1. Factor analysis
Factor analysis is sometimes called a data reduction technique. This method is frequently used to extract a few underlying components from a large initial set of observed variables and allow us to investigate whether there is an underlying structure. In this study, this method can extract important factors from product variables and reduce variables.
2. Reliability analysis
After factor analysis, reliability analysis is a way to test the reliability of questionnaires. To achieve that, Cronbach’s Alpha is the way to measure the internal consistency. In this study, Cronbach’s Alpha is used to measure the internal consistency between each variable and the whole variables.
3. Cluster analysis
Cluster analysis divides data into two or more mutually exclusive unknown groups that are meaningful, useful based on combinations of interval variables. It is used to discover a system of organizing observations, usually people, into groups, where members of the groups share properties in common.
4. ANOVA
ANOVA is a general technique that is widely used to test for differences experimental designs involving more than two groups or more than one independent variable. In this study, ANOVA is used to test the differences among each cluster in terms of population variables.
5. Discriminant analysis
is the variable that a researcher tries to explain or predict from the values of independent variables. It is close related to analysis of variance (ANOVA) and regression analysis. The difference between discriminant analysis and other two methods is with regard to the nature of dependent variable, which is a categorical variable.
CHAPTER4
DATA ANALYSIS
4.1
E
XTRACTING FACTORS4.1.1 Factor analysis
In this study, 26 product variables were designed in a 4-point Likert scale (1= very unimportant, 2= unimportant, 3= important, 4= very important). Through the KMO and Bartlett’s test, the data shows that it is suitable for factor analysis. As a measure of factorability, a KMO value of 0.5 is poor; 0.6 is acceptable; a value closer to 1 is better.
Table 4-1 KMO and Bartlett's Test
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. 0.736
Bartlett's Test of Sphericity
Approx. Chi-Square 1404.818
df 325
Sig. 0.000
The principle component analysis was used to extract factors in the study with the principle that the eigenvalues of components are greater than 1 and that the factor loadings are greater than 0.4. According to Hair, Anderson, Tatham and Black (1998) guideline for practical significance, a factor loading of ±0.3 indicates that the item is minimally significant; ±0.4 means that the item is more important; ±0.5 suggests that the factor is significant. Therefore, 7 factors are extracted. Furthermore, Varimax rotation method, the most commonly used orthogonal rotation, was also applied to maximize the factor loadings of each variable and to minimize the loading on other factors, so that the representing meaning of each factor is more apparent.
4.1.2 Internal consistency reliability
Items should all be fairly strongly correlated with each other if they are intended to measure aspects of the same construct. Cronbach’s Alpha, also called coefficient Alpha, is the most
commonly reported measure for assessing this. A scale should have a minimum Cronbach’s Alpha value of 0.7. In this study, the Cronbach’s Alpha value is 0.81, indicating that the overall reliability of the scale is fairly good.
The right-most column shows the values for Cronbach’s Alpha if item deleted. Since none of the items helps raise the Cronbach’s Alpha value over 0.81 if deleted. All of the items should be kept.
Table 4-2 Result of factor analysis- 1
Component Cronbach's Alpha if Item Deleted 1 2 3 4 5 6 7 1. CPU 0.295 0.523 0.007 -0.120 0.075 0.384 -0.009 0.801 2. Memory size 0.083 0.441 0.144 0.026 -0.014 0.352 -0.420 0.808 3. ISO 0.860 0.183 0.015 0.028 0.070 -0.020 -0.048 0.801 4. Camera pixel 0.860 0.076 0.087 0.071 -0.005 0.144 -0.095 0.801 5. Camera sensor 0.804 0.240 -0.023 0.091 0.101 0.107 0.035 0.797 6. Memory card 0.165 0.012 0.084 -0.063 -0.018 0.788 0.114 0.807 7. Compatibility 0.268 0.570 0.029 -0.086 0.076 0.189 -0.029 0.803 8. Screen size 0.609 0.239 0.015 -0.064 -0.163 0.089 -0.019 0.807 9. Multi-media 0.305 0.405 0.027 -0.069 0.172 0.508 0.227 0.796 10. Bluetooth -0.009 0.001 0.045 0.052 0.084 0.753 -0.019 0.809 11. Digital currency 0.169 0.510 0.036 0.121 0.012 -0.042 0.209 0.804 12. Apps 0.057 0.859 0.019 0.096 0.011 -0.124 -0.011 0.803 13. Other service 0.158 0.732 -0.128 0.162 0.067 0.031 -0.018 0.801 14. Advertisement 0.031 0.141 0.019 0.012 0.566 0.008 0.392 0.805 15. Life style 0.034 0.068 -0.059 0.463 0.473 -0.020 -0.006 0.807 16. Friends/ family 0.006 0.162 0.032 0.055 0.844 0.061 0.166 0.801 17. Brand familiarity -0.028 -0.112 0.076 0.160 0.785 0.102 -0.130 0.809 18. Appearance 0.069 0.129 0.165 0.685 0.209 -0.063 -0.097 0.804 19. Color selection 0.092 0.056 0.367 0.469 -0.006 0.073 0.386 0.801 20. Touch 0.068 0.030 -0.113 0.760 0.089 0.045 0.262 0.805 21. Material -0.085 0.022 0.281 0.802 -0.017 -0.051 0.047 0.806 22. Price of the phone -0.064 -0.107 0.872 0.090 -0.022 0.048 0.032 0.809 23. Cost/ price ratio 0.098 0.056 0.714 0.092 -0.072 0.230 0.134 0.804 24. Sales promotion 0.062 0.043 0.823 0.103 0.203 -0.087 0.114 0.803 25.Price of peripherals -0.107 -0.022 0.274 0.228 0.124 0.042 0.657 0.807 26. Price of app -0.046 0.083 0.086 0.026 0.056 0.124 0.773 0.809
The product variables of each factor are showed in following table. Generally, the internal reliability level should be higher than 0.6. If it is lower than 0.6, then, remodifying will be necessary. The right-most column shows the values for Cronbach’s Alpha if item is deleted. After modifying, each factor was named.
Table 4-3 Result of factor analysis- 2
Factor 1: Camera
(Cronbach’s Alpha= 0.837) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 4 The number of pixels of build-in camera 0.860
4.820 11.147%
0.758 3 The ISO performance of build-in camera 0.860 0.753 5 The sensor of build-in camera 0.804 0.782 8 The size of the phone screen 0.609 0.861
Factor 2: Integration of hardware and software (Cronbach’s Alpha= 0.738)
Factor
loading Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 12 The quality/ quantity of applications 0.859
3.462 10.590%
0.656 13 Other services (e.g. icloud, itunes) 0.732 0.673 7 Compatibility with computer 0.570 0.701 1 The core number and speed of the CPU 0.523 0.700 11 Supporting digital currency 0.510 0.737 2 The size of Build-in memory 0.441 0.729
Factor 3: price of the phone
(Cronbach’s Alpha= 0.776) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 22 The phone price 0.872
2.106 9.042%
0.616 24 Sales promotion event 0.823 0.684 23 The cost/ performance ratio 0.714 0.778
Factor 4: Design
(Cronbach’s Alpha= 0.721) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 21 The material of outward appearance 0.802
1.805 8.979%
0.567 20 The touch of outward appearance 0.760 0.668
19 The variety of color selection 0.469 0.718 Factor 5: Brand image
(Cronbach’s Alpha= 0.677) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 16 Friends/family are using the same brand 0.844
1.456 8.026%
0.476 17 The familiarity of the brand 0.785 0.606 14 The frequency of advertisement showed 0.566 0.658 15 Brand image that reflect personallifestyle 0.473 0.683
Factor 6: File transfer and display
(Cronbach’s Alpha= 0.634) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 6 Memory card expansion 0.788
1.312 7.390%
0.351 10 Bluetooth file sharing 0.753 0.634 9 Supporting multi-media formats 0.508 0.584
Factor 7: Price of additional purchase
(Cronbach’s Alpha= 0.617) loadingFactor Eigenvalues explainedVariance
Cronbach's Alpha if
Item Deleted 26 The price of application 0.773
1.172 6.875% - 25 The price of peripheral device 0.657 -
The Cronbach’s Alpha for factor 1 is 0.837. Since deleting the variable “8.The size of the cell phone screen” did not raise the Cronbach’s Alpha obviously, the variable was retained. This factor was composed of four variables that are all related to camera function. Thus, factor 1 was named as “Camera.”
The Cronbach’s Alpha for factor 2 is 0.738. The Cronbach’s Alpha value did not become greater if any of the variables were deleted, which indicated that these variables are rather representative. This factor was mainly composed of hardware specs of a smartphone such as CPU speed, build-in memory size, software such as phone applications, software compatibility, other service, and digital currency. Therefore, factor 2 was named as “Integration of hardware and software.”
The Cronbach’s Alpha for factor 3 is 0.776. Although the Cronbach’s Alpha can be increased by deleting the variable “the cost/ performance ratio”, this variable should not be eliminated since there was no apparent increase in Cronbach’s Alpha. This factor was formed by variable like phone price, Sales promotion event, and the cost/performance ratio. Accordingly, factor 3 was named as “Price.”
The Cronbach’s Alpha for factor 4 is 0.721. The Cronbach’s Alpha value did not increase if any of the variables was deleted, which suggested that these variables were relatively representative. This factor was mainly formed of the outward appearance. Consequently, factor 4 was named as “Design.”
The Cronbach’s Alpha for factor 5 is 0.677. The Cronbach’s Alpha value did not rise if any of the variables were deleted, which meant that these variables were respectably representative. This factor was mainly composed of variables such as brand familiarity and lifestyle related to brand image. Consequently, factor 5 was named as “Brand image.”
The Cronbach’s Alpha for factor 6 is 0.634. The Cronbach’s Alpha value did not increase if any of the variables was deleted, which showed that these variables were rather representative. This factor was mainly composed of variables such as memory card expansion and Bluetooth file sharing, which are related to file transfer, and supporting multi-media formats, which is related to file display. Hence, factor 6 was named as “File transfer and display.”
The Cronbach’s Alpha for factor 7 is 0.617. The factor loadings of the variable in this factor are relatively large. However, the Cronbach’s Alpha cannot be obtained if either one of the two variables was removed. Also, from the table, we can conclude that the overall reliability will not enhance if either one of the two variables was deleted. This factor was composed of the price of applications and peripheral devices, which can be referred to the cost of purchase activity after buying a phone. Therefore, the factor 7 was named as “Price of additional purchase.”
4.1.3 Validity
The questionnaire in this study was founded and corresponds to the real world based on related literature and on the characteristics of smartphones to represent the content domain of smartphone consumer behavior. The details of design of each question can refer to 3.1.2 variables. Before the formal questionnaire was distributed, the modification of each item was according to the pre-test questionnaires and based on the discussion with the thesis advisor. Therefore, the questionnaire possesses content validity.
4.2
C
LUSTER ANALYSISCluster analysis was conducted according to the seven factors we obtained from the factor analysis to explore factors that are considerable to each cluster.
In this study, K-means clustering was used to divide sample into 2, 3, and 4 groups. Nevertheless, when considering the number of cases in each group, both 3 groups and 4 groups had a very few number of cases in a certain group. Also, taking significant level into account, we found that dividing into 2 groups was the most appropriate approach.
The result is shown as below. Factor 3 price and factor 7 price of additional purchase have no significant difference between two clusters (sig. >0.05) and will not be discussed further.
Table 4-4 Result of cluster analysis- 1
Factor Sig. 1 Cluster 2 1. Camera 0.00 0.32121 -0.25398
2. Integration of hardware and software 0.00 -0.49181 0.38887 3. Price of the phone 0.995 -0.00054 0.00043 4. Design 0.00 -0.31748 0.25103 5. Brand image 0.00 -0.46675 0.36906 6. File transfer and display 0.00 -0.39943 0.31583 7. Price of additional purchase 0.279 0.09847 -0.07786
Cluster 1 has an average camera factor percentage that is higher than cluster 2. Thus, this cluster was named as “Camera group”. On the other hand, cluster 2 has higher average values for integration of hardware and software, design, brand image, and file transfer and display. Since integration of hardware and software and file transfer and display are related to cell phone performance and, also, design and brand image are related to outer image that consumers care about, cluster 2 was named as “Performance and outer image group”.
Table 4-5 Result of cluster analysis- 2
Cluster Factor (highest value) Factor (lowest value) 1.
Camera group
1. Camera performance
2. Integration of hardware and software 4. Design
5. Brand image
6. File transfer and display 2.
Performance and outer image group
2. Integration of hardware and software 4. Design
5. Brand image
6. File transfer and display
1. Camera performance
4.3
ANOVA
After dividing the respondents into groups via cluster analysis, ANVOA was used to analyze the population statistic variables in these clusters to have further understanding on the characteristics that these clusters carry.
ANOVA was used to analyze whether there is a significant difference between the two groups in population statistic variables, including gender, age, disposable income, and vocation. In this study, there is a significant difference with the Sig. value 0.03(<0.05) in gender between the camera group and performance and outer image group. However, there is no significant difference in age, disposable income, and vocation between the two groups.
4.4
D
ISCRIMINANT ANALYSISDiscriminant analysis is used to predict and explain people’s purchase decision on smartphone from predict variable, namely camera, integration of hardware and software, price of the phone, design, brand image, file transfer and display, and price of additional purchase factors that are derived from factor analysis. After repeatedly testing, we found that the function is statistically significant only in the case of Apple.
In discriminant analysis, the appropriateness of applying pooled covariance matrix in the classification phase is evaluated by the Box’s statistic. In this case, Box’s M statistic is statistically significant (Sig. < 0.05), which suggests that dispersion of two groups (Apple and other venders) is not homogeneous. The test was re-run using separate covariance matrices in classification.
Table 4-6 Box’s M statistic
Box’s M 45.220 F Approx. 1.519
df1 28
Df2 33131.506 Sig. 0.039
The canonical correlation is a correlation between the discriminant scores and the levels of the dependent variable. The present correlation of 0.328 is not extremely high.
Table 4-7 Eigenvalues
Function Eigenvalue % of Variance Cumulative % CorrelationCanonical 1 0.120 100.0 100.0 0.328
The Lambda of 0.893 has a significant value (Sig. = 0.018< 0.05); thus, the function is statistically significant. The factors of camera, integration of hardware and software, price of the
phone, design, brand image, file transfer and display, and price of additional purchase can effectively discriminate between Apple and other venders.
Table 4-8 Wilks’ Lambda
Test of
Function(s) Wilks' Lambda Chi-square df Sig.
1 0.893 16.849 7 0.018
The standardized canonical discriminant function coefficients show the importance of each factor. The greater the value is, the more important the factor is. According to the table below, design, integration of hardware and software, and file transfer and display are highly contributed to the discriminant function.
Table 4-9 Standardized canonical discriminant function coefficients
Camera 0.231
Integration of hardware and software 0.524 Price of the phone -0.312
Design 0.582
Brand image 0.018 File transfer and display -0.445 Price of additional purchase -0.363
In structure matrix, the greater the absolute value of a coefficient is, the greater the effectiveness is. In this case, design, integration of hardware and software, and File transfer and display are highly effective to the discriminant function.
Table 4-10 Structure matrix
Design 0.537
Integration of hardware and software 0.481 File transfer and display -0.405 Price of additional purchase -0.328 Price of the phone -0.281
Camera 0.207
Brand image 0.016
The classification results table is a summary of number and percent of subject classified correctly and incorrectly. 63 percent of original grouped cases were correctly classified, which indicates that the discrimination of these factors is good in certain level.
Table 4-11 Classification results
Predicted group membership
Total Other venders Apple
Original
Count Other venders 68 37 105
Apple 19 30 49
% Other venders 64.8 35.2 100.0 Apple 38.8 61.2 100.0
CHAPTER5
DISCUSSION
5.1
D
ISCUSSIONSThe major purpose of this study is to explore smartphone consumer behavior by finding the determents of smartphone purchase. The result shows that product performance, branding, product design, and price have the influence on people’s buying decision process, which echoes to the literature that whether the product can satisfy people’s needs, branding, appearance, and price can affect consumer behavior.
According to consumer behavior models, whether the product is able to satisfy people needs is one of the arguments that influence people’s buying decision. In the case of smartphone, the product performance, including integration of hardware and software, file transfer and display, and camera performance, is considered as the influence that defines whether one’s major needs could be satisfied; as what literature implies that brand name has the effect on certain level to the cognition of product quality, in this study, brand image of a smartphone vendor affects people’s purchase decision. However, the influence is rather small when comparing with other factor; appearance influences consumers’ behavior easily. Here, appearance is the determent that influences buying decision the most according to discriminant analysis, and is defined as the material, the touch, and the color selection of a smartphone; price has been pointed out that it can be treated as the signal that represents product quality, but, meanwhile, it also has negative correlation with needs. In the case of smartphone, the result suggests that the influence of representing product quality is relatively weak, and that people consider that price serves as an inverse indicator in terms of buying decision. Price is namely the price of a smartphone and the price of additional purchase in the study.
The main contributions of the thesis is to deconstruct existing theories of consumer behavior, proposing an alternative interpretation of consumer behavior in smartphone purchase in which major factors that affect smartphone buying were found through analysis. Findings are able to aid vendors
The factors in this study were defined strictly based on the literature. Hence, it is difficult to cover the whole scope from a certain dimension. For example, in this study branding is defined as the experience and the frequency of advertising that consumers perceived. Yet, the importance of whether one can shape his/her ideal self-image through brand image was ignored. This kind of questions could change the result enormously. On the other hand, the number of samples required to directly represent the distribution may be prohibitively large. As accidental sampling was applied in this study and the sample size is small, the generalizations about the total population cannot be made scientifically. In fact, most of the respondents of the sample are belong to the Y generation. Since external influences, such as culture, social, and personal factors, affect consumer behavior, the result is not unbiased. Also, this study is based on a one-time survey. A longitudinal study to show the measurement of motivation for better reliability is recommended.
5.2
C
ONCLUSIONSFactor analysis extracted factors from product variables such as product performance, branding, product design, and price. 3 factors were extracted from product performance and were named as camera factor, integration of hardware and software factor, and file transfer and display factor; 1 factor was extracted from branding and was named as branding image factor; 1 factor was extracted from product design and was named as design factor; 2 factors were extracted from price and were named as price of the phone factor and price of additional purchase factor.
In the aspect of market segmentation, smartphone market was divided into two segments via cluster analysis. According to the common characteristics of each segment, the first segment was named as camera group in which people consider camera performance is comparatively important. The second segment was named as performance and outer image group in which people consider smartphone performance, such as integration of hardware and software and file transfer and display, and the outer image that a smartphone carries, such as design and brand image, are comparatively important. ANOVA suggests that there is a significant difference in gender between two segments,
indicating that most females are belong to camera group while most males are belong to performance and outer image group.
Regarding purchase decision, the discriminant function is statistically significant only in the case of Apple and, in order of effectiveness, design, integration of hardware and software, file transfer and display, price of additional purchase, price of the phone, camera, and brand image factors affect people’s smartphone buying decision. Such a result implies that, when buying a smartphone, a consumer who considers that product design and integration of hardware and software are important and that the convenience of transferring file or media display are not important would choose Apple’s iPhone. In contrast, factors in this study seem to be hard to discriminate each smartphone venders from others, which implicates that the differentiation of other smartphones is not that strong.
5.3
I
MPLICATIONS5.3.1 Implications for business people
Figure 5-1 shows the relationship between actual prices and estimated prices of 28 smartphones from Apple, hTC, Samsung, and Sony. The estimated prices were calculated through regression analysis based on smartphone hardware specification namely CPU speed, number of CPU core, build-in memory size, screen size, battery capacity, and the number of camera pixels. Theoretically, a smartphone’s actual price and estimated price should be the same when its price is