45 / 330
臺大管理論叢
第
27
卷第
2
期
45
4.2 Regression Results and Analyses
As a comparison with the results of Beccalli (2007), we first examine the influence of
IT expenditure on banks’ profitability.
11
As mentioned in Section 2, the failure of Beccalli
(2007) to find the positive effect of IT expenditure may be due to inappropriate specification.
We subsequently follow Shalev (2009), Bauwhede (2009), and Brown and Caylor (2009) to
make some modifications to the empirical models of Beccalli (2007).
12
Specifically, one-
year-ahead profitability measures (return on assets, denoted as ROA; return on equities,
denoted as ROE) are regressed on current profitability, IT intensity, and two control variables
(the natural log of market value and debt ratio). The results are displayed in Table 4. As
shown in Column (1), IT intensity is positively and significantly related to the one-year-
ahead ROA; the coefficient is 0.003 while p value is 0.000. Similarly, in Column (2), the
coefficient on IT intensity is 0.039 (
p
= 0.000), even after controlling for the effect of size
(LNMV) and leverage (DER). Taken together, the results in Table 4 demonstrate that firms
investing high amount of IT expenditures enjoy greater profitability and provide preliminary
support for our conjecture that IT expenditures play a critical role in value creation.
11 Barely any study examined the influence of IT expenditure on banks’ profitability in the U.S. banking
industry. Hence, we preliminarily test the association between the above two items in this paper.
12 Even though Beccalli (2007) states in the footnote that she also uses ROA in subsequent periods as the
dependent variable, that paper fails to control for the influence of current ROA and other determinants
(i.e., the specification is deficient to some degree). In the typical specification applied in accounting
literatures (e.g., Shalev, 2009; Bauwhede, 2009; Brown and Caylor, 2009), which uses profitability
measures (e.g., ROA and ROE) as the dependent variable, not only the independent variable of interest
but also current profitability, proxies for size effect (usually the natural log of market values) and for
leverage (usually the debt ratios) are included, so we follow this specification.