The HBAT company case study: E-Commerce Activities, Customer Satisfaction, Likelihood of Recommendation and Likelihood of Future Purchase affect the Consideration of Partnership in the future
A submitted midterm report of Quantitative method class
Presented by Daranrat Jaitiang, P10522015 Department of Tropical Agriculture International Corporation Beny wahyudi, G10582009 Department of International Master’s Degree Program in Agribusiness Management Instructor
National Pingtung University of Science and Technology
The HBAT company case study: E-Commerce Activities, Customer Satisfaction, Likelihood of Recommendation and Likelihood of Future Purchase affect the Consideration of Partnership in the future Daranrat Jaitiang, P10522015 and Beny Wahyudi, G10582009
Abstract HBAT company is a premium manufacturer in United State, specialize in paper production. Additionally, the paper product is widely well-known and used in the new print industry and the magazine industry both inside and outside North America.
Introduction
Objective To identify the independent variables that impact the To determine group independent variable affecting the consideration of partnership in the future
Literature review This study looks over the HBAT consideration of partnership or alliance in the future. Consideration of Strategic Alliance/ Partnership in Future Strategic partnership consideration was being increased in American firm () Most industries pay attention to this. E-Commerce Activities
Customer Satisfaction
Likelihood of Recommendation
Likelihood of Future Purchase
Methodology This study used logistic regression with a binary dependent variable, which can explain and predict a two-group categorical viable (Pearson, 2014). Database The data was retrieved from HBAT Case Data, which included the dependent variables and the independent variables (Table1)
Model Perc = α0 + α1ECom + α2CSat + α3Rec + α4FPur + e Where: Perc ECom CSat Rec FPur
= Perception of future relationship with HBAT = E-Commerce Activities = Satisfaction = Likelihood of Recommendation = Likelihood of Future Purchase
Sample size Based on the HBAT data, data was consisted of 200 observation and 18 separate variables referring from the market segmentation study from HBAT of customer.
Hypothesis Hypothesis Model H0 = Independent variables are simultaniosly affecting to dependent variable H1 = Independent variables are not simultaniosly affecting to dependent variable Hypothesis Variable a.
E-Commerce Activities/Web Site H0 = E-commerce activities are not affeting to consider strateguc alliance H1 = E-commerce activities are affeting to consider strateguc alliance
b.
Satisfaction H0 = Satisfaction is not affeting to consider strateguc alliance
H1 = satisfaction is affeting to consider strateguc alliance c.
Likelihood of Recommendation H0 = Likelihood of recommendation is not affeting to consider strateguc alliance H1 = Likelihood of recommendation is affeting to consider strateguc alliance
d.
Likelihood of Futur Purchase H0 = Likelihood of Future Recommendation is not affeting to consider strateguc alliance H1 = Likelihood of Future Recommendation is affeting to consider strateguc alliance
Result and Discussion 1. Distribution Data a. Consider Strategic Alliance/ Partnership in Future Tabel 1. Distribution Data of Consider Strategic Alliance/ Partnership in Future Consider Strategic Alliance/ Partnership in Future (Perc) 0 1 Total
Frequency
Percent
Cumulatif
114 86 200
57.00 43.00 100.00
57.00 100.00
Based on table -------, frequency of 1 (could consider) and 0 (not consider) is not too lame. It means that data of strategic alliance partnership in the future is normal distribution. In other word, this data is spread evenly to 1 (firm could consider) and 0 (firm could not consider). So this data can use to analyze with logistics regretion. b. E-Commerce Activities
.8 .6 .4 0
.2
Density
2
3
4 E-commerce Activities/Website
5
6
.2 0
.1
Density
.3
.4
c. Customer Satisfaction
5
6
7 Satisfaction
d. Likelihood of Recommendation
8
9
10
.4 .3 .2 0
.1
Density
4
6 8 Likelihood of recommendation
10
0
.2
Density
.4
.6
e. Likelihood of Future Purchase
4
Variable E-Commerce Activities (Ecom) Customer Satisfaction (CSat) Likelihood of Recommendation (Rec) Likelihood of Future Purchase (FPur)
6 8 Likelihood of future purchase
10
Obs 200
Pr (Skewness) 0.0057
Pr (Kurtosis) 0.7146
Adj chi2 7.29
Prob > chi2 0.0261
200
0.5960
0.0004
11.19
0.0037
200
0.6773
0.5415
0.55
0.7590
200
0.2255
0.1188
3.94
0.1392
To dicede whether data of variables are normal distribution, we can look from skewness and kurtosis test. If probability chi2 of skewness kustosis is more than 0.005
2. Data Summary Variable
Obs
Mean
Std. Dev.
Min
Max
Consider Strategic Alliance/ Partnership in Future (Perc)
200
0.43
0.496318
0
1
E-Commerce Activities (Ecom) Customer Satisfaction (CSat) Likelihood of Recommendation (Rec) Likelihood of Future Purchase (FPur)
200
3.765
0.768916
2.2
5.7
200
6.952
1.241128
4.7
9.9
200
6.9525
1.082893
4
9.9
200
7.665
0.893233
4.3
9.9
From the summary table, number of observation is 200 firms. We use Perc as dependent variable. Perc is dummy variable with value 0 and 1, so minimum value of Perc is 0, and maximum value of Perc is 1. To predict Perc, we use some independent variables, there are Ecom, CSat, Rec, FPur. All of independent variables were measured on a grapich scale 0 (poor value) to 10 (excellent value). 3. Correlation
Consider Strategic Alliance/ Partnership in Future (Perc) E-Commerce Activities (Ecom) Customer Satisfaction (CSat) Likelihood of Recommendation (Rec) Likelihood of Future Purchase (FPur)
Consider Strategic Alliance/ Partnership in Future (Perc) 1.0000
ECommerce Activities (Ecom)
0.3069
1.0000
0.6928
0.3416
1.0000
0.6815
0.3040
0.7616
1.0000
0.2352
0.7258
0.6609
0.5861
Customer Satisfaction (CSat)
Likelihood of Recommendation (Rec)
Likelihood of Future Purchase (FPur)
1.0000
Based on matrix correlation table, there are not independent variables which have multicollinearity. Because there are not value of correlations more than or equal 0.8 between
independent variables (Gujarati, 2001). So all of independents variable can used to predict dependent variable 4. Logistic Regression Log Likelihood = -57.756677
Number of obs LR chi2 (4) Prob > chi2 Pseudo R2
Perc
Coef.
Std. Err.
z
P>z
Ecom CSat Rec FPur _cons
.6642076 .9403997 1.697149 1.045687 -29.59466
.3935321 .342572 .4214644 .4526921 4.672458
1.69 2.75 4.03 2.31 -6.33
0.091 0.006 0.000 0.021 0.000
= = = =
200 157.81 0.0000 0.5774 [95% conf. Interval]
-.1071012 .2689708 .8710938 .1584269 -38.75251
1.435516 1.611829 2.523204 1.932947 -20.43681
Table of coefficient logistic show that model in this study is statistically significant, because value of Prob > chi2 = 0.0000 or less than significant level at 0.01. In other word, rejecting H0 and accepting H1, so there are independent variables affecting to dependent variable. This table also show that independent variables can explain 57% dependent variable (Pseudo R2), and 43% of dependent variable is explained by other factors outside of model. All of the independent variables significantly affect to dependent variable, because the P > z of all of independent variables are less than 0.1 (significant level). The coefficient of independent variables have sign positive, it means that increase in the each independent variable is associated with an increase the firm probability to consider strategy alliance/partnership with HBAT. Coefficient value of constanta is negative, it is mean that if all of variable are zero (or firms are not consider all independent variables), so the firm probability to consider strategy alliance with HBAT is decreasing.
Log Likelihood = -57.756677
Number of obs LR chi2 (4) Prob > chi2 Pseudo R2
Perc
Odds Ratio
Std. Err.
z
P>z
Ecom CSat Rec FPur _cons
1.94295 2.561005 5.458363 2.845353 1.40e-13
.7646133 .8773286 2.300505 1.288069 6.56e-13
1.69 2.75 4.03 2.31 -6.33
0.091 0.006 0.000 0.021 0.000
= = = =
200 157.81 0.0000 0.5774 [95% conf. Interval]
.8984348 1.308617 2.389523 1.171666 1.48e-17
4.201814 5.011967 12.46848 6.909845 1.33e-09
Table of odd ratio show that odds ratio value of Ecom is 1.94295. This is mean that if Ecommerce activities from HBAT increase 1 scale, so the probability of firm to consider to strategy alliance or partnership increased 94% (1.94295 – 1 = .94295). The odds ratio of CSat is 2.561005, it is mean that if customer satisfaction increase 1 scale, so the probability of firm to consider to strategy alliance increase 156%. Odds ratio of Rec is 5.458363, it means that if likelihood of recommending HBAT increase 1 scale, so the probability of firm to consider to strategy alliance with HBAT increase 445%. Odds ratio of FPur is 2.845353, it means that if likelihood of future purchase from HBAT increase 1 scale, so the probability of firms to consider to strategy alliance increase 184%.
Marginal effects after logit y = Pr (Perc) (predict) = .3226457 Variable
dy/dx
Std. Err.
z
P>z
Ecom CSat Rec FPur
.1451595 .2055201 .3709041 .2285301
.08779 .07371 .09204 .09844
1.65 2.79 4.03 2.32
0.098 0.005 0.000 0.020
[95% C.I.] -.026912 .06106 .190515 .035594
.317232 .34998 551293 .421466
X 3.765 6.952 6.9525 7.665
From marginal effect table, we can look marginal effect from each independent variables to dependent variable in mean value of independent variables. From that table, we know that independent variable which has highest marginal effect is likelihood of recommending HBAT (Rec). Value of marginal effect from Rec is .3709041, it means that if likelihood of recommending HBAT increase 1 scale from it’s mean value, so the probability of firm to consider strategy alliance with HBAT increase 37%. Marginal effect of E-Commerce activities is .1451595, it means that if E-Commerce activities increase 1 scale from its mean value, so the probability of firm to consider strategy alliance with HBAT increase 14%. Marginal effect of CSat is 0.2055201, it means that if customer satisfaction increase 1 scale from its mean value, so the probability of firm to consider strategy alliance increase 20%. Marginal effect of FPur is .2285310, it means that if likelihood of future purchase from HBAT increase 1 scale from its mean value, so the probability of firm to consider strategy alliance with HBAT increase 22%.
5. Hypothesis Test To test our Hypothesis, we use goodness of fit, wald, and likelihood-ratio test. The result of three tests are:
Prob > chi2
Goodness of Fit
Wald Test
Likelihood-Ratio Test
1.0000
0.0000
0.0035
The first, we use goodness of fit to Hypothesis test (Hosmer and Lemeshow). The purpose of this test to know whether model that we use is in accordance with empirical data. The Hypothesis null of this test is “model fits the data well”. Based on table above, the probability chi2 of goodness of fit is 1.00. It means that rejected H1 and accepted H0, in other words that model fits the data well. The second test we use Wald test. The Wald test approximates the lr test, but with the advantage that it only requires estimating one model. The Wald test works by testing the null hypothesis that a set of parameters is equal to some value. In this study, the null hypothesis is that the 4 coefficients of independent variables are simultaneously equal to zero. Based on table above, value of prob > chi2 is 0.0000 less than 0.05. it means that rejected H0 and accepted H1, so coefficients of independent variables are not simultaneously equal to zero. In other word, that including Ecom, CSat, Rec, FPur results in a statistically significant improvement in the fit of the model. The third test we use Likelihood-Ratio Test. The LR test is performed by estimating two models and comparing the fit of one model to the fit of the other. The first model we use “Perc = α0 + α1Ecom + α2Rec + α3FPur + e (omit CSat)”. And second model we use “Perc = α0 + α1Ecom + α2CSat + α3Rec + α4FPur + e (complete model)”. The hypothesis null of this test is “smaller model is true model”. With probability chi2 of LR test is 0.0035 less than significant level (0.005), so rejected H0, and accepted H1, indicating that the model with 4 predictors (complete model) fits significantly better than model only 3 predictors.
6. LM Test Source
SS
df
MS
Model Residual
95.1054489 25.7758519
5 194
19.0210898 .132865216
Total
120.881301
199
.607443723
Number of obs F (5,194) Prob > F R-Squared Adj R-squared Root MSE
= = = = = =
200 143.16 0.0000 0.7868 0.7813 .36451
e
Coef
Std. Err.
t
P>t
Perc Ecom CSat Rec FPur _cons
2.060306 -.0943299 -.2682822 -.2923022 -.1175451 4.255712
.0771197 .03600179 .0381937 .039854 .0435935 .2956078
26.72 -2.62 -7.02 -7.33 -2..70 14.40
0.000 0.010 0.000 0.000 0.008 0.000
[95% conf. Interval] 1.908205 -.1653668 -.3436104 -.3709049 -.2035231 3.672694
2.212406 -.0232929 -.1929539 -.2136995 -.0315672 4.83872
7. Prediction Tabel --. Classification Table and Hit Rate Classified
True D 72 14 86
+ Total
Correctly classified
Total ͂D 14 100 114
86 114 200
86.00%
Based on actual data, the number of firms that consider the strategy of partnership with HBAT amounted to 86 units, while prediction by using four independent variables obtained results that 72 firms consider partnership strategy, so there are 14 firm that do not consider about the strategy partnership (the prediction error) , Based on actual data, the number of firms do consider about the strategy partnership with HBAT amounted to 114 units, but from prediction by using four independent variables obtained result that 100 firm do not consider about the strategy partnership, so there are 14 companies consider to strategy partnership (the prediction error ). Based on this, it can be concluded that the four independent variables able to predict the independent variable with the prediction accuracy of 86.00%.
1.00 0.75 Sensitivity
0.50 0.25 0.00 0.00
0.25
0.50 1 - Specificity
Area under ROC curve = 0.9478
Diagram ---. Reciver Operating Characteristic (ROC) Graph
Lsens graph
Conclusion
0.75
1.00