Table 8.

Decomposing the effect of control variables on the estimated difference in CIR and ROA between largest non-SIB and G-SIBs

	(1)	(2)	(3)	(4)	(5)	(6)
	Mean Q5	Mean G-SIB	CIR, 2010–2019		ROA, 2010–2019
			Coefficient	Implied effect on G-SIB vs. Q5	Coefficient	Implied effect on G-SIB vs. Q5
Deposits/Assets (%)	66.44	44.82	0.054	1.160	0.000	0.002
Mortgages/Assets (%)	69.01	14.45	0.012	0.654	− 0.002	− 0.118
Trading/Assets (%)	0.64	15.57	0.273	− 4.074	− 0.032	0.482
Net Int. Inc./Op. Inc. (%)	68.83	27.36	− 0.114	− 4.722	0.000	0.003
Commission Inc./Op. Inc. (%)	23.33	58.91	0.119	− 4.230	0.007	− 0.244
Trading Inc./Op. Inc. (%)	6.99	11.88	− 0.324	1.587	0.005	− 0.025
Capital/Assets (%)	8.17	4.85	− 1.379	− 4.581	0.027	0.090
RWA/Assets (%)	50.89	28.02	0.115	2.629	− 0.002	− 0.046
Domestic/Total Assets (%)	95.19	23.64	− 0.121	− 8.682	0.005	0.358
HHI of Mtg Holdings (/1000)	5.97	1.05	− 0.637	− 3.138	0.008	0.038
Avg. Local HHI (/1000)	1.99	1.90	0.726	0.060	0.035	0.003
Total				− 23.34		0.544

The first two columns show the average characteristics over 2010–2019 of non-SIBs in the largest quintile (Q5) vs. the G-SIBs. The third column shows the estimated coefficient on a given variable in the regression shown in column (6) of Table 6 (i.e., estimated on the non-G-SIB sample only). The fourth column then multiplies the difference between columns (1) and (2) with this coefficient from column (3). This yields the implied effect on non-SIB CIRs relative to G-SIBs’. The sum of these values adds up to the change in the gap between the Q5 and G-SIB coefficients between column (4) and column (6) of Table 6. The final two columns undertake a similar decomposition for the ROA regressions