Appendix 2 “The Hausman test based on equation 3”
(b) (B) (b-B) sqrt(diag(V_b-V_B))
random fixed Difference S.E. lnW1 0.665689 0.64337 0.022318 0.01193 lnW21 -0.40531 -0.39207 -0.01324 0.012573 lnW3 0.094908 0.106566 -0.01166 0.009254 F9lnW1 0.115203 0.109676 0.005527 0.019522 F10lnW1 0.109375 0.099365 0.010009 0.016164 F11lnW1 -0.00065 -0.01291 0.012263 0.016498 F12lnW1 -0.05248 -0.05883 0.006354 0.020004 F13lnW1 -0.18374 -0.19402 0.010283 0.018303 F14lnW1 -0.22436 -0.22658 0.002219 0.016964 F15lnW1 -0.30449 -0.31271 0.00822 0.019284 F9lnW21 -0.0139 -0.02649 0.012588 0.018169 F10lnW21 -0.00426 -0.02194 0.017679 0.019696 F11lnW21 -0.03392 -0.03952 0.005603 0.019239 F12lnW21 -0.04691 -0.05727 0.01036 0.022274 F13lnW21 -0.15515 -0.1782 0.023054 0.024162 F14lnW21 -0.0414 -0.05475 0.013347 0.025338 F15lnW21 -0.05053 -0.05647 0.005935 0.026617 F9lnW3 -0.07397 -0.06815 -0.00582 0.01235 F10lnW3 -0.08973 -0.08294 -0.00679 0.013019 F11lnW3 -0.06419 -0.0554 -0.00879 0.012504 F12lnW3 -0.06976 -0.06557 -0.00419 0.013631 F13lnW3 -0.15764 -0.15854 0.000905 0.014355 F14lnW3 -0.13178 -0.13606 0.004276 0.013836 F15lnW3 -0.08878 -0.08875 -3.2E-05 0.014621 F9 0.332548 0.2631 0.069448 0.103483 F10 0.436227 0.329558 0.106669 0.100758 F11 0.069435 0.015819 0.053616 0.093962 F12 -0.076 -0.13293 0.056937 0.114528 F13 -0.88583 -1.02208 0.136245 0.123848 F14 -0.51134 -0.58015 0.068817 0.12251 F15 -0.75474 -0.81678 0.062034 0.124525 lnY1 -0.26011 -0.22457 -0.03554 0.008818 lnY2 0.155082 0.138015 0.017067 0.007411 lnY3 -0.00399 0.007516 -0.0115 0.003811 lnY4 -0.00755 -0.00981 0.002261 0.00625 E 0.069867 0.051133 0.018734 0.009555 b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic
Appendix 3 “Information and comments on outlier selection of the dataset”
My outlier selection, the upper and lower bound & comments.Binom. Interp.
--Variable Obs Percentile Centile [95% Conf. Interval] Bounds
W1 2038 1 0.0038968 0 0.0069354 0 99 0.3510899 0.290469 0.6264719 0.4 W21 1617 1 0.0011935 0.000497 0.002199 0.001 99 0.0761873 0.066856 0.0827046 0.1 W3 1949 1 0.0210322 0.005373 0.0513435 0.005 99 17.61343 14.90786 24.31552 18 Y1 2088 1 -0.0616469 -0.131972 -0.0257633 0.01 99 0.7685214 0.61731 0.8478096 0.5 Y2 2076 1 0.0114444 0.002072 0.0189876 0 99 0.9395949 0.918553 0.955449 1 Y3 2005 1 0 0 0 0 99 1 1 1 0.98 Y4 2053 1 -0.0356734 -0.058435 -0.0152881 0 99 0.1567332 0.13104 0.2115402 0.15 P1 2054 1 0.0162688 0.012684 0.0197685 0 99 0.28864 0.274904 0.3455678 0.3 P11 2054 1 0.5 0.38325 0.6354058 0.5 99 1800.818 1428.526 2174.11 1800
Comments on the upper and lower bounds of my dataset
In order to evaluate my dataset and control for outliers I ran the option "centile (1, 99)" in Stata for every variable of my model, the results are shown above. I compared them with table 3 of the article by Bikker, Spierdijk and Finnie (2006). They also analyzed the Panzar-Rosse model and ended up with upper and lower bounds for several variables (based of the 1st and 99th centile). My upper and lower bounds resemble the upper and lower bounds suggested by Bikker, Spierdijk and Finnie to a great extend. It is important to keep in mind is that my dataset consists of
unconsolidated data and the fact that there are many foreign banks active in the market who have their headquarters in another countries.
I adjusted the lower bound of W21 to my dataset compared to the bounds suggested by
Bikker, Spierdijk and Finnie. Unconsolidated data can play a role in explaning a shift in the bound, since a considerable amount of personnel is stationed at the the headquarters of a bank.
But the shift in the bound made is not very big. The upper bound of W3 is far above the suggested upper bound, this can be explained by the fact that the calculation is different, the other costs are divided by Fixed Assets compared to Total Assets in the table plus again the argument that the unconsolidated data can play a role in this case. A headquater can be a big portion of the Fixed Assets which is not taken into account because
of the use of unconsolidated data.
The last comment is on the upper and lower bound of Y4, these are based on the 1st and 99th of this dataset, because Bikker, Spierdijk and Finnie do not provide an upper or lower bound for this variable.
The outlier selection causes 112 observations to be removed from the dataset, in the case when only the year up from 1997 are taken into account and when all observations (of W1,
Appendix 4 “Main calculation; the FE-model with TD, IT and control var. E”
P11 FE int. Term P21 FE int. Term
Appendix 5 “Main calculation 2-period sample with control var. E”
P11 G FE int. term P21 G FE int. term
lnW1 0.629*** lnW1 0.534*** [0.031] [0.031] lnW21 -0.456*** lnW21 -0.460*** [0.041] [0.042] lnW3 0.062** lnW3 0.072*** [0.027] [0.028] lnY1 -0.242*** lnY1 -0.256*** [0.035] [0.035] lnY2 0.170*** lnY2 0.179*** [0.026] [0.026] lnY3 0.004 lnY3 -0.007 [0.014] [0.014] lnY4 -0.003 lnY4 0.047** [0.023] [0.024] E 0.115*** E 0.128*** [0.037] [0.038] GlnW1 -0.253*** GlnW1 -0.262*** [0.030] [0.030] GlnW21 -0.07 GlnW21 -0.052 [0.046] [0.046] GlnW3 -0.058** GlnW3 -0.060** [0.025] [0.026] G -0.886*** G -0.848*** [0.219] [0.222] Constant 3.023*** Constant 3.032*** [0.201] [0.204] Observations 939 Observations 939
Number of indexnumber 218 Number of indexnumber 218
R-squared 0.53 R-squared 0.49
No. of Obs. 939 No. of Obs. 939
No. of groups 218 No. of groups 218
Standard errors in brackets
* significant at 10%; ** significant at 5%; *** significant at 1%
Period H-statistic Period H-statistic
1997-2000 0.24 ¹ 1997-2000 0.15 ¹
2001 - 2004 -0.15 ² 2001 - 2004 -0.23 ²
Appendix 6 “Main calculation with the second interaction term included”
P11 FE E int. Term P21 FE E int. Term
Appendix 7 “Equilibrium-test, with ROA as dependent variable”
ROA1 FE int. term ROA2 FE int. term
lnW1 0.021 lnW1 0.03 [0.068] [0.076] lnW21 -0.139 lnW21 -0.154 [0.120] [0.135] lnW3 0.062 lnW3 0.11 [0.069] [0.077] lnY1 0.435*** lnY1 0.525*** [0.113] [0.129] lnY2 0.072 lnY2 0.009 [0.076] [0.085] lnY3 -0.008 lnY3 0.014 [0.041] [0.046] lnY4 0.303*** lnY4 0.337*** [0.069] [0.077] E -0.089 E -0.083 [0.109] [0.122] Constant -2.739*** Constant -2.694*** [0.520] [0.585] Observations 817 Observations 814
Number of indexnumber 205 Number of indexnumber 204
R-squared 0.07 R-squared 0.07
No. of Obs. 817 No. of Obs. 814
No. of groups 205 No. of groups 204
H-statistic -0.06 ² H-statistic -0.01 ²
Standard errors in brackets
Appendix 8 “1 H-statistic for the entire dataset by OLS with control var. E, Y6 and Y7”
P11 OLS E P21 OLS E P11 OLS Y6 P21 OLS Y6 P11 OLS Y7 P21 OLS Y7
lnW1 1.027*** 0.909*** 1.117*** 0.995*** 1.086*** 0.965*** [0.057] [0.057] [0.064] [0.064] [0.062] [0.062] lnW21 -0.402*** -0.387*** -0.410*** -0.396*** -0.411*** -0.397*** [0.078] [0.078] [0.079] [0.079] [0.079] [0.079] lnW3 -0.107** -0.112** -0.087* -0.091** -0.080* -0.085* [0.045] [0.045] [0.045] [0.045] [0.046] [0.046] lnY1 -0.763*** -0.779*** -0.747*** -0.764*** -0.753*** -0.769*** [0.081] [0.081] [0.082] [0.082] [0.083] [0.083] lnY2 0.571*** 0.537*** 0.585*** 0.553*** 0.597*** 0.565*** [0.068] [0.068] [0.069] [0.069] [0.069] [0.069] lnY3 -0.312*** -0.316*** -0.237*** -0.241*** -0.234*** -0.239*** [0.037] [0.037] [0.036] [0.036] [0.036] [0.036] lnY4 -0.02 0.056 -0.001 0.074 -0.007 0.069 [0.063] [0.063] [0.064] [0.064] [0.064] [0.064] E 0.625*** 0.617*** [0.086] [0.086] Y6 2.304*** 2.223*** [0.454] [0.454] Y7 1.810*** 1.750*** [0.399] [0.399] Constant 2.832*** 2.930*** 2.486*** 2.601*** 2.615*** 2.725*** [0.404] [0.404] [0.423] [0.423] [0.421] [0.420] Observations 939 939 939 939 939 939 R-squared 0.44 0.41 0.43 0.39 0.42 0.39 R-square 0.44 0.41 0.43 0.39 0.42 0.39 H-statistic 0.52 ¹ 0.41 ¹ 0.62 ¹ 0.51 ¹ 0.59 ¹ 0.48 ¹
Standard errors in brackets