Regression Model for the Selection of Firm Value Determinants

---

To formally study the changes of patterns in three business models, we build a regression model which explains the impact of revenue and income to firm value. With this model, we attempt to evaluate the dynamic stages of pure e-tailers, C&M, and B&M. For each, we test the hypotheses at T1999–T2001 as follows.

H5a) The revenue has a positive (negative, or no) impact on market capitalization.

H5b) The income has a positive (negative, or no) impact on market capitalization.

H5c) The level of impact has changed during two consecutive time points.

H5d) The level of impact in C&M is significantly different from that in B&M.

We can derive the conclusion of H5a and H5b by testing the significance of coefficients b_R and b_I in model (1). We need to adopt dummy variables to test the hypotheses of H5c and H5d as in model (2) and (3) respectively. With the results derived from this study, we can conclude the dynamic stages of business models.

Impact of Revenue and Income

The estimated results from model (1) are summarized in Table 5. For the pure e-tailers, the revenue effect was positively significant at T2000 (1.8%) and T2001 (4.7%), and income effect became positively significant at T2001 (0.9%). For C&M, the income effect was positively significant for all times with 0.0% level of significance, but the revenue effect become significant (0.0%) only at T2001. However, B&M has both revenue and income effects significantly positive for all times. These results indicate significantly different patterns between the business models. The impacts measured by the coefficients are depicted in Figure 4.

Table 5: Revenue and Net Income Effects on Market Capitalization.

	Variable	T1999	T2000	T2001	Structural Changes
					T99 è T00	T00 è T01
Pure e-Tailers	Revenue (Standanized)	7.533 (0.492)	10.164 (1.314)	3.666 (0.637)	2.631	-6.497
	t	0.865	2.762	2.230	0.359	-1.574
	p-value	0.451	0.018	0.047	0.725	0.130
	Income	-4.365	9.928	14.792	14.292	4.864
	(Standanized)	(-0.214)	(0.625)	(0.530)
	t	-0.376	1.313	1.855	1.127	0.442
	p-value	0.732	0.216	0.091	0.279	0.663
C&M	Revenue (Standanized)	0.205 (0.151)	-0.159 (-0.105)	0.593 (0.441)	-0.364	0.752
	t	1.339	-1.198	7.060	-1.779	4.678
	p-value	0.183	0.234	0.000	0.077	0.000
	Income (Standanized)	37.768 (0.799)	50.287 (1.063)	20.990 (0.542)	12.519	-29.297
	t	7.114	12.112	8.688	1.832	-5.951
	p-value	0.000	0.000	0.000	0.068	0.000
B&M	Revenue (Standanized)	0.516 (0.686)	0.172 (0.444)	0.284 (0.627)	-0.344	0.112
	t	8.949	5.641	11.131	-5.469	2.735
	p-value	0.000	0.000	0.000	0.000	0.007
	Income (Standanized)	8.503 (0.273)	8.276 (0.529)	9.707 (0.390)	-0.227	1.431
	t	3.567	6.728	6.924	-0.088	0.765
	p-value	0.001	0.000	0.000	0.930	0.446

Figure 4: Dynamic Behaviors of e-Tailers and Retailers.

Pure retailers have moved from exploration stage to growth stage even though there was a big fluctuation in T2000. C&M fluctuated along the income effects, but B&M did not have prominent changes.

Time Effect

To confirm if β_R and/or β_I significantly changed between two consecutive time points, a dummy variable D is introduced as in model (2).

(2) 

where

D	=	0 if a datum belongs to the preceding time point, and
D	=	1 if a datum belongs to the succeeding time point.

The interaction variables, DR = D × Revenue and DI = D × Income, are included to test the structural changes between the two consecutive time points. The change in revenue can be examined by testing the null hypothesis of β₄ = 0, and the change in net income by testing β₅ = 0.

The results are summarized in the right-hand side column of Table 5. According to this study, both the revenue and income effects in C&M were significantly (10%) increased at T2001, and revenue effect in B&M was significantly (1%) increased both at T2000 and T2001.

Effect of Introducing Click to Retailers

To test the structural difference between C&M and B&M, we use model (3) with dummy variable M.

(3) 

where

M	=	0 if a datum belongs to C&M, and
M	=	1 if a datum belongs to B&M.

The interaction variables, MR = M × Revenue and MI = M × Income, are included to test the structural difference between C&M and B&M. The difference in revenue can be examined by testing the null hypothesis of β₄ = 0, and the difference in income by testing β₅= 0.

The estimated results are summarized in Table 6. According to this study, the structural differences between C&M and B&M are detected as follows:

1. Income effects of click were significant at T1999 and T2000 (0.5% and 0.0% respectively).

2. Revenue effect of click was significant at T2001 (7.8%).

Table 6: C&M vs. B&M: Retailers Click Effect.

		C&M	B&M	Click Effect B&M è C&M
T1999	Revenue	0.205	0.516	-0.311
	t	1.339	8.949	-1.216
	p-value	0.183	0.000	0.226
	Income	37.768	8.503	29.265
	t	7.114	3.567	2.861
	p-value	0.000	0.001	0.005
T2000	Revenue	-0.159	0.172	-0.331
	t	-1.198	5.641	-1.416
	p-value	0.234	0.000	0.159
	Income	50.287	8.276	42.011
	t	12.112	6.728	4.639
	p-value	0.000	0.000	0.000
T2001	Revenue	0.593	0.284	0.309
	t	7.060	11.131	1.773
	p-value	0.000	0.000	0.078
	Income	20.990	9.707	11.283
	t	8.688	6.924	1.246
	p-value	0.000	0.000	0.215

These results imply that by adding the e-tailing channel to retailers, the income effect was apparent in T1999 and T2000, but the revenue effect apparent in T2001.

---