Regulation fair disclosure and the cost of equity capital

Abstract

We examine the effect of Regulation Fair Disclosure (Reg FD) on the cost of equity capital. We find some evidence that (1) the cost of capital declines in the post-Reg FD period relative to the pre-Reg FD period, on average, for a broad cross-section of US firms, (2) the decrease in the cost of capital post Reg FD is mainly for medium and large firms but is insignificant for small firms, and (3) the decrease in the cost of capital post Reg FD is systematically related to firm characteristics indicative of selective disclosure before Reg FD. In contrast, we find little evidence of a decrease in the cost of capital for American Depositary Receipts and US-listed foreign firms, which are legally exempt from Reg FD. Overall, our findings do not support a conclusion in recent studies that the cost of capital has increased post Reg FD and, if anything, suggest the opposite.

Appendix: Firm-specific estimation of the implied cost of equity capital

We first define the following variables used in the three firm-specific cost of capital estimation models.

P _t :

Stock price of a firm’s common share at time t

E _t [.]:

Expectation based on information at time t

DPS _t+τ :

Future dividends per share for time t + τ

B _t :

Book value of equity per share from the most recent available financial statement at time t

EPS _t+τ :

Median earnings forecasts per share from I/B/E/S or derived earnings forecasts per share for time t + τ

POUT:

Forecasted dividend payout ratio. We use a firm’s previous fiscal year’s dividend payout ratio to measure the forecasted payout ratio. If earnings are negative, we assume a return on assets of 6% to calculate earnings

1.1 Gebhardt et al. (2001)—GLS model

The GLS model is derived from the dividend discount model

$$ P_{t} = \sum\limits_{\tau = 1}^{\infty } {{\frac{{E_{t} \left[ {{\text{DPS}}_{t + \tau } } \right]}}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} . $$

(GLS1)

Under the “clean surplus” assumption, that is, $B_{t + \tau } = B_{t + \tau - 1} + {\text{EPS}}_{t + \tau } - DPS_{t + \tau }$, the above equation can be rewritten as

$$ P_{t} = B_{t} + \sum\limits_{\tau = 1}^{T - 1} {{\frac{{E_{t} \left[ {{\text{EPS}}_{t + \tau } - R_{\text{GLS}} \times B_{t + \tau - 1} } \right]}}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} + \sum\limits_{\tau = T}^{\infty } {{\frac{{E_{t} \left[ {{\text{EPS}}_{t + \tau } - R_{\text{GLS}} \times B_{t + \tau - 1} } \right]}}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} $$

(GLS2)

or

$$ P_{t} = B_{t} + \sum\limits_{\tau = 1}^{T - 1} {{\frac{{E_{t} \left[ {{\text{ROE}}_{t + \tau } - R_{\text{GLS}} } \right] \times B_{t + \tau - 1} }}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} + \sum\limits_{\tau = T}^{\infty } {{\frac{{E_{t} \left[ {{\text{ROE}}_{t + \tau } - R_{\text{GLS}} } \right] \times B_{t + \tau - 1} }}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} , $$

(GLS3)

where ROE_t+τ is the return on equity for time t + τ. Assuming that the abnormal earnings at time T, that is, $\left[ {{\text{ROE}}_{t + T} - R_{\text{GLS}} } \right] \times B_{t + T - 1}$, becomes a perpetuity from T onward, Eq. GLS3 can be rewritten as

$$ P_{t} = B_{t} + \sum\limits_{\tau = 1}^{T - 1} {{\frac{{E_{t} \left[ {{\text{ROE}}_{t + \tau } - R_{\text{GLS}} } \right] \times B_{t + \tau - 1} }}{{\left( {1 + R_{\text{GLS}} } \right)^{\tau } }}}} + {\frac{{E_{t} \left[ {{\text{ROE}}_{t + T} - R_{\text{GLS}} } \right] \times B_{t + T - 1} }}{{\left( {1 + R_{\text{GLS}} } \right)^{T - 1} R_{\text{GLS}} }}}. $$

(GLS4)

To estimate Eq. GLS4, Gebhardt et al. (2001) first use I/B/E/S analysts’ earnings forecasts for the next 3 years as the expected earnings from year t + 1 to year t + 3. They then measure the expected earnings by assuming that the future return on equity declines linearly to an equilibrium return on equity from year t + 4 to year t + T. This equilibrium return on equity is a moving median return on equity for all firms in the same industry in the past 10 years. The return on equity (ROE) is calculated as income before extraordinary items (Compustat data item #18) scaled by the lagged total book value of equity (Compustat data item #60). We classify all firms into 48 industries defined by Fama and French (1997). Firm-year observations with a negative return on equity are eliminated in the calculation. The future book value of equity is estimated by assuming the clean surplus relation, that is, $B_{t + \tau } = B_{t + \tau - 1} + {\text{EPS}}_{t + \tau } - {\text{DPS}}_{t + \tau }$. The future dividend, DPS _t+τ , is calculated by multiplying EPS _t+τ by the payout ratio POUT. Following Gebhardt et al. (2001), we set T = 12.

1.2 Claus and Thomas (2001)—CT model

The CT model is similar to the GLS model, except that it assumes that the abnormal earnings after 5 years grow at a rate of g _lt in perpetuity. Under this assumption, we have

$$ P_{t} = B_{t} + \sum\limits_{\tau = 1}^{5} {{\frac{{E_{t} \left[ {{\text{EPS}}_{t + \tau } - R_{\text{CT}} \times B_{t + \tau - 1} } \right]}}{{\left( {1 + R_{\text{CT}} } \right)^{\tau } }}}} + {\frac{{E_{t} \left[ {{\text{EPS}}_{t + 5} - R_{\text{CT}} \times B_{t + 4} } \right]\left( {1 + g_{\text{lt}} } \right)}}{{\left( {1 + R_{\text{CT}} } \right)^{5} \left( {R_{\text{CT}} - g_{\text{lt}} } \right)}}}. $$

(CT1)

To estimate Eq. CT1, Claus and Thomas (2001) use the I/B/E/S analysts’ earnings forecasts to derive the abnormal earnings for year t + 1 to year t + 5. Earnings forecasts for year t + 4 and year t + 5 are derived from earnings forecasts for year t + 3 and the long-term earnings growth rate. If the long-term earnings growth rate is missing in I/B/E/S, an implied earnings growth rate from EPS_t+2 and EPS_t+3 is used instead. The long-term abnormal earnings growth rate (g _lt) is calculated using contemporaneous risk-free rate (the yield on the 10-year Treasury bonds) minus 3%. The future book value of equity is also estimated by assuming the clean surplus relation, that is, $B_{t + \tau } = B_{t + \tau - 1} + {\text{EPS}}_{t + \tau } - {\text{DPS}}_{t + \tau }$. The future dividend DPS _t+τ is calculated by multiplying EPS _t+τ by the payout ratio POUT.

1.3 Gode and Mohanram (2003)—GM model

The GM model is derived from the Ohlson and Juettner-Nauroth (2005) model, which in turn is based on the dividend discount model below.

$$ P_{t} = \sum\limits_{\tau = 1}^{\infty } {{\frac{{E_{t} \left[ {{\text{DPS}}_{t + \tau } } \right]}}{{\left( {1 + R_{\text{GM}} } \right)^{\tau } }}}} . $$

(GM1)

Defining the dividend-adjusted earnings (z _t ) as $z_{t} = {{\left[ {{\text{EPS}}_{t + 1} - {\text{EPS}}_{t} - R_{\text{GM}} \left( {{\text{EPS}}_{t} - {\text{DPS}}_{t} } \right)} \right]} \mathord{\left/ {\vphantom {{\left[ {{\text{EPS}}_{t + 1} - {\text{EPS}}_{t} - R_{\text{GM}} \left( {{\text{EPS}}_{t} - {\text{DPS}}_{t} } \right)} \right]} {R_{\text{GM}} }}} \right. \kern-\nulldelimiterspace} {R_{\text{GM}} }}$ and assuming that the sequence of $\left\{ {z_{t} } \right\}_{t = 1}^{\infty }$ grows asymptotically at a rate of g _lt, that is, $z_{t + 1} = (1 + g_{\text{lt}} )z_{t}$, Eq. GM1 can be rewritten as

$$ P_{t} = {\frac{{E_{t} \left[ {{\text{EPS}}_{t + 1} } \right]}}{{R_{\text{GM}} }}} + {\frac{{E_{t} \left[ {{\text{EPS}}_{t + 2} - {\text{EPS}}_{t + 1} - R_{\text{GM}} \left( {{\text{EPS}}_{t + 1} - {\text{DPS}}_{t + 1} } \right)} \right]}}{{R_{\text{GM}} \left( {R_{\text{GM}} - g_{\text{lt}} } \right)}}}. $$

(GM2)

Solving for R _GM, we obtain

$$ R_{\text{GM}} = A + \sqrt {A^{2} + {\frac{{E_{t} \left( {{\text{EPS}}_{t + 1} } \right)}}{{P_{t} }}}\left( {g_{\text{st}} - g_{\text{lt}} } \right)} , $$

(GM3)

where $A = {\frac{1}{2}}\left[ {g_{\text{lt}} + {\frac{{E_{t} \left( {{\text{DPS}}_{t + 1} } \right)}}{{P_{t} }}}} \right]$ and $g_{\text{st}} = {\frac{{{\text{EPS}}_{t + 2} - {\text{EPS}}_{t + 1} }}{{{\text{EPS}}_{t + 1} }}}$.

To estimate Eq. GM3, Gode and Mohanram (2003) use the average of the short-term earnings growth rate implied in EPS_t+1 and EPS_t+2 and the analysts’ forecasts of long-term growth rates as their measure of g _st. They further require that EPS_t+1 > 0 and EPS_t+2 > 0. The long-term asymptotic growth rate (g _lt) is calculated using the contemporaneous risk-free rate (the yield on the 10-year Treasury bonds) minus 3%. The future dividend DPS_t+1 is calculated by multiplying EPS_t+1 by the payout ratio POUT.