C4MB d Predictors: (Constant), C4PU, C4MB, C4PC Unstandardized Coefficients B Std Error Coefficientsa Standardized Coefficients Beta Model (Constant) ,719 ,283 C4PU C4MB (Constant) ,699 ,211 ,582 ,303 1,167 ,055 ,284 ,057 ,064 ,398 ,736 C4PU (Constant) ,546 ,290 -,168 ,056 ,062 ,051 ,575 ,273 -,178 C4PU C4MB C4PC a Dependent Variable: C4LU 136 ,612 ,286 t Sig Collinearity Statistics Tolerance VIF 2,538 ,012 12,627 ,745 10,204 4,766 2,929 ,000 ,457 ,000 ,000 ,004 1,000 1,000 ,813 ,813 1,230 1,230 9,740 4,704 -3,301 ,000 ,000 ,001 ,783 ,809 ,937 1,278 1,236 1,067 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Category 5: Accident detection and prevention Table 7.15: Multiple regression C5 Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,661a ,437 ,433 1,1091 ,688b ,474 ,466 1,0767 a Predictors: (Constant), C5PU b Predictors: (Constant), C5PU, C5MB Model Regression Residual Sum of Squares 129,010 Total Regression Residual Total ANOVAa df Mean Square 129,010 166,070 135 1,230 295,080 139,723 136 69,861 155,358 134 1,159 295,080 136 F 104,874 Sig ,000b 60,257 ,000c a Dependent Variable: C5LU b Predictors: (Constant), C5PU c Predictors: (Constant), C5PU, C5MB Unstandardized Coefficients B Std Error Model (Constant) ,539 ,432 C5PU (Constant) ,777 ,086 ,076 ,445 C5PU C5MB a Dependent Variable: C5LU ,713 ,207 ,077 ,068 Coefficientsa Standardized Coefficients Beta t Sig Collinearity Statistics Tolerance VIF 1,249 ,214 ,661 10,241 ,194 ,000 ,846 1,000 1,000 ,607 ,198 9,302 3,040 ,000 ,003 ,924 ,924 1,082 1,082 Excluded Variablesa Model Beta In t Sig C5MB ,198b 3,040 ,003 C5PC -,089b -1,369 ,173 C5PC -,085c -1,348 ,180 a Dependent Variable: C5LU b Predictors in the Model: (Constant), C5PU c Predictors in the Model: (Constant), C5PU, C5MB Partial Correlation ,254 -,117 -,116 Collinearity Statistics Minimum Tolerance VIF Tolerance ,924 1,082 ,924 ,991 1,009 ,991 ,990 1,010 ,917 137 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Category 6: Mobile accessibility Table 7.16: Multiple regression C6 Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,739a ,546 ,543 ,9459 a Predictors: (Constant), C6PU Model Regression Residual ANOVAa df Sum of Squares 145,223 Total Mean Square 145,223 120,777 135 ,895 266,000 136 F 162,325 Sig ,000b a Dependent Variable: C6LU b Predictors: (Constant), C6PU Model Unstandardized Coefficients B Std Error (Constant) -,025 ,403 C6PU a Dependent Variable: C6LU ,896 ,070 Coefficientsa Standardized Coefficients Beta ,739 t Sig -,063 ,950 12,741 ,000 Collinearity Statistics Tolerance VIF 1,000 1,000 Excluded Variablesa Model Beta In t C6MB ,112b 1,894 C6PC -,096b -1,638 a Dependent Variable: C6LU b Predictors in the Model: (Constant), C6PU 138 Sig ,060 ,104 Partial Correlation ,161 -,140 Collinearity Statistics Minimum Tolerance VIF Tolerance ,938 1,066 ,938 ,963 1,039 ,963 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Category 7: Personal dashboards Table 7.17: Multiple regression C7 Model Summary Adjusted R Std Error of the Model R R Square Square Estimate ,818a ,669 ,667 1,0707 ,836b ,699 ,694 1,0252 ,855c ,732 ,726 ,9710 a Predictors: (Constant), C7PU b Predictors: (Constant), C7PU, C7MB c Predictors: (Constant), C7PU, C7MB, C7PC Model Regression Residual Sum of Squares 312,749 Total Regression Residual Total Regression Residual Total ANOVAa df Mean Square 312,749 F 272,827 Sig ,000b 154,754 135 1,146 467,504 326,675 140,829 136 134 163,337 1,051 155,417 ,000c 467,504 342,108 136 114,036 120,951 ,000d 125,396 133 ,943 467,504 136 a Dependent Variable: C7LU b Predictors: (Constant), C7PU c Predictors: (Constant), C7PU, C7MB d Predictors: (Constant), C7PU, C7MB, C7PC Unstandardized Coefficients B Std Error Coefficientsa Standardized Coefficients Beta Model (Constant) ,058 ,238 C7PU C7MB (Constant) ,866 -,165 ,668 ,309 ,606 ,052 ,236 ,074 ,085 ,294 ,818 C7PU (Constant) ,607 ,383 -,186 ,072 ,083 ,046 ,573 ,316 -,187 C7PU C7MB C7PC a Dependent Variable: C7LU ,631 ,255 t Sig Collinearity Statistics Tolerance VIF ,244 ,808 16,517 -,701 9,017 3,640 2,063 ,000 ,484 ,000 ,000 ,041 1,000 1,000 ,459 ,459 2,176 2,176 8,454 4,646 -4,046 ,000 ,000 ,000 ,439 ,437 ,946 2,277 2,290 1,057 139 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies E4 Estimation General model Generic model Table 7.18: Multiple regression generic model Model Summary Std Error of the Estimate 1,09155 Model R R Square Adjusted R Square ,773a ,597 ,596 a Predictors: (Constant), ALLPC, ALLMB, ALLPU Model Regression Residual ANOVAa df Sum of Squares 1928,743 Total Mean Square 642,914 1301,096 1092 1,191 3229,839 1095 F 539,593 Sig ,000b a Dependent Variable: ALLLU b Predictors: (Constant), ALLPC, ALLMB, ALLPU Model Unstandardized Coefficients B Std Error (Constant) ,842 ,148 ALLPU ALLMB ALLPC a Dependent Variable: ALLLU ,650 ,213 -,135 ,023 ,023 ,019 140 Coefficientsa Standardized Coefficients Beta ,628 ,196 -,144 t Collinearity Statistics Tolerance VIF Sig 5,695 ,000 28,590 9,129 -7,283 ,000 ,000 ,000 ,764 ,800 ,939 1,308 1,250 1,065 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| Fit general model Table 7.19: Fit generic model Model Summary Std Error of the Model R R Square Adjusted R Square Estimate ,781a ,610 ,599 1,08774 a Predictors: (Constant), MB6, ALLPC, ALLMB, PC3, PU5, ALLPU, MB1, MB2a, MB2b, PC4, PU3, MB4, PC6, PC5, PC1, PC2b, PC2a, PU4, Cat6, MB3, MB5, PU2a, PU2b, Cat1, PU1, Cat2b, PU6, Cat4, Cat3, Cat2a, Cat5 Model Regression Residual ANOVAa df Sum of Squares 1970,944 Total 31 Mean Square 63,579 1258,895 1064 1,183 3229,839 1095 F 53,736 Sig ,000b a Dependent Variable: ALLLU b Predictors: (Constant), MB6, ALLPC, ALLMB, PC3, PU5, ALLPU, MB1, MB2a, MB2b, PC4, PU3, MB4, PC6, PC5, PC1, PC2b, PC2a, PU4, Cat6, MB3, MB5, PU2a, PU2b, Cat1, PU1, Cat2b, PU6, Cat4, Cat3, Cat2a, Cat5 Model Unstandardized Coefficients B Std Error (Constant) ,824 ,161 ALLPU ALLMB ALLPC Cat1 Cat2a Cat2b Cat3 Cat4 Cat5 Cat6 PU1 PU2a PU2b PU3 PU4 PU5 PU6 PC1 PC2a PC2b PC3 PC4 PC5 PC6 MB1 MB2a MB2b MB3 MB4 MB5 MB6 a Dependent Variable: ALLLU ,639 ,230 -,136 -,349 ,380 ,774 ,108 ,343 -,429 -,610 ,015 -,017 -,080 -,054 -,093 ,065 ,197 ,005 -,012 -,040 ,009 -,033 ,065 ,055 ,044 -,076 -,026 -,007 ,060 -,025 -,123 ,025 ,026 ,019 ,384 ,455 ,391 ,430 ,408 ,467 ,509 ,070 ,062 ,061 ,060 ,059 ,072 ,078 ,047 ,054 ,050 ,051 ,052 ,050 ,049 ,070 ,072 ,066 ,063 ,063 ,065 ,059 Coefficientsa Standardized Coefficients Beta ,618 ,212 -,145 -,102 ,111 ,225 ,031 ,100 -,125 -,178 ,021 -,024 -,110 -,075 -,131 ,097 ,296 ,007 -,016 -,057 ,013 -,044 ,081 ,071 ,056 -,095 -,034 -,008 ,066 -,028 -,142 t Collinearity Statistics Tolerance VIF Sig 5,134 ,000 25,107 8,920 -7,147 -,908 ,836 1,978 ,251 ,840 -,918 -1,197 ,211 -,278 -1,310 -,912 -1,584 ,900 2,525 ,106 -,212 -,809 ,179 -,633 1,313 1,122 ,631 -1,065 -,399 -,105 ,939 -,378 -2,081 ,000 ,000 ,000 ,364 ,404 ,048 ,802 ,401 ,359 ,231 ,833 ,781 ,190 ,362 ,113 ,368 ,012 ,916 ,832 ,419 ,858 ,527 ,189 ,262 ,528 ,287 ,690 ,916 ,348 ,706 ,038 ,605 ,651 ,887 ,029 ,021 ,028 ,023 ,026 ,020 ,017 ,036 ,050 ,052 ,054 ,054 ,032 ,027 ,085 ,064 ,075 ,075 ,076 ,096 ,091 ,047 ,046 ,051 ,068 ,074 ,066 ,079 1,652 1,537 1,127 34,117 47,850 35,458 42,901 38,605 50,425 60,037 27,594 19,827 19,296 18,654 18,596 31,611 37,530 11,766 15,583 13,351 13,377 13,164 10,387 11,003 21,152 21,868 19,738 14,794 13,457 15,230 12,732 141 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies F Conjoint analysis This appendix provides background information and output files for the conjoint survey/analysis First appendix F1 provides an overview on effect coding and appendix F2 provides the choice-set output file constructed by Ngene This output file formed the basis for the choice-sets and attribute-levels as included in the conjoint survey Eventually, appendix F3 presents an overview of the Biogeme statistical tools and the input files F1 Effect coding Table 7.20: Conjoint analysis effect coding F2 Ngene Table 7.21: Ngene choice set Choice situation keuze1.kr keuze1.rwkeuze1.av keuze1.gr keuze1.b keuze2.kr keuze2.rwkeuze2.av keuze2.gr keuze2.b 0 0 0 1 0 0 0 1 0 1 0 0 1 0 142 0 0 1 10 11 12 0 1 1 0 1 1 1 1 1 0 1 1 2 0 0 1 1 0 1 1 1 1 0 1 2 2 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| F3 Biogeme Figure 7.3: Biogeme interface // File MODELRUNFINAL.mod [ModelDescription] "privacy assessment PAYD" [Choice] CHOICE [Beta] // Name Value asc1 asc2 beta1 beta2 beta3 beta4 beta5 beta6 LowerBound UpperBound -10000 10000 -10000 10000 -10000 10000 -10000 10000 -10000 10000 -10000 10000 -10000 10000 -10000 10000 status (0=variable, 1=fixed) 0 0 0 [Utilities] // Id Name Avail linear-in-parameter expression (beta1*x1 + beta2*x2 + ) keuze1 AV1 asc1 * CONST + beta1 * K1KR1 + beta2 * K1RW1 + beta3 * K1AV1 + beta4 * K1GR1 + beta5 * K1B1 + beta6 *K1B2 keuze2 AV2 asc1 * CONST + beta1 * K2KR1 + beta2 * K2RW1 + beta3 * K2AV1 + beta4 * K2GR1 + beta5 * K2B1 + beta6 *K2B2 keuze3 AV3 asc2 * CONST [Model] $MNL Figure 7.4: Biogeme MODfile 143 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Figure 7.5: Biogeme DATfile (example) 144 |SEPAM Master Thesis |December 2014 | S.A.J.P Derikx| F4 Output Biogeme output // This file has automatically been generated // 11/08/14 16:14:34 // Michel Bierlaire, EPFL biogeme 2.2 [Thu Mar 15 14:58:02 WEST 2012] Michel Bierlaire, EPFL privacy assessment PAYD Model: Number of estimated parameters: Number of observations: Number of individuals: Null log-likelihood: Cte log-likelihood: Init log-likelihood: Final log-likelihood: Likelihood ratio test: Rho-square: Adjusted rho-square: Final gradient norm: Diagnostic: Iterations: Run time: Variance-covariance: Sample file: Utility parameters ****************** Name Value Std err - asc1 -1.20 0.203 asc2 0.00 fixed-beta1 -0.288 0.159 beta2 -0.378 0.163 beta3 0.369 0.155 beta4 -0.351 0.166 beta5 -1.42 0.171 beta6 0.304 0.116 Multinomial Logit 660 660 -725.084 -608.270 -725.084 -530.289 389.589 0.269 0.259 +5.042e-003 Convergence reached 00:00 from analytical hessian RUNFINAL2.dat t-test p-val -5.93 0.00 -1.82 -2.31 2.38 -2.11 -8.31 2.62 0.07 0.02 0.02 0.03 0.00 0.01 Rob std err Rob t-test Rob p-val - -0.197 -6.11 0.00 * 0.158 0.165 0.152 0.164 0.178 0.117 -1.83 -2.28 2.42 -2.14 -7.99 2.61 0.07 0.02 0.02 0.03 0.00 0.01 * Utility functions ***************** keuze1 AV1 asc1 * CONST + beta1 * K1KR1 + beta2 * K1RW1 + beta3 * K1AV1 + beta4 * K1GR1 + beta5 * K1B1 + beta6 * K1B2 keuze2 AV2 asc1 * CONST + beta1 * K2KR1 + beta2 * K2RW1 + beta3 * K2AV1 + beta4 * K2GR1 + beta5 * K2B1 + beta6 * K2B2 keuze3 AV3 asc2 * CONST Correlation of coefficients *************************** Coeff1 Coeff2 Covariance Correlation beta2 beta4 0.00169 0.0624 beta3 beta6 -0.00140 -0.0779 beta1 beta4 0.0111 0.423 beta1 beta2 -0.000356 -0.0137 asc1 beta5 0.00542 0.156 asc1 beta2 -0.0162 -0.489 asc1 beta4 -0.0184 -0.545 beta1 beta3 -0.00101 -0.0411 asc1 beta1 -0.0154 -0.478 beta1 beta6 -0.000538 -0.0292 beta3 beta4 4.46e-005 0.00173 beta4 beta6 -0.000795 -0.0413 beta2 beta6 0.000695 0.0367 beta2 beta3 0.00653 0.258 beta2 beta5 0.00190 0.0681 beta4 beta5 0.00309 0.109 asc1 beta3 -0.0151 -0.479 beta1 beta5 0.00285 0.105 asc1 beta6 -0.00321 -0.136 beta5 beta6 -0.0148 -0.745 beta3 beta5 0.00201 0.0756 t-test 0.12 0.32 0.36 0.39 0.89 -2.61 -2.63 -2.90 -2.94 -2.97 3.17 -3.17 -3.46 -3.85 4.58 4.76 -5.09 5.13 -6.10 -6.42 8.07 * * * * * Rob covar 0.00319 -0.00208 0.0110 0.000440 0.00102 -0.0168 -0.0177 -0.000688 -0.0148 -0.00203 0.000161 -0.00191 -0.000884 0.00697 0.00522 0.00541 -0.0144 0.00513 -0.00129 -0.0158 0.00406 Rob correl -0.118 -0.117 0.426 0.0169 0.0292 -0.514 -0.548 -0.0286 -0.476 -0.110 0.00644 -0.100 -0.0458 0.276 0.177 0.186 -0.478 0.183 -0.0562 -0.763 0.150 Rob t-test -0.12 0.32 0.36 0.39 0.83 -2.62 -2.69 -2.95 -3.00 -2.87 3.23 -3.11 -3.30 -3.90 4.74 4.91 -5.22 5.27 -6.43 -6.22 8.28 * * * * * Smallest singular value of the hessian: 14.0193 Figure 7.6: Biogeme REPfile 145 Overcoming Privacy Concerns in the Consumer Use of Insurance Services based on Mobile Technologies Scientific Article Scientific article 146 147