Demographic Data
Based on previous studies, we test four demographic variables in our research: age, education level, average monthly income, and job occupation. Before further analysis, we first use descriptive statistics to describe the image of the whole sample, and for facilitating subsequent analysis and interpretation.
Table 4.1 is the descriptive statistics of all data arranged by demographic variables.
It gives a clear framework of the subject’s image. After one week’s online questionnaire distribution, 237 samples were collected. Roughly view of the table, it is easily told that the most replies are from 20~25 years old, bachelor degree students with limited income.
Although we try our best to reach different classifications people to sampling broadly, the samples do not seem comprehensive overall.
Also, in table 4.1, we show the differentiates between the subjects who finally chose yearly payment and those chose monthly payment. In terms of payment decisions, only 52 samples choose to make annual payments, while the number of people who chose monthly payments is about 3.5 times the former. The result is obviously inconsistent with our first hypothesis, and more detailed analysis relate to hypothesis 1 will be tested and presented in the next section.
Table 4.2 is the t-test of demographic data versus payment decisions. We try to figure out if there is any differentiation between the subjects in two payment decisions. In the four demographic variables, only average monthly income level has a statistically significant difference between the two payment decisions. But if we take back to check the detail number in Table 1, we can find that there is only 6% of subjects’ who have an average income level above 50k, and only one subject of them chose yearly payment. Given the small sample size, the significant result is not strong evidence and very likely is caused by the defect of the sample comprehensive.
Furthermore, other insignificant results imply that the three demographic variables do not influence the payment decision questions in this research. In other words, we don’t need to concern any potential influence those might from demographic differences.
Table 4.1.
Sample Descriptive Statistics
Table 4.2.
T-test of Payment Decision Versus demographic Data
To illustrate more, we run a binary logistic regression with payment decision versus demographic variables, the result is shown in table 4.3. There is no statistical significant differences between the subjects chose yearly payment and monthly payments. The results show that demographics variables can not be take to explain payment preference and won’t be a problem in the payment decision question.
Table 4.3.
Binary Logistic Regression – Payment Decision Versus Demographics
Preference of Payments Payment Preference Hypothesis Test
To more specific mention out how strongly people preferred integrate loss, the payment decision question which formed six options within monthly payment and yearly payment is recoding into 0 to 5, as 5 represent the most strong preference to choose yearly payment. Table 4 shows the distribution of integrate loss preference, with mean 1.32 it is obviously weak evidence to support our first hypothesis.
H1: Consumers should prefer to pay by years than pay by months, as a smaller but more frequent loss payment.
Table 4.4.
Descriptive statistic – integrate loss preference
It is easy to say that we should reject H1. As table 4 shows, 50.6% of subjects had no doubt to choose the monthly payment when there was only 13.1 people chose yearly payment surely. This inconsistent with the hypothesis is beyond our original suppose. We try to find some possible explanations for this result, but the result may imply an uncompleted frame of the questionnaire. Further discussion will include in Chapter 5.
Table 4.5.
T-test of Integrate Loss Preference Versus Payment Decision
To illustrate more, a t-test with payment decision versus yearly payment preference is shown on table 5. The test showed two groups of payment decision are statistically significant different on intergrate loss preference.
Discount Rate
From Kirby’s research, we give each subject a k value to represent their intertemporal preference which is counted from the answers of 27 alternative questions. To fulfill homogeneity test, we recoding k into ln(k*1000) and also classified them into ten levels which are suggested from Kirby.
Discount Rate Descriptive Statistics Table 4.6.
Descriptive statistic - Discount Rate
Table 4.6 is the descriptive statistic of whole data’s discount rate. Following the former research, we can give each subjects one k value to represent their discount rate, and also we can sign them into ten scaled discount rate level.
Discount Rate Hypothesis Test
For the purpose of testing if there exist moderator effect from discount rate, we further recording subjects those who have discount rate rank under 3 as low and those above 7 as high.
H3: People who have a lower Intertemporal discount ratio will be less willing to pay by years than pay by months.
To test H3, we first run ANOVA of payment decision versus discount rate 3 classes and it shows a statistic significant result. That is, we can say there are difference between three classes of discount rate in integrate loss preference.
Table 4.7.
ANOVA – Integrate Loss Preference versus Discount Rate
Second, we run binary logistic regression of 10 ranks discount rate versus payment decision. As table 8 shows, the result support H3 that discount rate can somehow explained the payment decision, and discount rate showed the explanation 0.314 in the regression.
Table 4.8.
Binary Logistic Regression – Payment Decision Versus Discount Rate
Finally, we provide the evidence to support H3 that discount rate can somehow explained the payment decision, and discount rate with medium bonus showed the highest explanation 2.565 in the regression.
To further understand the distribution of discount rate and yearly payment preference, we use χ-test to show the distributions. The original discount rate data is in form of k-value and related level from zero to ten, for the data present needs, we recode ten levels into five levels. In the vertical site is 5 level discount rate, and the horizon site is 6 level of yearly payment preference. Although we got a significant results here, the distribution is not strictly follow any rules or show any trend.
Table 4.9.
χ-test of Preference of Yearly payment versus Discount Rate
See table 4.9, the percentage under n. column first, the percentage is show the preference of yearly payment within five discount level. In yearly payment preference 0, we can find a obviously rather high percentage compared to total data in discount rate level
2, and a rather low percentage in level 4. Then take a look in the column 5 of yearly payment preference, compared to total data, it can also find a obviously rather low percentage in discount rate level 2 and a rather high percentage in level 4. Those numbers support our hypothesis that lower discount rate people will less preferred yearly payment.
Second, the percentage in the right of n. column is the distribution within yearly payment preference in the specific discount level. The distribution within yearly payment preference is no big different in different discount rate level.
Product familiarity Product Familiarity Descriptive Statistics
Following the former research, each subject got their familiarity score from their answer, the score is between 0 to 40, and the mean of data is 20.3 and the median is 21. To take a closer look, whole data was recoding into 4 levels. Each level contains ten scores, and the distribution is shown in Table 10. The data that distincted by two payments decision is also presented in table 10.
Table 4.10.
Descriptive Statistic – Familiarity Score
Product Familiarity t-test
Table 11 is the T-test of payment decision and familiarity score, unfortunately it shows insignificant differentiation between two groups.
Table 4.11.
T-test of Payment Decision versus Familiarity Score
To further illustrate, we also run t-test on each question in familiarity questions, and we find significant only in question five and question eight.
Familiarity question 5: When speak to streaming, you will think yourself is?
a.Totally unfamiliar b.Unfamiliar c.Ordinary d.Familiar e. Specially familiar Familiarity question 8: Please do your best to list at most 4 brands of streaming services.
Table 4.12.
T-test of Payment Decision versus Familiarity Score Question
Product Familiarity Hypothesis Test
H2: People who have higher product familiarity will be more clear about their willingness to make the decision between pay by years or pay by months, then those have lower product familiarity.
Before testing the H2 we first test correlations between preference intensity and familiarity score. Preference intensity is the data recoding from payment decision question. In the 6 options which are chosen between monthly and yearly payment, the two options that most near two extremes are recoding into 3, and following two are 2, and the last two options in the middle represent the weakest intensity are recode into 1.
Table 4.13.
Correlation of Preference Intensity and Familiarity Score
Based on the Pearson correlation analysis, preference intensity and familiarity score have positive relationship [r (235) =0.110, p < 0.05]. P-value under 0.05 means there are correlations between familiarity scores and preference intensity.
Table 4.14.
Multinomial Logistic Regression - Preference Intensity and Familiarity Score χ-test - Preference Intensity and Familiarity Score
To test H2, we classified whole subjects into three levels of familiarity scores, and run multinomial logistic regressions. As table 4.14 shows, the results are not statistically significant. The result reject H2 that implies familiarity scores can not somehow explain the distribution of preference intensity.
Table 4.15.
χ-test - Preference Intensity and Familiarity Score
Finally, we try χ-test to make sure there is difference distribution between familiarity scores and preference intensity. Unfortunately, the result on table x still insignificant.