Page 40 - 32.3

The Effect of Information Opacity on the Weighting of Performance Measures in the Compensation Contracts
of CEOs: Evidence from U.S. Firms

Next, we test equations (3) and (4). The estimated coefficients of ROA× |DACCa| and
RET× |DACCa| are -7.786 (p-value = 0.021) and 1.699 (p-value = 0.109), respectively; the
estimated coefficients of ROA × |DACCb| and RET × |DACCb| are -3.924 (p-value = 0.000)
and 0.246 (p-value = 0.623), respectively. The signs of the estimated coefficients are in

line with our expectations, although the estimated coefficients of RET × |DACCa| and RET
× |DACCb| are not statistically significant.
Furthermore, we conduct robustness tests by changing the formula used to calculate

ROA. In the main tests, we calculate ROA by using net income as our numerator. In
the robustness tests, we use EBITDA (Earnings before Interests, Taxes, Depreciation,
and Amortization) and EBIT (Earnings before Interests and Taxes), respectively, as our
numerator to calculate ROA and re-estimate the same regression models. Our robustness
tests yield similar empirical results as the main tests.

Lastly, we take industrial differences into account to investigate whether different
business models across industries lead to variations in the weighting of the performance
measures. We find that different business models across industries do not result in

significantly higher or lower weightings of the performance measures.

4. Implications and Research Limitations

Through this empirical study, we verify the theory proposed by Banker and

Datar (1989). We first show that the volume of information can affect the weighting of
performance measures. For instance, if the volume of stock-related information is limited,
the weighting of the accounting-based performance measure increases significantly,
and, in the meantime, the weighting of the stock-based performance measure decreases

significantly. Next, we show that the quality of information can also affect the weighting
of performance measures. For instance, if the magnitude of discretionary accruals is large,
we see a negative effect on the accounting-based performance measure.
In general, we find that both information volume and information quality are of great

influence. Therefore, when asking for more information to be publicly released, we should
also be cautious about the quality of the information provided.
On the other hand, we acknowledge two research limitations existed in this research.
First, another potential candidate, such as the frequency of a firm being reported by mass

30