Empirical Results

My main results are shown in Panel A, Table 2, and the coefficients of each variable are shown above in four lengths of periods. To simplify the coefficient listed in Panel A and analyze the different earnings responses (DERs) of good (bad) news, the incremental coefficients in current period and future periods are listed in Panel B.

Panel B shows the different earnings responses to bad news between Future Group and Current Group. The DERs of bad news between two groups are positive in current period (θ0 + λ0), denoting that bad news in Future Group is recognized earlier than in Current Group in current period. In future periods, DERs of bad news between Future Group and Current Group in future periods (θ1+ λ1) are positive in shorter-term periods (max n=1), and turn negative in longer-term periods (max n=2, 3 or 4). The

26

results show that information of bad news in Future Group is still recognized more than in Current Group next year, and is recognized less than in Current Group as time periods increase. These results are consistent with H1.

The different earnings responses (DERs) of good news between Future Group and Current Group (θ0) in current period are shown in Panel A. The value of θ0 is negative and statistically significant in current period. This indicates that good news in Future Group is recognized less than in Current Group in current period, consistent with H2. In future periods, the DERs of good news between Future Group and Current Group (θ1) remain negative, inconsistent with H2. Under conservatism, good news in Future Group is recognized in longer periods than in Current Group, but due to constraints², the time lengths in regression are only up to five years. Hence, the good news in Future Group might not be fully recognized in five years, but is reflected in longer periods.

Due to different speeds of recognition of good (bad) news in Future Group and Current Group, it’s suggested that the asymmetric timeliness of earnings is different in Future Group and Current Group. The following paragraph measures the difference in asymmetric timeliness between Future Group and Current Group in current period and future periods. The empirical results are shown in Panel A, Table 2.

2 To avoid data missing due to long lags, I refer to Lin and Liu (2012), and adopt the time lags (5 years), containing more signal contents of news.

27

The results show the bad news in Future Group is recognized in even faster pace than in Current Group, and earnings responses to bad news in Future Group are more than in Current Group in current period (θ0 + λ0). In another aspect, good news in Future Group is recognized in slower pace than in Current Group, and thus earnings responses to good news in Future Group are less than in Current Group in current period (θ0). Such reverse moving directions (increase or decrease) in recognition of bad news and good news magnify the asymmetric timeliness in Future Group. The difference in asymmetric timeliness (λ0) between Future Group (γ0 + λ0) and Current Group (γ0) remains positive and statistically significant, indicating that the magnification effect exists in Future Group in current period. The results are consistent with H3.a.

As mentioned above, the results show the good news in Future Group is still recognized less than in Current Group after 4 years from current period (θ1). But due to more decay in earnings responses to bad news in Future Group than in Current Group (θ1 + λ1), the asymmetric timeliness in Future Group decays more than in Current Group in long-term periods. Hence, the magnification effect in Future Group in future periods (λ1) still exists but decreases. The results are consistent with H3.b.

28

Table 2 Results for Effect of News in “Future Group” and “Current Group” on Multi-Period Earnings

Panel A. Choosing Positive (Negative) 35% Samples of 3-Day Return and Using 3 Day Return β0 0.0317 0.249 -0.0037 0.871 0.0024 0.919 0.0103 0.668 β1 -0.0482 0.044^b -0.0103 0.193 -0.0074 0.214 -0.0045 0.318 γ0 0.0106 0.789 0.0602 0.087^c 0.0319 0.366 0.0311 0.396 γ1 0.1319 0.004^a 0.1087 0.002^a 0.0983 0.003^a 0.0812 0.013^b θ0 -0.0619 0.05^b -0.0618 0.029^b -0.0532 0.066^c -0.0443 0.133 θ1 -0.0288 0.35 -0.0321 0.179 -0.0451 0.041^b -0.0692 0^a λ0 0.089 0.008^a 0.0917 0.031^b 0.095 0.026^b 0.0824 0.063^c λ1 0.0321 0.558 0.009 0.841 0.0122 0.779 0.0571 0.168 δ0 -0.0085 0.001^a -0.0078 0.003^a -0.0088 0^a -0.0093 0^a ζ0 0.0272 0^a 0.0255 0^a 0.0286 0^a 0.0289 0^a

Adj-R² 64.01% 51.85% 63.76% 63.72%

N 3242 3229 3202 2995 Notes: RETi,t means short window(3-day) stock return of firm i at year t. Eit indicates the

earnings of firm i at year t deflated by starting stock price of year t . ∑ E_it+n mean the aggregate EPS of firm i at future periods. Eit+n are each deflated by starting stock price of year t+n, and the stock dividend factors. In this model, I drop the 3-day return samples fall in median

±15% to capture the effect of one piece of news on multi-period earnings. To eliminate the effect of outliers, I separately drop 1% of outliers in 3-day return. E_it and ∑ E_it+n.A superscript of

‘a’,’b’ or ‘c’ indicates that result is significant at 0.01,0.05 and 0.1 level in a one-tail test if the coefficient has the predicted sign.

29

Table 2 Results for Effect of News in “Future Group” and “Current Group” on Multi-period Earnings

Panel B. Difference in earnings responses to good news in Each Period

Max n 1 2 3 4

Notes: Panel C. shows the difference in earnings responses between Future Group and Current Group.

The difference between incremental coefficients shows asymmetric timeliness of good news in two groups. A superscript of ‘a’,’b’ or ‘c’ indicates that result is significant at 0.01,0.05 and 0.1 level in a one-tail test if the coefficient has the predicted sign.

Panel C. Difference in earnings responses to bad news in Each Period

Max n 1 2 3 4

Notes: Panel C. shows the difference between incremental coefficients shows asymmetric timeliness of bad news in two groups. A superscript of ‘a’,’b’ or ‘c’ indicates that result is significant at 0.01,0.05 and 0.1 level in a one-tail test if the coefficient has the predicted sign.

30

Table 2 Results for Effect of News in “Future Group” and “Current Group” on Multi-period Earnings

Panel C. Difference in asymmetric timeliness of earnings in Each Period

Max n 1 2 3 4

Notes: Panel D extends the estimation of Panel C., showing the incremental effect of different kinds of news s. I use the incremental coefficient to estimate the asymmetric timeliness of different kinds of news recognition. A superscript of ‘a’,’b’ or ‘c’ indicates that result is significant at 0.01,0.05 and 0.1 level in a one-tail test if the coefficient has the predicted sign.