The effect of a firm’s reporting lag on the post-earnings-announcement drift

(1)

The effect of a firm’s reporting lag on the

post-earnings-announcement drift

Name: Daan Stam

Student number: 11403357

Thesis supervisor: Ms S. Schafhäutle MSc Date: June 25, 2018

Word count: 23393

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Daan Stam who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study examines the effect of a firm’s reporting lag, defined as the number of days between a firm’s fiscal quarter end date and the date of its quarterly earnings announcement (EA), on the post-earnings-announcement drift (PEAD). Over time, a firm’s information environment has changed because of technological developments like the internet and social media. Therefore, I hypothesize that the magnitude of the PEAD becomes smaller as the reporting lag becomes longer, because of a higher amount of information being available to investors prior to the EA date. The sample consists of 5,934 quarterly EAs between 2011 and 2016 of 381 firms from the STOXX Europe 600 Index. The results of a regression analysis are in contrast with the hypotheses. The length of the reporting lag has no significant effect on the complete PEAD and the PEAD that takes place in the first five trading days of the PEAD window. However, when the reporting lag becomes longer, the PEAD within the three trading days around the subsequent EA date is significantly more positive if unexpected earnings (UE) are positive, while its effect is insignificant if UE are negative. This study contributes to the literature by combining two streams of literature that received much attention in the past, namely the PEAD and the timing of a firm’s EA. Moreover, this study provides mixed evidence on the existence of the PEAD within a European setting.

Keywords: post-earnings-announcement drift, unexpected earnings, abnormal returns, earnings announcement, timing, reporting lag, stock price reaction, Europe, technological developments, internet reporting, social media, archival study

(4)

List of figures and tables

Figures:

Figure 1: Overview of timeframes 24

Figure 2: Expectations hypothesis 1 25

Figure 3: Expectations hypothesis 2a and 2b 26

Figure 4: Distribution of the PEAD over time 40

Tables:

Table 1: Sample selection procedure 28

Table 2: Predicted effects on the PEAD 33

Table 3: Descriptive statistics 35

Table 4: Distribution of reporting lag over time 36

Table 5: Pearson correlation matrix 37

Table 6: Distribution of CARs over time 41

Table 7: Regression analysis - Hypothesis 1 42

Table 8: Regression analysis - Hypothesis 2a 44

Table 9: Regression analysis - Hypothesis 2b 47

(6)

1 Introduction

Over the last five years, stocks from the Standard & Poor’s 500 Index faced an average price change of 1.1% on positive quarterly earnings per share (hereafter EPS) surprises during the five days around the earnings announcement (hereafter EA) date (Butters, 2018). When the quarterly EPS surprise is negative, the average price change was 2.4% negative. Closer to home, bpost SA is a nice example to show that markets react on a firm’s EA. On 13 March 2018, bpost SA released its fourth quarter and full-year results of 2017 (bpost SA, 2018). On the day after, the stock price plunged about 22%. At the end of May 2018, their stocks are valued approximately 45% lower compared to 13 March 2018. This effect applies in the other direction as well. For example, Nokia Corporation published its fourth quarter and full-year 2017 results on 1 February 2018 (Nokia Corporation, 2018). On this day, the stock price increased around 12%. Moreover, on 31 May 2018, the stock price is about 28% higher relative to the day before the fourth quarter EA date.

In the literature, this phenomenon is called the post-earnings-announcement drift (hereafter PEAD), which is the positive association between the sign and magnitude of a firm’s unexpected earnings (hereafter UE) and its abnormal stock returns after the EA date. Ball and Brown (1968) were the first who demonstrated this capital market anomaly, which deviates from the idea of an efficient market. Over time, several studies replicated and extended their findings by using different data sources (Bernard & Thomas, 1989; Foster, Olsen, & Shevlin, 1984; Rendleman, Jones, & Latané, 1982). In general, the PEAD is a result of the initial underreaction from investors at the EA date, because they are unable to identify immediately the effect of current earnings for the realization of future earnings (Bernard & Thomas, 1990). I expect the chance of information from a firm’s EA being available to investors before the EA date to be higher when the reporting lag, which is the number of days between the fiscal quarter end date and the EA date, becomes longer. Therefore, this study answers the following research question:

RQ: Does the length of the reporting lag prior to a firm’s EA affect the magnitude of the

post-earnings-announcement drift.

Answering this research question is relevant for many reasons. The first and most important one is as follows: No similar research exists which specifically examines the effect of a firm’s reporting lag on the PEAD. The PEAD and the timing of a firm’s EA are two well-developed streams of literature. Unfortunately, there is only limited and mixed evidence about the combination of these two subjects, derived from studies with a different emphasis. According to Givoly and Palmon (1982), the stock price change in the two weeks after the annual EA date is higher if the reporting lag is shorter. On the other hand, Chambers and Penman (1984) showed

(7)

that the stock price reaction during the fifty trading days after the EA is stronger for firms with a longer reporting lag. Research on the timing of a firm’s EA is mostly interested in the short-term stock price reaction around the EA date or in the unexpected part of the reporting lag. The objective is to observe whether a delay relates to the forthcoming release of bad news and vice-versa (Bagnoli, Kross, & Watts, 2002; Begley & Fischer, 1998; Chambers & Penman, 1984; Givoly & Palmon, 1982; Kross, 1981; Kross & Schroeder, 1984).

Over time, the information environment has changed dramatically. Inventions like social media or the internet in general, have created new information channels for the distribution of firm-specific information (Healy & Palepu, 2001; Miller & Skinner, 2015). This could influence the magnitude of the PEAD in general, but especially in case of a longer reporting lag. The production of firm-specific information has increased, because digital platforms provide outsiders with the opportunity to share their information with others. Hence, the firm has lost part of the control over its information environment. Furthermore, information is distributed in a more quickly manner and is broadly accessible on a real-time basis for everybody who is interested. Therefore, it is relevant to examine whether these changes had an impact on the informativeness of a firm’s EA at the time of the EA date, especially when the reporting lag becomes longer.

I created a sample of 5,934 quarterly EAs from 381 firms. These firms are part of the STOXX Europe 600 Index anytime between 2011 and 2016. I conducted a linear OLS regression analysis, with the PEAD as the dependent variable, which I measured as the size-adjusted cumulative abnormal returns (hereafter CARs). Reporting lag is the independent variable. To control for other factors, which prior literature indicated to affect the PEAD, I added multiple control variables to the regression model: UE, unexpected reporting lag, firm size, beta, book-to-market ratio, analyst following and a dummy variable to distinguish fourth quarter EAs from all other quarterly EAs.

During the sample period, the mean reporting lag decreases from 37.1 days in 2011 to 35.5 days in 2016. Using size-adjusted CARs, I found evidence of the PEAD. On average, the total size-adjusted CARs of firms from the lowest UE decile are 1.7% negative, while they are 0.7% positive with those from the highest UE decile. Unfortunately, these results are not robust to using Market-Model CARs instead of size-adjusted CARs.

To answer the research question, I formulated three hypotheses. Each hypothesis refers to a unique part of the PEAD. According to H1, the magnitude of the PEAD, measured from two trading days after the EA to one trading day after the following EA, is smaller if the reporting lag increases. From the regression analysis, I find no evidence for a significant relation between the

(8)

length of the reporting lag and the PEAD, regardless of the sign of UE, indicating that investors do not obtain more information from the firm’s EA before the EA date when the reporting lag is longer. Another reason for these results could be that investors still wait with their response until after the release of the EA. H2a stated that the magnitude of the PEAD taking place during the first five trading days of the PEAD window is larger when the length of the reporting lag increases. If investors obtain more information in advance, they should be able to process the remaining information at the EA date in a more quickly manner. However, as with H1, I find no support for this hypothesis. Finally, H2b predicted the magnitude of the part of the PEAD observable in the three trading days around the next quarterly EA date to be smaller if the reporting lag becomes longer. Because of a lower level of uncertainty amongst investors after the EA date, there is less need to wait for the next EA to confirm their initial thoughts. In contrast, I find that the magnitude becomes significantly more positive after an EA revealing positive UE. In this case, investors delay a larger part of their response until the next EA date after a longer reporting lag. When UE is negative, the effect is insignificant. Therefore, I reject H2b as well.

This study contributes to the literature in different ways. First, as mentioned before there is only limited evidence about the effect the reporting lag on the PEAD. Although I found no evidence for such an effect, this study could serve as a starting point to answer this research question within another setting or by using data from different firms or different timeframes. Second, the literature on those two subject is mostly based on data from the United States. Using a sample of large European firms, this study provides mixed evidence on the existence of the PEAD in Europe. The presence of the PEAD depends on the way I measured CARs. Future research could apply other methods to measure CARs to provide extensive evidence on whether the PEAD has disappeared over time. Thirdly, this study shows that information from a firm’s EA does not leak to the market when the reporting lag becomes longer, despite the dramatic changes over time in a firm’s information environment.

In practice, the findings of this study could be of interest to different groups. Since the impact of the reporting lag on the PEAD is insignificant, there is no need for preparers of financial statements to change the reporting lag in order to affect the magnitude of the PEAD. Analysts may need to consider the effect of the length of the reporting lag in their stock recommendations and quarterly EPS forecasts. Moreover, regulators and standard setters may be interested in the average length of the reporting lag. Furthermore, they do not want firms to change their reporting lag to influence the magnitude of the PEAD. To enhance transparency, they may decide to shorten the maximum allowed reporting lag.

(9)

The remainder of this study is outlined as follows. Chapter 2 reviews the literature on the PEAD, the timing of a firm’s EA and the effects of the change of times. In Chapter 3, I explain the hypotheses of this study. Thereafter, Chapter 4 describes the sample selection, the operationalization of variables and the regression model. In Chapter 5, I present the descriptive statistics and the results of the hypotheses testing, including the outcome of the robustness test. In Chapter 6, I end with a discussion and conclusion, in which I answer the research question.

(10)

2 Literature review and theory

In this chapter, I review the most relevant literature in relation to the research question. I divided the topic into four key concepts. The review starts with a description of one of the most proven capital market anomalies, namely the PEAD, in Section 2.1. Further, I present evidence on the effects of the timing of a firm’s EA on stock prices in Section 2.2. In Section 2.3, I combine those two topics by discussing current evidence on the effect of different forms of timing of a firm’s EA on the PEAD. The literature about changes over time in the information content of EAs and the value relevance of earnings is reviewed in Section 2.4. Lastly, in Section 2.5 I discuss recent literature about the impact of changes in a firm’s disclosure landscape over time.

2.1 The post-earnings-announcement drift

Ideally, we assume capital markets to be efficient. In an efficient market, “given the available information, actual prices at every point in time represent very good estimates of intrinsic values” (Fama, 1965a, p. 90). Hence, when the intrinsic value changes, the actual price should respond instantaneously, because of competition between all the intelligent participants. Hence, actual prices “already reflect the effects of information based both on events that have already occurred and on events which, as of now, the market expects to take place in the future” (Fama, 1965b, p. 56). On average, within an efficient market, a portfolio consisting of randomly picked stocks should not perform any worse than a portfolio of stocks carefully selected by professional analysts, under the assumption that both portfolios are exposed to the same degree of risk. Professional analysts should be unable to discover and take advantage of any discrepancies between a firm’s intrinsic value and actual price. There are three different forms of market efficiency, of which the semi-strong form is the most relevant for the present study. The semi-strong form means that stock prices change efficiently when publicly available information is released (Fama, 1970). According to Fama (1970), there is no evidence against the semi-strong form of market efficiency. Nevertheless, Ball and Brown (1968) discovered that abnormal stock returns move in the same direction as the earnings news for as long as two months beyond the EA date. Unfortunately, they were unable to explain their findings. Over time, several other studies found evidence inconsistent with the existence of market efficiency (Jensen, 1978). Ball (1978) conducted a literature survey and reported that prior literature found excess returns after a firm’s EA. He argued that if a firm’s EA is a public good, which seems reasonable, it should not lead to private returns. Although these findings are at first sight inconsistent with the concept of market efficiency, he claims that they are a result of methodological misspecifications. However, Watts (1978) found

(11)

significant abnormal returns subsequent to a firm’s EA after controlling for the deficiencies reported by Ball (1978). Therefore, he concluded that these findings are to some extent a consequence of market inefficiency. Fama (1998), on the other hand, stuck to his prior beliefs that the efficient market hypothesis is still valid and concluded that anomalous evidence is based on chance results. However, he admitted that some anomalies “are above suspicion” (p. 304), including the PEAD, which he also called “the granddaddy of underreaction events” (p. 286).

Later on, other studies replicated the study of Ball and Brown (1968) discussed above and extended the data to quarterly EAs. The phenomenon became known as the PEAD (Bernard & Thomas, 1989; Foster et al., 1984; Rendleman et al., 1982). These studies also examined the effect of the magnitude of earnings news on the PEAD and found that, except for the sign, also the magnitude of UE is associated with the PEAD. That is, higher (lower) UE are associated with higher (lower) abnormal returns after a firm’s quarterly EA. Foster et al. (1984) as well as Bernard and Thomas (1989) observed differences in the magnitude of the PEAD across small and large firms. This is consistent with prior literature on the effect of firm size on stock returns (Banz, 1981). On average, smaller firms face higher abnormal returns after they have announced their quarterly earnings than do larger firms.

When it became clear that the findings of Ball and Brown (1968) were not due to coincidence, a new stream of literature emerged, aimed at explaining the PEAD. A major study in this area is that of Bernard and Thomas (1989), who provided the most compelling reason amongst all competing explanations of the PEAD. The first explanation they assessed is that researchers fail to adjust raw returns fully for risk when calculating abnormal returns. They argued that firms that have announced high (low) UE became more (less) risky. In this case, abnormal returns should be interpreted as a compensation for risk. The second explanation assumes that there is a delay in part of the price response to new information. Either this could be due to investors being unable to process the available information or due to the costs of using the information, such as transaction costs, being higher than the benefits that could be obtained from it. Most of their results confirmed a delayed response to information, while it was difficult to find support for the first explanation. However, they were not fully satisfied about the presence of transaction costs as the explanation of the existence of the PEAD.

Hence, Bernard and Thomas (1989) presented an alternative explanation for this delayed response. They argued that some investors fail to identify immediately at a firm’s EA what current earnings mean for the realization of future earnings. This leads to an initial underreaction at the EA date. After receiving supportive evidence from subsequent analyst forecast revisions or a firm’s

(12)

next quarterly EA, those investors correct gradually for their initial underreaction, leading to the PEAD. Consistent with this idea, they found that a large part of the PEAD takes place around the EA of the next quarter. This is in line with the findings of Freeman and Tse (1989). In their follow-up study, Bernard and Thomas (1990) investigated this explanation in more detail. It seemed that investors assume earnings to follow a naïve seasonal random walk model. With this model, current expected earnings are modelled as following the earnings of four quarters ago. However, it turned out that the actual behavior of earnings deviates from this model. Therefore, by using information about the historical behavior of earnings, the sign and magnitude of a firm’s UE in the eyes of investors are predictable up to four quarters ahead. As a result, the future stock returns around upcoming EAs are predictable, since abnormal returns are a function of a firm’s UE. More specifically, Bernard and Thomas (1990) showed that, by using only current earnings, they were able to predict the sign and magnitude of the stock price reaction around the upcoming four quarterly EAs. Abnormal returns around the subsequent three quarterly EAs remained positive after a firm has just announced positive UE over the current quarter, while abnormal returns moved in the opposite direction around next year’s quarterly EA. In short, investors adjusted their initial underreaction over time, in particular around subsequent EAs, which leads to the PEAD.

More recently, several studies showed that the PEAD has survived the test of time (Collins & Hribar, 2000; Nichols & Wahlen, 2004). Other authors tried to explain, from a different point of view, the existence of the PEAD. One set of studies examined the role of investor’s information processing. For instance, Liang (2003) hypothesized and found that investors tend to overreact to private information compared to public information and put a higher weight on information that is less reliable. Consequently, investors underreact to the current EA. Francis, Lafond, Olsson and Schipper (2007) examined the PEAD in relation to information uncertainty. Investors delay their response to UE if there is more uncertainty related to those earnings. A portfolio consisting of stocks from firms with extreme UE is usually associated with higher information uncertainty than a portfolio of stocks from firms with small UE. Those stocks characterized by the highest amount of information uncertainty also showed higher abnormal returns than other stocks. Therefore, they concluded that information uncertainty reinforces investor’s underreaction at the EA date.

If underreaction is the explanation for the PEAD, arbitrageurs should take advantage of it and, consequently, eliminate the PEAD (Jacob, Lys, & Sabino, 2000). However, it seems that risk related to the application of such a trading strategy, so as arbitrage risk, impedes them from doing so. Indeed, Mendenhall (2004) found that arbitrage risk is positively associated with the level of the PEAD. However, as liquidity and trading activity increased over time due to technological

(13)

several anomalies, among them the PEAD (Chordia, Subrahmanyam, & Tong, 2014). Active institutional investors with the objective to take advantage from the PEAD ensure that the meaning of current earnings for future earnings is incorporated into a firm’s stock price in a more quickly manner (Ke & Ramalingegowda, 2005). Unfortunately, they trade less actively when transaction costs are high, wherefore the PEAD continued to exist. This is consistent with prior results from studies that point to transaction costs as an impediment to trade on the PEAD (Bhushan, 1994; Ng, Rusticus, & Verdi, 2008).

Multiple studies have examined whether the activity of a particular type of investor leads to the PEAD. Their results are, however, not unanimous. According to Bartov, Radhakrishnan and Krinsky (2000), the behavior of unsophisticated investors is the reason behind the PEAD. That is, they observed a negative association between investor sophistication, as measured by the fraction of a firm’s stocks hold by institutional investors, and the PEAD. Battalio and Mendenhall (2005) corroborated those findings by concluding that investors who carry out small transactions reply to a signal that does not fully incorporate the meaning of current earnings for future earnings. On the other hand, Shanthikumar (2004) confirmed that small traders underreact to a firm’s EA, but she found that large investors do so as well. Hirshleifer, Myers, Myers and Teoh (2008) added to this by stating that the PEAD does not exist purely because of trading by individual investors. Another set of studies discovered factors that lead to differences in the observed PEAD across firms. For instance, as showed by Narayanamoorthy (2006), investors fail to understand fully the effect of accounting conservatism on future earnings. An attribute of conservatism is that positive changes in earnings are more often persistent than negative changes (Basu, 1997). A trading strategy designed to exploit the PEAD generates even higher returns by taking into account the properties of accounting conservatism. Looking at the magnitude of the PEAD, we could infer that investors prefer UE originated by unexpected growth in revenues compared to unexpected cost reductions, consistent with the notion that higher revenues are more likely to be persistent (Jegadeesh & Livnat, 2006). The magnitude of the PEAD is reduced to a certain extent if a firm is followed by more experienced analysts (Mikhail, Walther, & Willis, 2003). However, Zhang (2008) stated that analysts are unsuccessful in using all of the information in a firm’s EA. This underreaction from analysts adds to the PEAD. On the other hand, Zhang (2008) found that when analyst responsiveness after a firm’s quarterly EA increased, the PEAD reduced significantly. According to Kimbrough (2005), holding a conference call leads to a significantly lower PEAD for smaller firms and a decrease in the delayed portion of stock price response to a firm’s EA.

(14)

Nowadays, researchers still found evidence for the existence of the PEAD, also outside the United States. For instance, Griffin, Kelly and Nardari (2010), based on data from 1994 through 2005, found evidence for the PEAD within 15 of the 38 countries they investigated. In addition, they showed that there is no difference in the magnitude of the PEAD between developed and emerging markets. Using data from shortly before and after the mandatory adoption of International Financial Reporting Standards in 2005, Hung, Li and Wang (2015) discovered the existence of the PEAD within a number of countries in their international sample. However, they also noticed that the size of the PEAD decreased over time. The PEAD was present as well within a large sample of Bird, Choi and Yeung (2014) between January 1986 and September 2009. However, as time went by, the PEAD became smaller in magnitude, potentially due to technological developments (Chordia et al., 2014).

2.2 The timing of the earnings announcement

Capital markets are sensitive to manager’s decisions about how and when to communicate firm-specific information (Miller & Skinner, 2015). Accordingly, it is well known that “managers spend considerable time thinking about how to manage their ﬁrms’ disclosures and that managers believe their disclosure decisions have ﬁrst-order value implications” (p. 222). Further, the International Accounting Standards Board (hereafter IASB) (2015) sees timeliness as one of the enhanced qualitative characteristics of useful financial information. They define timeliness as “having information available to decision-makers in time to be capable of influencing their decisions” (p. 32). According to the IASB, in most cases, older information is less useful.

In this section, the decisions around the timing of a firm’s EA and its influence on the subsequent stock price reaction will be discussed. The literature about the timing of a firm’s EA consists of three different forms of timing. The first and most studied form is about whether a firm released its EA on time, earlier or later than the date expected by investors or the date scheduled by the firm itself. The second form is about examining the differences between the nature and impact of EAs that take place on different days of the week or before, during or after market. The last form of timing investigated by prior literature has to do with a firm’s reporting lag, which is the number of days between the fiscal period end date and the EA date. Although prior literature mostly measured the stock price reaction over a short window around the date of the EA, the last category is of particular interest for the present study, because the objective of this study is to examine the effect of a firm’s reporting lag on the PEAD. However, all three forms have their relevant aspects for the current study.

(15)

With regard to the first form, prior research suggested numerous reasons why a firm would advance or delay its EA as compared to the expected EA date. In general, management must consider whether the benefits of delaying outweigh its costs (Begley & Fischer, 1998). Most benefits relate to management’s preference to postpone the release of bad news. In essence, by delaying the firm’s EA, management might warn the firm’s shareholders that it is going to announce bad news in the future (Kross, 1981). By doing so, they provide shareholders with the opportunity to sell their stocks before its release. Management also buys time to complete current negotiations or other contracts under current circumstances and has the possibility to improve the results by engaging in earnings management practices (Givoly & Palmon, 1982; Trueman, 1990). They also get the opportunity to think of arguments that could curb shareholder criticism and to work on a turnaround plan (Begley & Fischer, 1998). Through delaying, management has time to wait for the appearance of good news to disclose along with the bad news (Livnat & Zhang, 2015). There is also a chance that an earlier EA from a competitor contains even worse news. On the other hand, Begley and Fisher (1998, p. 357) argued that the incorporated news might not be the most important determinant of the timing of a firm’s EA. They found that the sign and magnitude of UE explained only 4% of a firm’s decision to advance or delay its EA. Moreover, they are cautious when it comes to the contemporary applicability of these motives, because of increased litigation risk and reputation costs. Based on Skinner (1994) and Francis, Philbrick and Schipper (1994), a firm’s incentives to delay bad news have declined, while there is more demand for extensive verification of good news.

Multiple authors studied the hypothesis that, as compared to the expected date, bad news' EAs are made later than expected, while good news’ EAs are released earlier than expected (Bagnoli et al., 2002; Begley & Fischer, 1998; Chambers & Penman, 1984; Givoly & Palmon, 1982; Haw, Qi, & Wu, 2000; Kross, 1981; Kross & Schroeder, 1984). All of them found evidence in support of the delayed dissemination of bad news. However, Bagnoli et al. (2002) and Givoly and Palmon (1982) found only weak support for the earlier than expected announcement for good news. In addition, Bagnoli et al. (2002) found a positive association between the amount of bad news and its delay, which was not observable in case of early EAs that contain good news. Therefore, investors could anticipate the arrival of bad news when a firm does not announce its earnings on the expected date. In line with their expectations, they observed a negative change in stock price from the expected date to the actual EA date. On the other hand, remarkably, Bagnoli et al. (2002) found no difference in abnormal stock return on the delayed actual EA date as compared to EAs made on time, which reveal the same amount of news. This indicates that, in total, the abnormal stock returns are more severe for later than expected EAs, because, as

(16)

mentioned before and different from on time EAs, investors already reacted to late EAs in the days between the expected and actual EA date. Furthermore, the abnormal stock return on early EAs is more pronounced in comparison to EAs made on time, regardless of whether the nature of the news is good or bad. That is why Bagnoli et al. (2002) conclude that the magnitude of the change in a firm’s stock price around the EA date is in part a function of whether the EA is earlier or later than expected.

Recently, Johnson and So (2017) used the fact that most firms publicly disclose their planned EA date. In this case, investors know whether the EA is going to be on time or not and thus, could form an expectation about the nature of the news. The results showed that firms who plan their EA date earlier than expected eventually reveal better news than those who deviate to a later date. This is consistent with the results of Livnat and Zhang (2015). However, advancers and delayers did not experience a different stock price reaction around the day they disclosed their future EA date. This means that despite being useful, investors do not immediately use the information as such. In the month thereafter, by contrast, advancers outperform the market, while delayers exhibit underperformance. The majority of the price reaction, however, still surrounds the actual EA date. Finally, Johnson and So (2017) showed that investor’s failure to respond directly to the signal provided by a firm when disclosing their intended EA date is not caused by not receiving the information on time.

The second form of timing is about whether managers release bad news EAs intentionally on a day or time when market attention is likely to be lower. Here, prior research showed that the stock price reaction on a firm’s EA date is lower on Fridays (DellaVigna & Pollet, 2009) and on days with more EAs from other firms (Hirshleifer, Lim, & Teoh, 2009). Other authors found contradictory evidence when it comes to the idea that EAs after trading hours do more often contain bad news as compared to EAs before the opening of the market (Doyle & Magilke, 2009; Patell & Wolfson, 1982). Further, the timing of an individual firm’s quarterly EA differs over time (deHaan, Shevlin, & Thornock, 2015). Most firms switch to another day of the week almost within every single year, while a quarter of all firms changed whether its EA takes place before, during or after market. However, most of these changes are legitimate. As a result, firms are able to switch from a certain day and time without getting unwanted attention from investors. According to deHaan et al. (2015), a decline in market attention is observable after market close and on busy days, but not on Fridays. Furthermore, they found that, in general, UE are lower for EAs made after trading hours, on Fridays and on busy days. It turns out that firms choose to announce bad (good) news during periods in which they expect lower (higher) market attention.

(17)

Other studies investigated the stock price reaction in the days around the EA with a short and long reporting lag. This is the third category of studies and the one that is of most interest for the present study. There are many different determinants of a firm’s reporting lag. Some studies indicated that large firms have a shorter reporting lag compared to small firms (Chambers & Penman, 1984; Givoly & Palmon, 1982). However, Givoly and Palmon (1982) concluded that, instead of firm characteristics, industry patterns and traditions determine to a larger part the timing of a firm’s EA. In addition, attributes like firm size, the quality of a firm’s internal control system, the complexity of its operations and the nature of the news are likely to influence the length of the audit, which could play an important role in the timing of a firm’s EA as well. Disagreements between a firm’s management and the auditor, either personally or content wise, could as well cause a delay in the EA (Bagnoli et al., 2002). Another important aspect to note is that management has to wait for divisional managers to provide the information needed (Kross, 1981). Hence, a higher organizational complexity could extend a firm’s reporting lag. Sengupta (2004) examined the drivers of a firm’s decision to release its EA earlier than others do. He found that firms confronted with higher investor demand for information and higher litigation cost have a shorter reporting lag. Longer reporting lags are associated with greater block ownership and operations that are more complex to grasp in accounting practices. Hence, firms try to avoid litigation and try to meet the demands of their shareholders by disseminating private information in a timely manner. Givoly and Palmon (1982) showed that the average reporting lag has become shorter over time. This either could be due to developments in the use of computers, a higher auditor involvement or increased investor demand. Overall, the length of a firm’s reporting lag could be influenced by many different factors, of which some are controllable by the firm itself. Others are, however, beyond their control.

Ball and Brown (1968) noted that stock prices already include a large part of the accounting information before the earnings release. One might expect that private information from a firm’s EA leaks to the market and that other information channels disseminate a large part of the news when the reporting lag becomes longer (Givoly & Palmon, 1982). Competitors could as well influence the informativeness of a firm’s EA. For example, prior literature showed that the release of the EA of a particular firm leads to a change in stock price of other firms in the same industry, consistent with the idea that some of the information in a firm’s EA is transferable to other firms operating in the same industry (Foster, 1981; Han & Wild, 1997). Hence, Givoly and Palmon (1982) argued that when a firm’s reporting lag becomes longer, more EAs have already taken place within the same industry, such that the amount of new information revealed by a firm’s own EA would be lower. On the other hand, they discovered that most of the information disseminated by

(18)

other sources is disclosed shortly before or after a firm’s expected EA date. Nevertheless, Givoly and Palmon (1982) reported a difference in the magnitude of the market response between EAs after a short and long reporting lag. In fact, they showed that early annual EAs led to more pronounced subsequent price movements during the week of the EA compared to late EAs. A few years later, Chambers and Penman (1984) proposed that the market is able to form expectations about the earnings report, such that the stock price reaction may be dependent on the reporting lag. They also argued that other sources distribute more of the information from the report when the reporting lag becomes longer. As an example, they mentioned search activity by investors, other voluntary disclosures or estimates made based on the earnings reported by other firms. Therefore, they expected that less timely reports are associated with less price variability around the EA date than more timely reports. However, they found no significant relationship between reporting lag and the variability of stock returns related to annual and quarterly earnings releases. As a result, they concluded that accounting reports include some firm-specific information that is not distributed by other sources.

2.3 Timing and the post-earnings-announcement drift

To date, many researchers examined the effect of the timing of a firm’s EA on the short-term stock price response. There is only limited evidence on its relationship with the PEAD. According to Chambers and Penman (1984), the subsequent stock price reaction during the fifty days after a firm’s EA was extraordinarily high after the release of unexpectedly early EAs presenting good news and unexpectedly late EAs containing bad news. In this case, the expected EA date was determined by looking at last year’s EA date. The other way around, this effect was not observable. In a subsequent study, Penman (1984) tried to produce an implementable trading strategy based on those results. Taking a short position in stocks of firms, which report earnings at a later than expected date, could be profitable, because abnormal returns on those stocks would be negative during a period up to fifty trading days after the actual EA date. In addition, the advancing of a good news EA might lead to a positive drift in abnormal stock returns over the next fifty trading days, such that a long position in those stocks should be profitable as well. Moreover, Johnson and So (2017), showed that firms that have set their planned EA date to an earlier than expected date obtained positive abnormal returns in the two weeks after the EA, while those firms who delayed their EA date had negative abnormal returns during this period. Livnat and Zhang (2015) conducted a similar study and found the same results over a period starting two days after a firm scheduled its EA date to ninety days after the EA.

(19)

Another set of studies observed the effect of investor’s attention on the PEAD. For example, DellaVigna and Pollet (2009) hypothesized and found that EAs on a Friday do receive less initial attention from investors in comparison to EAs made on any other day of the week. As a result, a proportionately larger part of investor’s response took place after the EA date, which in turn caused a higher PEAD. In a similar study, Hirshleifer et al. (2009) examined whether a firm’s PEAD is influenced by the amount of other EAs taking place on the same day. Consistent with the idea that investors can process only a limited amount of new information at once, they found a lower price reaction on the EA date and a more pronounced PEAD when a firm’s EA took place on a day with multiple EAs from other firms.

Givoly and Palmon (1982) did not only examine the stock price reaction in the week of the EA, but also in the surrounding weeks. Consistent with the results from the EA week, they found a higher stock price reaction on annual EAs with a short reporting lag during the two weeks after the EA week. After two weeks, the stock price change of EAs with a longer reporting lag is more severe. Overlap between the periods of the publication of late annual EAs and the distribution of first quarter EAs could be a possible explanation for this finding. Besides the findings mentioned earlier, Chambers and Penman (1984) explored whether the variability in abnormal return at the reporting date temporarily persisted in the fifty days after the EA. By doing so, they provided preliminary evidence on the effect of the timing of a firm’s EA on the PEAD. They found significant stock price changes after the EA date for those EAs that also had a significant impact on stock price at their release date, but not for those EAs that led to little or no changes in stock price on the reporting date. Moreover, the stock price reaction after a firm’s EA is more severe for EAs characterized by bad news. Accordingly, there is only a minimal stock price reaction after EAs with a relatively short reporting lag, while those EAs with a longer reporting lag are followed by strong stock price changes. This was inconsistent with their beliefs. They expected to observe a higher stock price change after EAs with a short reporting lag, because of the subsequent publication of EAs from other firms, which could distribute relevant information.

2.4 Changes in information content and value relevance over time

Prior research showed that “No other single event has been found to explain more of the cross-sectional variation in stock returns than the earnings announcements” (Dechow, Sloan, & Zha, 2014, p. 344). Nevertheless, there is a lively debate about the informativeness of EAs over the last decades. Beaver (1968) was the first to discover the existence of higher trading volume and stock price changes at a firm’s annual EA date. These findings brought him to the conclusion that the availability of other information prior to a firm’s annual EA did not make it completely redundant.

(20)

However, Ball and Brown (1968, pp. 175-176) demonstrated that, even before its release, the market received almost all information included in a firm’s earnings number. In fact, only 10% to 15% is unknown. Moreover, of all the information flowing to the market during the announcement month, only 20% arises from information included in a firm’s annual earnings number.

More recently, Lev and Zarowin (1999) observed a reduction in the usefulness of earnings, cash flows, and book values over the past twenty years, while Brown, Lo and Lys (1999) found evidence for a drop in the value relevance of accounting earnings and book values of equity over the last forty years. On the other hand, Landsman and Maydew (2002) did not find a deterioration in the information content of quarterly EAs over the last thirty years. In contrast, their results indicated an increase in the information content over time. Others proposed an explanation for these results. Following Francis, Schipper and Vincent (2002b), it is a result of more detailed EAs through the inclusion of more information, like detailed income statements. The findings of Collins, Li and Xie (2009), however, indicated that the information content of quarterly EAs has increased over time, because of a more powerful reaction to a firm’s Street earnings. These are non-GAAP earnings metrics provided by analysts simultaneously with a firm’s EA. Ball and Shivakumar (2008) found that circa 6% to 9% of the total annual information is associated with a firm’s quarterly EAs. That is why they believe that disseminating timely new information to the market is not the main role of reported earnings. They concluded that quarterly EAs distributed “a modest but not overwhelming amount of incremental information to the market” (p. 975). Contrary to what one would expect, in the weeks before and after EAs there is no overwhelming flow of information. Management forecasts and analyst forecast revisions prior to EAs appear to be more important information sources. On the other hand, Francis, Schipper, and Vincent (2002a) did not find that the informativeness of EAs has decreased due to competing information disseminated by analyst reports. Instead, they found evidence for a complementary relationship between those two.

Two recent studies, based on data collected over a long period, confirmed the conclusions discussed above. Accounting information became less value relevant for equity holders within a sample from 1985 to 2013 (Givoly, Hayn, & Katz, 2017). On the other hand, however, the information content of quarterly EAs increases between 1971 and 2011, especially after 2001 (Beaver, McNichols, & Wang, 2018).

2.5 Changing times

(21)

impacts this may have on the conclusions drawn in prior literature. Besides publishing regulated financial reports, firms are free to make voluntary additional disclosures like non-GAAP management forecasts and conference calls (Healy & Palepu, 2001). In addition, information intermediaries, such as financial analysts, industry experts and the financial press, disclose information about a firm. Overall, the production, dissemination and processing of firm-specific information has changed due to recent developments in the field of information technology, the media and capital markets (Miller & Skinner, 2015). As Healy and Palepu (2001) pointed out, a striking difference with the disclosure landscape of the last century is the availability of computers and the rise of the internet. These innovations made it possible for firms to tap into new channels to communicate with investors and the public. For instance, the internet allows a firm to keep in touch with all investors and inform them about any type of important information at any time. Most firms use their website to share all kinds of firm related information, like press releases and analyst forecasts. According to Healy and Palepu (2001), it is likely that investors use the internet more often, wherefore firms tend to provide more voluntary disclosures on the internet. From a firm’s point of view, engaging in more voluntary disclosure and using more information channels could lead to higher stock liquidity, a lower cost of capital and higher analyst following. In addition, the disclosure of more firm-specific information reduces stock price comovement and results in a more informative stock price (Haggard, Martin, & Pereira, 2008).

One of the advantages is that “internet reporting enables information to be disseminated to a wide range of stakeholders in a timely and convenient manner” (Lodhia, Allam, & Lymer, 2004, p. 64). One set of studies investigated the consequences of a firm’s online disclosure practices. For example, providing higher quality web-based non-financial voluntary disclosures is negatively associated with information asymmetry and the cost of finance of continental European firms (Orens, Aerts, & Cormier, 2010). Similar, a higher level of Internet-based voluntary disclosure by French listed firms lowers their information asymmetry (Gajewski & Li, 2015). On the other hand, investors use the internet to search for available information about a firm in order to meet their information demand (Drake, Roulstone, & Thornock, 2012). Within the weeks around a firm’s quarterly EA, the authors found evidence for abnormal Google search volume on a firm’s ticker symbol, with the highest value measured on the EA date. Moreover, they showed that a higher search volume in the weeks before a firm’s quarterly EA leads to a lower stock price reaction at the EA date.

Moreover, other developments started to play a role in the supply of corporate information. Nowadays, the media uses all kinds of smart algorithms controlled by computers that give them the opportunity to collect and distribute news in real-time (Miller & Skinner, 2015).

(22)

Prior literature showed that higher press following can reduce information asymmetry during a firm’s EA (Bushee, Core, Guay, & Hamm, 2010). In addition, stock returns are larger for firms that got less attention from mass media (Fang & Peress, 2009). Miller and Skinner (2015) argued that firms use social media as an accessible platform to communicate with investors as part of their disclosure strategy. On the other hand, they pointed out that the emergence of social media made a firm’s information environment less controllable, because outsiders could use it as well to share their views and content about a particular firm to the public. This could lead to undesirable effects, especially if people who share information are relatively uninformed. The effects of these innovations are strengthened by a tremendous growth in mobility and the continuous availability of devices connected to the internet, which allow investors to obtain information whenever they want. Technological developments also had their influence on capital markets when it comes to trading activity and the speed with which news is discounted into stock prices by, for example, by the rise of high-frequency trading.

Due to its increased use and importance, academic researchers started to pay attention on the effects of social media on capital markets. For instance, one can use different social media measures, reflecting the nature of the content posted on different forms of social media, to predict a firm’s future equity value (Luo, Zhang, & Duan, 2013; Yu, Duan, & Cao, 2013). Of all messages posted by a sample of firms on Facebook and Twitter from 2009 to 2013, 7.06% and 3.45% respectively relate to corporate disclosures (Zhou, Lei, Wang, Fan, & Wang, 2015, p. 25). More specifically, 16.8% and 30.24% of these corporate disclosures are financial disclosures. Twitter makes it possible for a firm to supply investors with information, instead of investors having to look this up themselves (Blankespoor, Miller, & White, 2014). Within a sample of technology firms, information asymmetry lowered when firms started to share a hyperlink to their press releases via Twitter. These results were stronger for less visible firms, consistent with the idea that these firms look for disclosure channels outside the traditional media in order to keep in touch with investors. Moreover, Lee, Hutton and Shu (2015) found that the use of social media could contribute to a less negative stock price reaction on a firm’s product recall announcement. However, by looking especially at the use of Twitter, they found this effect to be dependent on the amount of control lost by a firm over its information environment due to comments from outsiders and the ability of a firm to respond to these comments. In their experiment, Trinkle, Crossler and Bélanger (2015) found that comments related to a firm’s financial disclosures on social media posted by other investors had influence on the opinion and decision making of nonprofessional investors. The impact of good news is softened by negative comments, while bad news is viewed less negative when combined with positive comments. Somewhat similarly, the

(23)

nature of the content posted on Seeking Alpha, a popular social platform for investors and analysts to share their opinions, is associated with a firm’s stock price movement and UE after its publication (Chen, De, Hu, & Hwang, 2011).

Based on the evidence discussed above, it seems that due to technological developments the availability of firm-specific information has increased. The internet and in particular the rise of social media created new channels for firms to keep in touch with investors and the public at large. In addition, investor’s possibilities to share their own views and discuss with others became more extensive. These changes should lead to less uncertainty among investors prior to and after a firm’s EA, such that the underreaction around a firm’s EA should decrease. In addition, investors should attach less value to and be less dependent of a firm’s EA, because of a higher availability of firm-specific information via other channels. Collectively, these developments have the ability to lower the magnitude of a firm’s PEAD.

(24)

3 Hypothesis development

In this chapter, I describe and substantiate the hypotheses in order to answer the research question, by means of the literature reviewed in Chapter 2. Figure 1 below illustrates a graphical representation of the PEAD window examined in this study, showing which timeframes are of interest with respect to the hypotheses. I included “The Predictive Validity Framework” (Libby, 1981) in the Appendix section in order to offer an overview of the concepts involved in each hypothesis.

Figure 1: Overview of timeframes

Results from prior literature are contradictory about the effect of the length of a firm’s reporting lag on the subsequent stock price reaction (Chambers & Penman, 1984; Givoly & Palmon, 1982). However, to date, as far as I know, no similar research exists which specifically examines the effect of a firm’s reporting lag on the PEAD. I believe that due to rapid technological developments during the last years (Healy & Palepu, 2001; Miller & Skinner, 2015) the overall amount of information available to investors has increased tremendously. Over time, it has become much easier for investors to obtain relevant information in a timely manner, because of the rise of, for instance, the internet and social media. This is not limited to information provided by the firm itself. Other market participants can produce information and opinions as well. As shown by prior research, the value relevance of earnings decreased over time (Brown et al., 1999; Lev & Zarowin, 1999), while, the information content of EAs increased (Landsman & Maydew, 2002). This could indicate that the earnings number itself is not as important anymore as other information disclosed at the EA date. Thus, other information could serve as a replacement for a firm’s earnings number. The content of the other information may be better predictable well in advance by other market participants.

I expect that as the amount of days between the end of a firm’s fiscal period and its EA date becomes longer, investors obtain more information included in a firm’s EA in advance by means of various information channels. As a result, the amount of new information revealed by a

(25)

Therefore, for those firms I expect to observe a decrease in the magnitude of the PEAD. This leads to the hypothesis formulated below. In Figure 2, I plotted the expectations underlying this hypothesis in a graph for both positive and negative PEADs.

H1: The post-earnings-announcement drift is smaller in magnitude for firms with a longer reporting lag

compared to firms with a shorter reporting lag.

Figure 2: Expectations hypothesis 1

Most studies argued that the PEAD is a result of an initial underreaction from investors on the information revealed by a firm’s EA. Investors delay their response, because they are unable to understand immediately the consequences of current earnings for the realization of future earnings (Bernard & Thomas, 1989; 1990). Therefore, they wait for other information to confirm their initial thoughts. As a result, the part of the PEAD taking place within a relatively short window around the EA of the next fiscal quarter is much higher than if the PEAD would grow monotonically over time (Bernard & Thomas, 1989). When the reporting lag becomes longer, I expect the market to receive less new information at the EA date and from a firm’s EA. Therefore, the uncertainty among investors immediately after a firm’s EA should be lower. They are less surprised, and hence, need less time to process the remaining information. Combined, less new information and a lower level of uncertainty should result in less underreaction from investors. It is less complicated and time-consuming for investors to form expectations about a firm’s future earnings and to determine the appropriate value of a firm’s stock immediately after a firm’s EA. Therefore, investors should have fewer incentives to wait for a firm’s next quarterly EA to confirm their initial thoughts.

While I want to examine the magnitude of a firm’s complete PEAD with the first hypothesis, the objective of H2a and H2b is to discover the magnitude of the PEAD in two

-10% -5% 0% 5% 10% 0 10 20 30 40 50 60 70 80 90 100 M agn it ud e of th e P EA D

Length of reporting lag

Positive PEAD Negative PEAD

(26)

different timeframes. Specifically, I want to test whether the length of the reporting lag affects the absolute part of the PEAD taking place at the beginning and at the end of the PEAD window. Based on the arguments above, I expect the magnitude of the PEAD occurring within the first five trading days of the PEAD window to be larger when the reporting lag becomes longer. On the other hand, the magnitude of the PEAD being observable within the three trading days around the subsequent quarterly EA would be smaller. This leads to the two hypotheses formulated below. Figure 3 provides a graphical expression of the expectations underlying those hypotheses. Here, I assume the magnitude of the total PEAD to be 10%. Note that the numbers used in this figure are not part of my expectations, but are only there for clarification purposes. Together, the three blocks represent the complete PEAD, which I test with H1.

H2a: The magnitude of the post-earnings-announcement drift occurring within the first five trading days is

larger for firms with a longer reporting lag compared to firms with a shorter reporting lag.

H2b: The magnitude of the post-earnings-announcement drift occurring within the three trading days around

the subsequent quarterly earnings announcement is smaller for firms with a longer reporting lag compared to firms with a shorter reporting lag.

Figure 3: Expectations hypothesis 2a and 2b

(H2a) (H2a) (H2a)

(H2b) (H2b) (H2b) 0.0% 2.0% 4.0% 6.0% 8.0% 10.0%

Short reporting lag Medium reporting lag Long reporting lag

M agn it ud e of th e P EA D

(27)

4 Research method and design

In this chapter, I explain the research method designed to answer the research question. In Section 4.1, I start with a description of the initial sample and the sample selection criteria applied to obtain the final sample. I also refer to the data sources used to retrieve my data. The way I operationalized my variables is described in Section 4.2. In Section 4.3, I explain the statistical model that I used to test my hypotheses.

4.1 Sample selection

While prior literature is mostly built on samples from the United States, this study use firms from STOXX Europe 600 Index. This index consists of the 600 largest firms, based on free-float market capitalization, from 17 European countries (STOXX Limited, 2018a; 2018b). Together, they represent approximately 90% of the total European free-float market capitalization. Every quarter, a review of the composition of the index takes place.

The sample consists of quarterly EAs related to fiscal quarters between 1 January 2011 and 31 December 2016 from all firms that are part of the STOXX Europe 600 Index anywhere during this period. By starting in 2011, I exclude the potential effects of the financial crisis on my research question. When a firm is deleted from the STOXX Europe 600 Index, its quarterly EAs are also completely dropped from the sample. By doing so, the sample consists of approximately the same amount of observations every fiscal quarter.

I collect the components of the STOXX Europe 600 Index within the sample period from the Compustat Global database. Due to additions and deletions from the STOXX Europe 600 Index, 809 unique firms fall into the initial sample. I use the database I/B/E/S to retrieve data about quarterly EA dates. In order to keep the data manageable, I delete observations from firms of which the fiscal quarter end dates are not equal to calendar quarter end dates. Theoretically, the initial sample consists of a maximum of 14,400 observations of quarterly EAs (600 firms * four quarterly EAs* six fiscal years). However, after collecting data about EA dates, only 7,932 observations from 464 firms are left.

Besides the EA dates, I collect data on analyst forecasts from I/B/E/S. The stock market and financial statement data I retrieve from Compustat Global. The historical prices of the STOXX Europe 600 Index were available on the website of The Wall Street Journal. As the sample consists of European firms listed in different countries, the raw data is recorded in various currencies. To make fair comparisons between firms, I convert all data to Euros by using the

(28)

exchange rates applicable on the end of each fiscal quarter. I gather the daily exchange rates from I/B/E/S.

Due to the unavailability or incompleteness of data, I needed to drop another 1,998 observations from 83 unique firms from the sample. This leads to a final sample of 5,934 unique quarterly EAs from 381 firms. Table 1 below summarizes the full sample selection procedure. Table 1: Sample selection procedure

Quarters Firms

Maximum number of observations theoretically 14,400 809

Exclusions:

- Missing EA date data from I/B/E/S (6,468) (345)

- Missing financial statement data from Compustat (199) (4) - Missing actual and/or forecast EPS data from I/B/E/S (1,720) (72)

- Missing stock price data from Compustat (79) (7)

Remaining number of firm quarter observations 5,934 381

4.2 Operationalization of variables

The total set of variables consists of three types: the independent, dependent and control variables. I include control variables in line with prior literature. In order to verify the robustness of the statistical results (Section 5.4), I choose to operationalize the dependent variable, CAR, and the first control variable, UE, in two ways.

4.2.1 Independent variable

The independent variable employed is reporting lag, which is the number of days between the end of a firm’s fiscal quarter and its quarterly EA date (Chambers & Penman, 1984; Givoly & Palmon, 1982; Sengupta, 2004).

𝑅𝑒𝑝𝐿𝑎𝑔𝑖,𝑡= 𝐸𝑎𝐷𝑎𝑡𝑒𝑖,𝑡− 𝐸𝑛𝑑𝐷𝑎𝑡𝑒𝑖,𝑡

Where:

𝑅𝑒𝑝𝐿𝑎𝑔𝑖,𝑡 = the number of days between the end of the fiscal quarter and the

quarterly EA date of firm i in quarter t.

𝐸𝑎𝐷𝑎𝑡𝑒𝑖,𝑡 = the quarterly EA date of firm i in quarter t.

(29)

4.2.2 Dependent variable

I construct three dependent variables, one for each hypothesis. Hence, I calculate CARs to measure the PEAD. However, there are many different methods to compute abnormal returns. Since there is no right model to determine expected returns, I use two different methods. One of them is the size-adjusted approach, as proposed by Foster et al. (1984). With this approach, I compare a firm’s daily return to the daily return of a portfolio consisting of firms of approximately the same firm size. The difference between a firm’s raw return and the size-adjusted portfolio return is a firm’s abnormal return for that day. The formula is as follows:

𝑆𝐴𝐴𝑏𝑛𝑅𝑒𝑡𝑖,𝑡= 𝑅𝑎𝑤𝑅𝑒𝑡𝑖,𝑡− 𝑃𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜𝑅𝑒𝑡𝑖,𝑡

Where:

𝑆𝐴𝐴𝑏𝑛𝑅𝑒𝑡𝑖,𝑡 = size-adjusted abnormal return of firm i on day t.

𝑅𝑎𝑤𝑅𝑒𝑡𝑖,𝑡 = raw return of firm i on day t.

𝑃𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜𝑅𝑒𝑡_𝑖,𝑡 = equally weighted mean return of firm size decile i on day t of which firm

i is a member at the start of the fiscal quarter.

Because portfolios are equally weighted, the daily portfolio return is equal to the mean of all daily raw returns from firms within the firm size decile. I form firm size deciles by sorting on market capitalization at the start of each fiscal quarter. In order to test H1, I sum abnormal returns from two days after a firm’s quarterly EA date to one day after the next quarterly EA date (𝑇𝑜𝑡𝑎𝑙𝐶𝐴𝑅𝑖,𝑡).

I measure until after the next EA date since a relatively large part of the PEAD takes place during a short-term period around the next quarterly EA (Bernard & Thomas, 1989).

As robustness test, I perform the analysis again while using the Market-Model to calculate abnormal returns (Strong, 1992). To do so, I estimate abnormal returns by regressing a firm’s daily raw return on the return of a market portfolio, in this case the STOXX Europe 600 Index. The estimation period consists of the trading days from the most recent fiscal quarter. The formula of the Market-Model states:

𝑅𝑎𝑤𝑅𝑒𝑡𝑖,𝑡= 𝛼𝑖+ 𝛽𝑖𝑀𝑘𝑡𝑅𝑒𝑡𝑡+ 𝜀𝑖

Where:

𝑅𝑎𝑤𝑅𝑒𝑡𝑖,𝑡 = raw return of firm i on day t.

𝛼𝑖 = firm-specific constant.

𝛽𝑖 = firm-specific beta.

𝑀𝑘𝑡𝑅𝑒𝑡𝑡 = raw return of the STOXX Europe 600 Index on day t.

(30)

After I obtain the estimates of a firm’s constant and beta from the regression, I derive abnormal returns with the following formula:

𝑀𝑀𝐴𝑏𝑛𝑅𝑒𝑡𝑖,𝑡 = 𝑅𝑎𝑤𝑅𝑒𝑡𝑖,𝑡− (𝑎̂𝑖+ 𝛽̂𝑖𝑀𝑘𝑡𝑅𝑒𝑡𝑡)

To test H2a and H2b, the timeframe for accumulating abnormal returns is shorter. In H1, I examine the magnitude of the total PEAD. In H2a and H2b, I concentrate on the part of the PEAD occurring within two short-term periods, namely the first five trading days (𝑆𝑡𝑎𝑟𝑡𝐶𝐴𝑅𝑖,𝑡)

of the drift window and the three trading days around the subsequent EA date (𝐸𝑛𝑑𝐶𝐴𝑅𝑖,𝑡).

Calculating the magnitude of the PEAD taking place during shorter timeframes is something that is done in prior literature as well. For example, Bernard and Thomas (1989) scrutinized which percentage of the PEAD is taking place at the start of the PEAD window and around the following quarterly EA. They found that the arrival of the PEAD is far from constant, with higher than expected parts being observable at the start and around the subsequent quarterly EA date.

4.2.3 Control variables

Unexpected earnings

UE is the first control variable included, because prior literature found evidence for a positive association between the sign and magnitude of a firm’s UE and the PEAD (Bernard & Thomas, 1989; Foster et al., 1984; Rendleman et al., 1982). Usually, two different measures are employed. First, the standardized UE measure of Foster et al. (1984), which Bernard and Thomas (1989; 1990) and many others as well. In this case, I assume quarterly earnings to follow a seasonal random walk model. Therefore, I calculate UE as the difference between a firm’s quarterly earnings and its respective quarterly earnings of four quarters ago. Finally, I scale UE by the market value of equity at the end of the fiscal quarter, which is consistent with, for example, Ng et al. (2008). I use data on a firm’s quarterly income before extraordinary items as a proxy for its quarterly earnings. The calculation is summarized by the formula below:

𝑈𝑛𝐸𝑎𝑟𝑖,𝑡=

𝑄𝑖,𝑡− 𝑄𝑖,𝑡−4

𝑀𝑉𝑖,𝑡

Where:

𝑄𝑖,𝑡 = realized quarterly earnings of firm i in quarter t.

𝑄𝑖,𝑡−4 = realized quarterly earnings of firm i four quarters prior to quarter t.

𝑀𝑉𝑖,𝑡 = market value of equity of firm i at the end of quarter t.

I use a second measure to compute a firm’s UE as a robustness test. In this case, I compare a firm’s actual quarterly EPS with the consensus forecast from analysts and scale the difference by

The effect of a firm’s reporting lag on the post-earnings-announcement drift