![Sample our Area Studies journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5922/TOC-Subject-Sample-2021-Area-Studies.jpg"})
Abstract
We examine the relationship between weather effects and investor response to a firm's quarterly earnings announcements using data from 20 countries. Our results show that market cumulative abnormal returns are associated with unexpected earnings and with weather measures. Investors in some countries respond more negatively (less positively) to bad (good) earnings news when earnings are announced on cloudy days than on sunny days. We also find an asymmetric weather effect that is more significant for bad earnings announcements, on average. Moreover, the extent of this weather effect is less for countries with higher financial information transparency and those with common-law systems.
Notes
aIn our sample countries, the begin date of Brazil, Greece and Turkey in I/B/E/S is 1992/7, 1992/11 and 1991/12, respectively. Other countries’ begin dates are all earlier than 1990.
bObservations that firm codes between I/B/E/S and CRSP cannot be matched are also deleted.
cThe CARs over the event window (−1, 1) is calculated by firm actual returns minus expected returns from the market model estimated over the estimation window (−60, −2), with day zero being the earnings announcement date. Thus, observations with estimation data less than 59 days or event data less than three days are deleted.
1. Prior studies suggest a number of proxies for investor sentiment. For example, Lee et al. (Citation1991) extract indexes directly from the financial markets, such as the close-end fund discount, NYSE share turnover and the equity share in new issues. According to psychology evidence, soccer outcomes, weather conditions, lunar cycles, seasonal affective disorder, etc. are also used to measure investor sentiment (e.g. Kamstra et al. Citation2003).
2. Hirshleifer and Shumway (Citation2003) find that 18 of the 26 coefficients are negative, while only four are significant (Brussels, Milan, Australia and Vienna).
3. Following Saunders (Citation1993) and Hirshleifer and Shumway (Citation2003), we use weather in cities with stock exchanges to proxy investor mood. Saunders (Citation1993, p. 1337) states that “listed stocks are traded in NYC but represent claims on geographically diversified firms. Hence a local weather effect must work through local trading agents rather than the productive agents of listed firms … The focus on traders physically present in NYC has another benefit: test results are biased to favor the rational-markets hypothesis”.
4. While the ISWO data are of very high quality, a complete record of sky cover observations is not available for all of the cities with major stock exchanges, including Tokyo, Hong Kong, Seoul, Lisbon, Mexico City, Toronto, Jakarta, Frankfurt and Wellington (Hirshleifer and Shumway Citation2003). Thus, these cities are not included in the sample.
5. DeFond et al. (Citation2007) offer the first study in the accounting literature to use international earnings announcement dates from I/B/E/S. These authors perform tests to verify the accuracy of the announcement dates.
6. This provides a conservative measure of the cloud cover effect, since any effects caused by cloud-cover associated seasonal return patterns are excluded (Hirshleifer and Shumway Citation2003). However, we obtain similar results if we do not use the deseasonalized measure.
7. As Shon and Zhou (Citation2009, p. 218) state: “because earnings announcements are regularly occurring economic events that receive widespread attention, they are arguably least vulnerable to investor irrationality, and hence, least likely to exhibit a sunshine effect”. We also consider firm-specific returns only for days surrounding earnings announcements, which provide a conservative litmus test of the cloudiness effect.
8. We thank the anonymous referee for comments. Our conclusion is unchanged after controlling for merger events and calendar effects, such as the Monday and January effects.
9. Because changes in country-level institutions are a slow process, we do not measure country-level variables each year, due to data limitations as seen in some cross-country studies (Leuz et al. Citation2003, DeFond et al. Citation2007).
10. We thank the anonymous referee for the reminder that not all countries require quarterly reporting. The percentage of annual earnings announcements in our final sample is 32% (53,340/165,534).
11. The number of samples per country in DeFond et al. (Citation2007) ranges from 153 for Pakistan to 21,573 for the USA.
12. The stronger effect of bad mood is also evident in research on learning, neurological processes, feedback and memory (Baumeister et al. Citation2001).
13. However, if we divide countries into high and low financial information transparency groups based on the sample median, we find a significantly positive coefficient of PUE × × RULE (0.000013, t = 10.87) and a significantly negative coefficient of PUE ×
× RULE (−0.000005, t = −2.96).
14. The enforcement of accounting standards data is based on Hope (Citation2003). However, a number of sample countries are not included: Brazil, Greece, Malaysia, Philippines, Singapore, Taiwan, Thailand and Turkey.