Chapter 3: Methodology
3.2 Empirical Tests of Market Reaction
3.2.1 Measurement of Variables
Most investors’ stock-buying decisions are driven by an attention-grabbing event, and the releases of annual reports surely attract the attention of investors. As the annual reports are disseminated to the market, there is a condensed focus on them resulting in a change of return variability and trading volume (Holthausen and Verrecchia, 1990; Bamber and Cheon, 1995; Bamber, Barron, and Stober, 1997). I employ SAR (standardized absolute return) and SAAR (standardized absolute abnormal return), as used in previous literature, to measure the market reaction on return changes (e.g., Cready and Mynatt, 1991; Asthana and Balsam, 2001). In addition, I also calculate SV (standardized volume) to measure the market reaction on the trading volume, as used in Asthana and Balsam (2001) and Asthana, Balsam, and Sankaraguruswamy (2004).
SAR is obtained by subtracting the mean absolute return μ
( )
Rit during the non-filing period (-60, -11) from the absolute return Rit during the event period (-1, +5), and then deflating by the standard deviation of absolute returns σ( )
Rit duringthe non-filing period. It is defined as:
( ) ( )
The abnormal return is calculated by using the market model, with a 150-day (-210, -61) estimation period. Thus, SAAR is obtained by subtracting the mean absolute abnormal return μ(
ARit)
during the non-filing period (-60, -11) from the absolute abnormal return ARit during the event period (-1, +5), and then deflating by the standard deviation of absolute abnormal returns σ(
ARit)
during the non-filing period. It is defined as:
( )
The above two return reactions are measured as an unsigned daily stock return (abnormal return), because the primary concern of this study is with the magnitude of investors’ reaction to the release of the annual reports. Asthana and Balsam (2001) and Griffin (2003) posit that prior research has used the square of abnormal return as a form of measure, but the result is likely to be biased in favor of a few extreme stock returns. Hence, the use of standardized absolute return instead of standardized square return precludes the disturbance of extreme values and gives a more powerful test (Rohrbach and Chandra, 1989).
SV is obtained by subtracting the mean trading volume μ
(
VOLit)
measured in shares during the non-filing period (-60, -11) from the trading volume VOLit during the event period (-1, +5), and then deflating by the standard deviation of trading volume σ(
VOLit)
during the non-filing period. It is defined as:If the release of annual reports cannot attract investors’ attention, then SAR and SAAR are expected to be indistinguishable from zero.6 On the contrary, if annual reports attract investors’ attention and have information content, then the market reaction will be different from the non-filing period. The release of annual reports is said to contain information if it can alter the beliefs of market participants in a systematic way (Beaver, 1998).
With the arrival of new information during the filing period, Holthausen and Verrecchia (1990) identify two effects of new information: an informedness effect and a consensus effect. If information contained in the annual reports increases informedness, then the variance in unexpected return change and trading volume will both increase. However, if new information increases consensus, then the variance in unexpected return change will increase, but trading volume will decrease. Note that the trading volume is influenced by both informedness and consensus effects, and it may shift upward or downward depending upon which effect dominates.
3.2.2 Control Variables
In analyzing investors’ limited attention, we cannot ignore the information environment, and several characteristics of a firm are taken as control variables.
Hence, after controlling for possible information environmental factors, a multiple regression test allows us to examine if systematic behavioral biases are still encountered in the market.
The control variables herein are: (1) firm size — the uploading of annual reports to MOPS helps firms transmit their financial information to more market participants.
6 Prior research studies have posited that trading volume is one indicator of the attention a stock is receiving (Barber and Odean, 2007), or an indicator of sentiment (Hong and Stein, 2007). However, there is no natural definition of trading volume.
Thus, smaller firms generally convey more unexpected information and their stock returns respond more than larger firms (Atiase, 1985; Bamber, 1987; and Freeman, 1987). However, Griffin (2003) finds a positive association between firm size and market response; (2) unexpected earnings — prior evidence finds that average abnormal returns associated with the release of financial reports published earlier (later) than expected are positive (negative), indicating that early (delayed) reports carry good (bad) news (Kross, 1981; Givoly and Palmon, 1982; Chamber and Penman, 1984; and Kross and Schroeder, 1984); (3) debt level — this information is an important indicator of a company’s financial risk. A firm’s debt level is associated with a different market response (Dhaliwal, Lee, and Fargher, 1991; Dhaliwal and Reynolds, 1994; and Billings, 1999). When a firm’s debt level is too high, there is concern over future default and financial distress. In turn, a firm with a low debt level normally shows a solid financial structure, and investors pay attention to this information; (4) liquidity — Lee, Fox, and Liu (2001) find that information and liquidity trading both play an important role in explaining the intraday pattern of trading volume. Since more liquid stocks may be attracted by a larger group of investors, I include share turnover as a proxy variable to control for liquidity needs (Grullon, Kanatas, and Weston, 2004); (5) institutional holdings — attention is not as scarce a resource for institutional investors as it is for individuals (Barber and Odean, 2007).7 Dey and Radhakrishna (2007) find that institutions are most active in the immediate aftermath of an earnings announcement while individual investors are slow to explain the differences between individual and institutional trading volume reactions.
To examine different annual report filing timings that have an impact on the
7 Barber and Odean (2007) argue that institutions can use computers or pre-selection criteria to reduce their attention demands.
market reaction, I conduct the following multiple regressions:
(
,)
0 1 2 3 4it it it it it it it
SAR SAAR SV =β β+ DATTNT +β DEARLY +β DLATE +β LOGMV
5
SUE
it 6DTA
it 7LIQ
it 8INST
it itβ β β β ε
+ + + + + . (3.9)
The dependent variable is either SAR, SAAR, or SV measured over the release of the annual reporting period for company i at filing date t. The three indicator variables related to the different filing timings are as follows: DATTNT (1 = report filing belongs to the “attention-grabbing” zone, 0 = otherwise); DEARLY (1 = report filing belongs to the first 20%, 0 = otherwise); DLATE (1 = report filing belongs to the last 20%, 0 = otherwise). According to the Secretary Problem, the optimum strategy is to wait until after 37% of the earnings announcements and then to select the next relatively best one. Thus, I choose those sequential annual report filings which fall within the 37% to 57% area of the earnings release position as the
“attention-grabbing” zone.
The control variables related to a firm’s characteristics are as follows: LOGMV is the log of the market value of common equity at two days prior to the annual reports’
filing date; SUE is standardized unexpected earnings measured as the fourth quarter’s EPS minus the EPS from four quarters ago, deflated by the standard deviation of EPS changes over the preceding eight quarters (see Chan, Jegadeesh, and Lakonishok, 1996); DTA is total debts deflated by total assets; LIQ is share turnover measured as an annual average of total monthly volume divided by shares outstanding; and finally,
INST is the percentage of common shares held by institutions.
8
8 Due to informedness effect, I expect these control variables to be positively associated with stock return reactions — SAR and SAAR. However, I do not predict the sign of trading volume reactions because trading volume is influenced by both informedness and consensus effects, and it may shift upward or downward depending upon which effect dominates.