Cutting Through Complexity: Segment Disclosure and Pricing Efficiency

Abstract

In 1997, Statement of Financial Reporting Standards (SFAS) 131 introduced a substantial change in how segment information is reported in US-GAAP (Generally Accepted Accounting Principles) financial statements. We seek to examine whether this change in financial reporting policy increases investors’ ability to process information relevant to conglomerate firms (i.e., those operating in multiple industries) more quickly, and whether this increased efficiency in information processing varies cross-sectionally based on firm complexity and the direction of industry news. Our results indicate that SFAS 131 has increased the speed with which stock prices capture information about conglomerates, relative to focused firms, although we find that some frictions remain with regard to disclosing bad news. This study documents an example of how a change in disclosure policy can enhance pricing efficiency, and hence, it may be of interest to the Financial Accounting Standards Board (FASB) in its ongoing initiative to consider potential changes to SFAS 131.

Keywords

return predictability segment reporting SFAS No. 131 conglomerate market frictions information processing

Introduction

In this article, we examine how Statement of Financial Reporting Standards (SFAS) No. 131, which changed segment reporting requirements, may have altered the processing and impounding of industry information into firm values for firms with multiple operating segments.¹ We use the novel approach of Cohen and Lou (2012, CL hereafter), which distinguishes between two types of firms: one set that operates in a single industry (focused or standalone firms) and requires simple information processing and one that has multiple industry segments (conglomerate firms) and requires more complicated analysis to impound the same piece of information into price. The authors show that using this straightforward versus complicated information processing classification leads to predictability in the cross-section of returns. We exploit this property and study whether SFAS 131 increases investors’ ability to process information relevant to conglomerate firms and whether this increased efficiency in information processing leads to increased price efficiency for conglomerates. Furthermore, we investigate how the complexity of a firm’s operations, its level of disaggregation, and the direction of industry information (i.e., good vs. bad news) affect this complicated information processing before and after the standard went into effect.

The premise of CL is that complexity in information processing leads to a significant delay in the impounding of information into prices. Therefore, information is first reflected in the easy-to-analyze (focused) firms, relative to complicated (conglomerate) firms, which in turn generates predictability in the returns of conglomerates. We follow CL’s method and construct “pseudo-conglomerate” portfolios for each conglomerate in our sample, where a “pseudo-conglomerate” is a portfolio of the conglomerate firm’s industry segments constructed using solely focused firms in each industry, weighted by the percentage of sales contributed by each industry segment within the conglomerate. To illustrate how a pseudo-conglomerate is constructed, consider ABC Corp., a hypothetical conglomerate that sells autos, bacon, and candy, where 50%, 30%, and 20% of ABC’s revenues are generated by its auto, bacon, and candy segment, respectively. The pseudo-conglomerate return for ABC would be (50% × the portfolio returns for focused auto firms) + (30% × the portfolio returns for focused bacon firms) + (20% × the portfolio returns for focused candy firms). The portfolio returns for each group of focused firms is calculated as the value-weighted (VW) average return for the standalones in the respective two-digit Standard Industrial Classification (SIC) industry. As these pseudo-conglomerates are composed of easy-to-analyze focused firms subject to the same industry shocks as their corresponding conglomerate firms, focused firms’ prices should be updated and should reflect the information first. Consequently, the returns of the pseudo-conglomerate portfolios predict the future returns of their corresponding conglomerate firms.

Following this approach, we examine whether SFAS 131 has an effect on the documented conglomerate return predictability; specifically, we are interested in whether the different disclosure requirements influence investors’ ability to process complex information and to impound it into prices of complex firms operating in multiple industries. We split the sample into pre (1977-1996) and post (2000-2016) regulation period and show that the conglomerate firms’ return predictability is driven by the pre SFAS 131 period and mostly disappears in the post period. This result suggests that the regulation is effective in alleviating the delay in complicated information processing associated with conglomerate firms. More complicated conglomerates are associated with an incremental information processing delay prior to SFAS 131, but this differential effect disappears after the standard went into effect. Interestingly, we also find that the return predictability is driven by bad news, and some of the information processing frictions related to negative news continue to exist, even after the implementation of SFAS 131. This implies that the regulation was not completely successful at eliminating all difficulties associated with impounding complex information into prices of conglomerates. Therefore, while SFAS increased the speed of information processing for good as well as bad news, as bad news were processed more slowly prior to the regulation, some predictability remains in the post period for bad news.

This article presents a new approach to studying the effect of SFAS 131, and it is the first to document that the regulation affects the speed with which stock prices capture value-relevant information about conglomerates, although the effect varies in the cross-section. Our investigation is important because if the standard helps prices capture information about complex firms quicker, it likely improves market efficiency and resource allocation. However, our article shows that the anomaly may still exist in the presence of bad industry-level news. Although the standard was implemented in 1996, our inquiry is timely, because in June 2019, the Financial Accounting Standards Board (FASB) asked “public companies to participate in a study that focuses on potential improvements to the segment disclosure requirements.”² Subsequently, FASB held meetings and issued decisions related to segment reporting disclosures as recently as May 2021 and intends to consider further analysis of the issues in future meetings.³ Our article provides evidence on how the current rule affects existing frictions and constraints that impede information processing and hence may be of interest to the FASB in its consideration of potential standard improvements.

Our article makes several important contributions to the literature. First, we add to the literature on biased or delayed information processing, which generally shows that if investors are constrained by their limited resources and capacity to process available information, asset prices will impound the information only gradually. To this end, we add to research on the post earnings announcement drift (PEAD; e.g., Bernard & Thomas, 1990; Livnat & Mendenhall, 2006), investor inattention (e.g., DeHaan et al., 2017; DellaVigna & Pollet, 2009), and the literature on investor under- and overreaction to public information (e.g., Bernard & Thomas, 1989; Hui & Yeung, 2013; Zhang, 2006). These studies generally find that investors’ inefficient processing of available information leads to significant return predictability based only on publicly available information. In that vein, CL find a noteworthy return predictability of conglomerate firms’ returns, and we add to this research by documenting that this anomaly has mostly disappeared following SFAS 131, except, to some extent, in the case of bad news.

Second, we add to the literature on the effects of SFAS 131. Prior research finds that the regulation improved both the overall information environment (Berger & Hann, 2003; Botosan & Stanford, 2005; Ettredge et al., 2005; Park, 2011) and managers’ behavior by increasing loss recognition timeliness (Bens et al., 2018) and improving managers monitoring (Berger & Hann, 2003; Cho, 2015). We add to this research by showing that SFAS 131 enhanced the speed with which investors can process information about complex firms, but more importantly, we provide evidence on how this effect varies cross-sectionally based on firm complexity, level of disaggregation, and industry news.

Third, this article provides evidence of how a change in financial disclosure can improve market efficiency and as such, contributes to the literature on the effect of disclosure quality on valuation (e.g., Akins, 2018). Specifically, we add to Hope et al. (2008) that shows that mispricing of foreign earnings decreases and to Kang et al. (2017) that documents a decrease in PEAD following increased disclosure under SFAS 131. Although our article confirms the beneficial effect of the rule, it also raises the concern that some market frictions remain related to the direction of industry news, which should be of interest to FASB in evaluating further regulation related to segment reporting.

Finally, our article also complements the vast literature related to diversification discounts. Focusing on the valuation differences between diversified and focused firms, this literature originally contended that diversified firms are consistently worth less than the sum of their parts (Berger & Ofek, 1995). In a recent article, Hund et al. (2021) reexamine the question and show that the diversification discount is in fact an artifact of the methodology used in the previous literature. Using a newly created measure, the authors show that focused and diversified firms have comparable values. Similar to the aforementioned literature, we examine conglomerate firms and how their prices respond to information. However, in contrast to this literature that focuses on valuation, our results are cross-sectional among diversified firms, in the sense that we are exploring how diversified firms respond to industry information shocks (in the same vein as CL).⁴

The rest of the article is organized as follows. “Background and Hypothesis Development” section provides background information and develops the hypotheses. “Data and Methodology” section shows the research design, and “Empirical Results” section discusses the empirical results. Finally, “Discussion of the Results and Robustness Checks” section discusses robustness checks, and “Conclusion” section concludes.

Background and Hypothesis Development

Background on Segment Reporting

Based on a survey of over 100 financial analysts, Epstein and Palepu (1999) report that analysts consider segment performance data as the most useful for their investment decisions. Consequently, it is not surprising that the transition from SFAS 14 to SFAS 131 was an important event for U.S. capital markets. Starting in 1976, under SFAS 14, publicly traded firms were required to report information such as assets, sales, earnings, and cash flows from operations separately for each industry segment using the industry approach. A major concern with SFAS 14 was that the discretion in the definition of “industry” allowed many firms to report fewer segments to external users than what was reported internally. SFAS 131 superseded SFAS 14 and became effective in 1997, for fiscal years beginning after December 15, 1997. Under SFAS 131, firms are required to report segments based on the management approach, with which segment information is presented based on the segments that management uses internally to evaluate the operating performance of its business units. As a result, firms increased the number of reported segments and the disaggregation of their data, which arguably improved the amount and quality of reported information under SFAS 131 (Berger & Hann, 2003).

In an early study assessing the effects on SFAS 131, Street et al. (2000) document that overall, the standard improves segment reporting. Chen and Liao (2015) focus on the extent to which firms comply with disclosure requirements under SFAS 131 as a measure of segment disclosure quality and find that it is negatively related to firm’s cost of debt documenting a benefit to complying with the rule. Consistently, Herrmann and Thomas (2000a) show significant changes in segment disclosure following the regulation, including an increase in the number of firms providing segment disclosure and an increase in the number of disclosed items for each segment.

Although SFAS 131 has clear advantages over the previous regulation, the standard also has a number of weaknesses. For example, one concern with imposing increased disaggregation of disclosed data is an increase in proprietary costs. Empirical research confirms that such costs drive segment aggregation (Bens et al., 2011; Botosan & Stanford, 2005). However, Berger and Hann (2007) find mixed results on the proprietary cost motive hypothesis, which predicts that managers withhold information about segments with relatively large abnormal profits, while Ettredge et al. (2006) show that managers continue having some discretion to conceal segment information that could be competitively harmful even post SFAS 131. Furthermore, following the regulation, there are now fewer firms reporting earnings by geographic area, which is no longer required for firms that define their operating segments on any basis other than geographic area (Herrmann & Thomas, 2000a). This non-disclosure of potentially important accounting information may have significant effects on future earnings predictability. Indeed, Hope et al. (2009) imply that this decrease in disclosure results in a decrease in the generation of event window private information in the spirit of Kim and Verrecchia (1997), which could decrease trading. In contrast, Hope et al. (2006) do not find an effect of this change on analyst forecast accuracy and dispersion, concluding that this disclosure change has not hindered investors’ ability to predict future earnings of multinationals.

Given the regulation’s potential weaknesses, FASB has been working since 2016 on possible improvements to the segment aggregation criteria and disclosures to provide interested users with more decision-useful information about reportable segments and its efforts are still ongoing, as of the writing of this article. Some recent notable decisions that are especially relevant in light of our study include the development of a requirement to, first, disclose significant segment expense categories that are regularly provided to the “chef operating decision maker” and included in reported segment profit or loss and, second, disclose such expense categories on an interim in addition to an annual basis. This suggests that the FASB may be concerned about whether segment information, as currently provided by companies, is sufficient, which calls for more information on the issue. Therefore, the effect of SFAS 131 on the efficiency with which stock prices capture information about complex firms is an unresolved and timely question, which we intend to address in this article.

Hypothesis Development

A primary role of financial accounting is to level the informational playing field and possibly alleviate frictions or constraints that preclude capital market participants from incorporating all available information into stock prices as efficiently and quickly as possible. Given investors’ limited resources and capacity to collect and interpret available information, overly complex or undetailed financial disclosure may result in a significant delay in the processing and impounding of information into stock prices. Indeed, CL find that industry information applicable to conglomerate firms takes longer to get captured by price due to the complicated analysis necessary to figure out the implications of this information for conglomerates. One way to deal with this problem is through accounting standards intended to improve financial disclosures and alleviate at least some constraints associated with the efficient processing of information. For example, the adoption of SFAS 131 was intended to significantly increase the amount and quality of information provided to market participants, which should result in improved processing of segment data. Consistent with this goal, prior research documents that following the implementation of SFAS 131, there is an increase in analyst forecast accuracy (Berger & Hann, 2003), price informativeness for future earnings (Ettredge et al., 2005; Park, 2011), and segment profitability disclosure transparency (Ettredge et al., 2006). This evidence suggests that the regulation likely improved the overall information environment of affected firms and an important mechanism that explains that this change relates to the speed with which stock prices capture information about conglomerates.

Studies that analyze investors’ inefficient or delayed reactions to information generally attribute this behavior to investors’ limited resources and capacity to collect, interpret, and then trade on publicly available, value-relevant information. There are two streams of literature that contribute to this broad area of research—investor under- or overreaction to various public announcements and investor inattention. We discuss the relevant findings below.

Prior research finds that investors tend to underreact to publicly available information resulting in a predictability of future returns from past returns (e.g., Chan et al., 1996; Zhang, 2006). One such phenomenon is PEAD, which refers to the continuance of a stock’s abnormal returns in the direction of an earnings surprise after an earnings announcement. The literature has proposed a number of hypotheses to explain this anomaly, but the most widely accepted explanation is investor underreaction to earnings announcements. Furthermore, Livnat and Mendenhall (2006) show that the drift is stronger when earnings surprises are based on analyst forecasts relative to time-series models, and the authors argue that this is because financial analysts, or investors, do not fully incorporate past information in their earnings forecasts. Extending this research, Kang et al. (2017) find that this expectations inefficiency of financial analysts is more pronounced for firms with international diversification, but the effect is less pronounced subsequent to the implementation of SFAS 131. Together, this literature suggests that there is an information processing inefficiency associated with earnings surprises, which increases in firms’ diversification, and disclosure regulation can alleviate this effect.

Another stream of research attributes the information processing inefficiencies to investors’ limited attention. This literature generally finds that investors respond more quickly to information that attracts their attention, such as that in media articles (e.g., Barber & Odean, 2008; Blankespoor et al., 2019), and tend to ignore information that is less salient yet value relevant, such as Securities and Exchange Commission (SEC) comment letters (Dechow et al., 2016) or information about related firms (Cohen & Frazzini, 2008). Similarly, DellaVigna and Pollet (2009) find a lower immediate response and higher drift for earnings disclosed on Fridays, relative to other days, and DeHaan et al. (2017) show that bad weather may also result in sluggish reaction to publicly disclosed information. Consequently, managers use this investor inattention strategically and disclose negative information after trading hours (e.g., DeHaan et al., 2015; Segal & Segal, 2016) or on Fridays (e.g., Louis & Sun, 2016; Michaely et al., 2016). Together, this research suggests that investors may react slower and less fully to information that they pay less attention to.

Overall, the literature argues that investors’ inefficient processing of available information leads to significant return predictability based only on publicly available information. In this vein, CL find that stock returns of conglomerate firms, those operating in multiple industries, are associated with prior returns of related focused firms. There are several ways in which SFAS 131 has improved disclosure and has likely made conglomerates’ publicly available information easier and faster to process. First, pre SFAS 131 (i.e., under SFAS 14), firms used the so-called industry approach to classify each segment and exercised discretion in defining industry. As a result, several firms disclosed fewer segments to investors relative to the number of segments that was used internally. Under SFAS 131, firms follow the so-called management approach that requires that segment information be presented consistent with how management assesses the operating performance of its segments internally. This increased disaggregation, which increased the amount of available information (e.g., Herrmann & Thomas, 2000b). Second, prior research finds an increase in analyst forecast accuracy (Berger & Hann, 2003), price informativeness for future earnings (Ettredge et al., 2005; Park, 2011), and segment profitability disclosure transparency (Ettredge et al., 2006), which is consistent with an increase in the quality of information after SFAS 131. Therefore, we posit that the increase in the amount and quality of conglomerates’ publicly available information has increased the speed of information flow and hence decreased these firms’ return predictability. Following these arguments, we state our first hypothesis in the alternative form:

Hypothesis 1 (H1): The predictability of conglomerates’ returns, based on publicly available information for standalones, is lower following the adoption of SFAS 131.

As CL find that the return predictability effect is stronger for more diversified firms, we seek to examine whether SFAS 131 has a differential effect on the predictability of more complex, relative to less complex, firms. The complexity of a firm’s operations adds another layer that affects the speed with which information is processed by market participants. The predictability effect documented in CL is predicated on the fact that news is processed faster for simple firms than it is for complicated firms. When regulation such as SFAS 131 modifies the way in which these firms report their financial information, it has the potential to differentially increase the speed of information flow for complex, as compared to less complex, conglomerates. For example, the increase in disaggregation following SFAS 131 provides a potential mechanism for the differential effect, where more complex firms increased their disaggregation more relative to less complex firms. Hence, we test the following hypothesis, stated in the alternative form:

Hypothesis 1a (H1a): The effect of SFAS 131 on the predictability of conglomerates’ returns, based on publicly available information for standalones, is stronger for more complex, relative to less complex, firms.

Prior research suggests that good and bad news disseminate differently in the capital markets. For example, Hou (2007) finds that bad news diffuses in an industry significantly more slowly than good news, likely due to the more pronounced effect of market frictions, such as short sale constraints, when bad news arrives. The disclosure literature also suggests that positive and negative news generate very different and asymmetrical responses in the capital markets (e.g., Kothari et al., 2009; Skinner & Sloan, 2002). Given these findings, it is interesting to examine whether there is a differential impounding of good and bad information around SFAS 131. Interestingly, Berger and Hann (2003) find some evidence that prior to the regulation, the market was less informed about new segment information that results in downward forecasts of earnings and revenues, that is, negative news, than information that leads to more positive projections. This suggests a potential asymmetry in the way good and bad news disseminate in the market. We state our second hypothesis in the alternative form as follows:

Hypothesis 2 (H2): The predictability of conglomerates’ returns, based on publicly available information for focused firms, is higher for bad, relative to good, news.

The asymmetric effect of the regulation for good and bad news is in line with models of gradual information diffusion, such as Hong and Stein (1999) or Hong et al. (2000). Specifically, these models show how underreaction in prices (i.e., prices adjusting too slowly to news) generates return predictability patterns (such as momentum, reversal, etc.). More pertinent to our article, models of gradual information flow specifically predict an asymmetric effect of return predictability for good and bad news: stocks with slow information flow seem to react more sluggishly to bad news than to good news. To proxy for the rate of information flow, papers in the current literature use constructs, such as size or analyst coverage. For example, Hong et al. (2000) argue that the marginal contribution of analysts in getting the news out is likely to be greater when the news is bad (when the news is good, the firm will find a way to speedily communicate good news to investors).⁵ We expect that, while SFAS 131 may be able to close the gap in processing speed when it comes to good news (and hence eliminate predictability when good news is involved), it will have a stronger effect in terms of being able to reduce predictability for bad news (where the gap in processing speed was larger to begin with). We argue that the difference in processing speed, and hence the predictability, is larger when the news is bad than when the news is good. If that is indeed the case, a standard like SFAS 131 that improves financial disclosure for complex firms may have an asymmetric effect on the processing of good/bad news. We express our final hypothesis as follows:

Hypothesis 2a (H2a): The effect of SFAS 131 on the predictability of conglomerates’ returns, based on publicly available information for focused firms, is stronger for bad news, relative to good news.

Data and Methodology

Sample

As discussed in the previous section, starting in 1976, firms are required by SFAS 14 to report information such as assets, sales, earnings, and cash flows from operations separately for each industry segment. After 1997, SFAS 131 supersedes SFAS 14 and significantly changes the way segments are defined, in effect changing the way firms disclose segment-level financial information. We use Compustat segment files to extract all segment data available for the period between 1977 and 2016. As our hypothesis is that this change in financial reporting should have a significant effect on the way investors interpret the reported information, we focus our analysis on comparing the sample pre (1977-1996) with the sample post SFAS 131 (2000-2016).⁶

We follow the methodology described in CL to categorize firms as conglomerates or standalones. Specifically, standalones (or focused firms) are defined as those that operate in only one industry and whose segment sales (as reported in the Compustat segment files) account for more than 80% of the total sales reported in the Compustat annual file. Conversely, conglomerates operate in more than one industry and have aggregate segment sales that account for more than 80% of the total sales of the firm. Using this approach, we use industries defined based on two-digit SIC codes to categorize each firm in the segment Compustat database.

We then merge this sample with the Center for Research in Security Prices (CRSP) monthly stock files and Compustat annual files. To ensure that the segment information is publicly known before we conduct our stock return tests, we impose at least a 6-month lag between firm fiscal-year ends and stock returns; specifically, we use segment financial information from a given fiscal year only after June of the following year. We further remove observations with missing market or book value of equity as well as firms that do not have a fiscal-year end in December.⁷ After applying these procedures, we are left with a sample of 234,665 firm-year observations, of which 122,744 (111,921) firm-year observations are associated with conglomerates (standalones) and 107,561 (105,311) firm-year observations are covering the period pre (post) SFAS 131.⁸ Table 1 presents summary statistics for our sample.

Table 1.

Summary Statistics.

Description	Minimum	Maximum	Mean	SD	Median
Panel A: Full sample (1977-2016)
Number of conglomerate firms in the sample year	1,683.0	4,792.0	2,798.0	792.9	2,619.0
Number of standalones in the sample per year	878.0	1,552.0	1,139.5	193.5	1,150.0
Full sample % coverage of CRSP universe (EW)	48.1	66.8	57.5	5.0	57.1
Full sample % coverage of CRSP universe (VW)	32.6	66.4	49.3	13.5	56.4
Conglm firms % of CRSP universe (EW)	10.4	24.1	17.1	3.8	17.8
Conglm firms % of CRSP universe (VW)	20.1	41.8	29.5	6.9	27.6
Standalones of % CRSP universe (EW)	31.8	54.7	40.4	5.7	38.4
Standalones of % CRSP universe (VW)	10.1	30.3	19.8	7.6	22.3
Number of industries per conglomerate	2.0	11.0	2.7	0.9	2.0
Percentage of sales per industry segment	0.0	1.0	0.2	0.3	0.1
Panel B: Pre sample (1977-1996)
Number of conglomerate firms in the sample year	1,683.0	4,792.0	2,657.3	934.7	2,541.0
Number of standalones in the sample per year	878.0	1,165.0	989.6	95.0	949.5
Full sample % coverage of CRSP universe (EW)	48.8	66.1	56.0	4.6	56.2
Full sample % coverage of CRSP universe (VW)	56.4	66.4	61.7	2.9	62.4
Conglm firms % of CRSP universe (EW)	10.4	24.1	16.0	4.6	14.0
Conglm firms % of CRSP universe (VW)	26.6	41.8	35.4	4.1	35.7
Standalones of % CRSP universe (EW)	33.2	54.7	40.0	6.6	37.3
Standalones of % CRSP universe (VW)	20.9	30.3	26.4	2.5	26.2
Number of industries per conglomerate	2.0	10.0	2.7	1.1	2.0
Percentage of sales per industry segment	0.0	1.0	0.4	0.3	0.3
Panel C: Post sample (2000-2016)
Number of conglomerate firms in the sample year	2,234.0	3,444.0	2,736.4	346.2	2,603.0
Number of standalones in the sample per year	1,150.0	1,552.0	1,301.0	114.4	1,293.0
Full sample % coverage of CRSP universe (EW)	48.1	66.8	58.5	5.5	56.8
Full sample % coverage of CRSP universe (VW)	32.6	37.9	34.9	1.5	34.9
Conglm firms % of CRSP universe (EW)	16.4	21.1	18.9	1.4	19.1
Conglm firms % of CRSP universe (VW)	20.1	26.8	23.0	1.9	22.7
Standalones of % CRSP universe (EW)	31.8	45.6	39.6	4.2	38.7
Standalones of % CRSP universe (VW)	10.1	13.1	11.9	0.8	12.0
Number of industries per conglomerate	2.0	11.0	2.7	0.9	3.0
Percentage of sales per industry segment	0.0	1.0	0.1	0.3	0.0

Note. This table shows the time-series averages of summary statistics as of December of each year, for the entire sample period 1977 to 2016 (Panel A, based on 234,665 firm-years observations), the pre SFAS 131 period (Panel B, based on 107,561 firm-years observations), and the post SFAS 131 period (Panel C, based on 105,311 firm-year observations). Full sample % coverage of CRSP universe (EW) is the number of stocks included either in the conglomerate or standalone sample for a given year divided by the number of stocks in the CRSP universe. Full sample % coverage of CRSP universe (VW) is the total market capitalization of stocks included in the conglomerate or standalone for the given year divided by the total market value of stocks in the CRSP universe. Observations with negative sales and negative book value of equity were eliminated as well as firms that do not have a December fiscal year or firms that had price lower than US$5. EW = equally weighted; VW = value weighted; CRSP = Center for Research in Security Prices; SFAS = Statement of Financial Reporting Standards.

In Panels A, B, and C of Table 1, we present the coverage of our sample as a fraction of the CRSP universe for the full, pre, and post sample, respectively. Combined, our full sample covers approximately 57.5% (49%) of the CRSP common stock universe in terms of the number of firms (market capitalization). Panel A shows that on average, conglomerates have 2.74 segments, and an average segment accounts for about 26% of the total sales reported by a conglomerate. Panels B and C show that the proportion of conglomerates and standalones out of the CRSP universe is relatively stable across the pre and post sample period. Specifically, the percentage of firms that represent conglomerates (standalones) in the pre period of 16% (40%) is comparable to the corresponding percentage in the post period of 18.9% (39.6%). Similarly, the number of firms covered in our sample relative to the total number of firms in CRSP is relatively stable in the pre and post periods (56% vs. 58%, on average). It is interesting to note, however, that in terms of market capitalization, there is a stark decline from the pre to the post sample coverage relative to the CRSP stock universe (in the pre period, our full sample covers about 61% of the CRSP universe, while in the post period, our sample covers only about 35% of the CRSP universe).

Research Design

The premise of CL is that stock prices capture complex information with a delay. Therefore, information will first be reflected in the easy-to-analyze standalones to the difficult-to-analyze conglomerates. This delay generates predictability in the returns of conglomerates. Following CL, we construct a “pseudo-conglomerate” portfolio for each conglomerate in the sample—specifically, a “pseudo-conglomerate” is a portfolio of the conglomerate firm’s industry segments constructed using solely standalones in each industry. The segment portfolios are then value weighted by the percentage of sales contributed by each industry segment within the conglomerate.⁹ The returns of these pseudo-conglomerate portfolios are then used to predict the returns of conglomerates and, more importantly, to investigate whether this predictability is influenced by the introduction of SFAS 131 and whether it varies by the degree of complexity, disaggregation, and type of news.

Complicated information processing before and after SFAS 131

To capture the effect of “complicated processing,” we start by sorting firms into portfolios based on how high (low) the returns on their pseudo-conglomerates are during the previous month. At the beginning of each month (starting in July), all conglomerates are sorted into deciles based on the returns of their corresponding pseudo-conglomerate portfolios in the previous month.¹⁰ The decile portfolios are then rebalanced at the beginning of each month.

The idea is that if the price movements of standalones (and hence the pseudo-conglomerates) predict future price movements of conglomerates, one should observe that firms with high (low) lagged pseudo-conglomerate returns produce significantly higher (lower) raw and risk-adjusted returns relative to their counterparts. Therefore, after sorting firms into portfolios based on the lagged returns of their pseudo-conglomerates, we investigate whether the average monthly returns of the firms in the top decile are significantly higher than the average monthly returns of the firms in the bottom decile.¹¹ To make sure that the difference is not driven by the risk profile of each portfolio, we also investigate whether the risk-adjusted returns (i.e., Alphas) of these portfolios are significantly different. We control for risk using various risk factor mimicking portfolios such as the Fama and French (1993) three factors (MKT, SMB, and HML), the Carhart (1997) momentum factor (UMD), and the Pástor and Stambaugh (2003) liquidity factor (LIQ).

As our focus is on the effect of the regulation on the way that information is incorporated into prices, to test H1, we implement the above strategy in the pre and post samples and compare the results. Specifically, to compare raw or risk-adjusted returns before and after SFAS 131, we use the whole sample of returns and run the following regression:

EXRE T_{t} = α + β_{1} INTE R_{t} + β_{2} POS T_{t} + \sum β_{i} RiskFacto r_{i, t} + ε_{t},

(1)

where EXRET is the return in excess of the risk-free rate, INTER is a dummy variable that takes value 1 if the month in question is in the intermediary period (1997.01-2000.06) and 0 otherwise, and POST is a dummy variable that takes value 1 if the month in question is post regulation (2000.07 and beyond) and 0 otherwise. Our focus is on the $β_{2}$ coefficient, which would indicate if the raw (or risk-adjusted) returns are significantly different in the post, relative to the period pre SFAS 131.

In addition to the portfolio sorts tests described above, we also employ the Fama and MacBeth (1973) cross-sectional regressions to investigate the predictability generated by complicated portfolio processing and the effect of SFAS 131 on this predictability. Specifically, we follow CL and conduct cross-sectional regressions of the following form:

Y_{t} = θ PCRE T_{t - 1} + λ_{1} RE T_{t - 1} + λ_{2} IRE T_{t - 1} + λ_{3} SIZE + λ_{4} B / M + λ_{5} MOM + λ_{6} TURN + ε_{t},

(2)

where the dependent variable (Y_t) is either (a) raw conglomerate returns (RET) or (b) industry-adjusted returns computed as the difference between the conglomerate firm return and its paired pseudo-conglomerate return (RET_t– PCRET_t).¹² The independent variable of interest is the return of the conglomerate’s paired pseudo-conglomerate in month t– 1 (PCRET_t₋₁). A significant $θ$ coefficient indicates that delay in information processing generates predictability in returns. To investigate the magnitude of this predictability before and after SFAS 131, we run the above test in the pre and post samples separately and compare the results. To provide a meaningful comparison of the average $θ$ coefficient across the samples, we first standardize all variables within each regression sample to make sure that differences in their magnitudes across samples do not affect the magnitudes of the $θ$ coefficients. As controls, we incorporate the conglomerate’s past return (RET_t₋₁), the lagged VW primary industry returns, as used in Moskowitz and Grinblatt (1999) (IRET_t₋₁), as well as lagged size, book-to-market, price momentum, and turnover.¹³ Cross-sectional regressions are run every calendar month, and the time-series standard errors are adjusted for heteroskedasticity and autocorrelation up to 12 lags, following Newey and West (1987).

Level of complexity and information processing

As CL argue, if the return predictability effect is truly driven by investors’ limited capacity and resources, combined with the valuation difficulty of conglomerate firms, we would expect that the more complicated the firm, the more severe the lag in incorporating information into prices and thus the stronger the return predictability. To test this prediction, we follow CL and use a Herfindahl index based on the firm’s segment sales as a measure of conglomerate complexity. Specifically, the Herfindahl Index is calculated as the sum of the squared ratios of segment sales over total sales. For example, a firm with segment sales of 1, 3, and 6 would have a Herfindahl index of (1/10)²+ (3/10)²+ (6/10)² = 0.46, and a firm with sales of 1, 1, and 8 would have a Herfindahl index of (1/10)²+ (1/10)²+ (8/10)² = 0.66. Hence, the higher the degree of dispersion across the segments, the higher the complexity and the lower the Herfindahl index. To distinguish between relatively complex and relatively simple conglomerates, we assign observations with a Herfindahl index above (below) its annually estimated median to a simple (complex) group. We investigate whether predictability is influenced by complexity by running a Fama and MacBeth (1973) regression as shown in Equation 2 for the full (1977-2016), the pre (1977-1996), and the post (2000-2016) sample period.

To test our H1a hypothesis, we examine the θ coefficient across the different subsamples.¹⁴ As we expect the predictive ability of PCRET_t₋₁ to be weaker for simpler firms, we expect $θ_{Simple}^{Full}$ to be smaller than $θ_{Complex}^{Full}$ . To test whether the perceived complexity is influenced by SFAS 131, we use a difference-in-difference research design and test whether the change in predictability for complex conglomerates between the pre and post SFAS 131 period is significantly different from the corresponding change for simple conglomerates. More specifically, we test whether $(θ_{Simple}^{Pre} - θ_{Simple}^{Post})$ = $(θ_{Complex}^{Pre} - θ_{Complex}^{Post})$ .¹⁵

Differences in processing good and bad news

As discussed in “Background and Hypothesis Development” section, good and bad news disseminate at different speeds in capital markets (Hou, 2007; Skinner & Sloan, 2002). Therefore, we need to consider whether return predictability varies with the direction of industry news and whether SFAS 131 has a differential effect on that phenomenon. To test this potential asymmetry, we consider that a conglomerate received market-adjusted good (bad) news if the pseudo-conglomerate return at time t– 1 (PCRET_t₋₁) is higher (lower) than the market return in t– 1.¹⁶ Using this definition, we separate our sample into good and bad news subsamples.¹⁷

First, to establish whether good and bad news are indeed processed differently, we rerun the portfolio tests described in “Complicated information processing before and after SFAS 131” section separately within each subsample (over the whole period) and investigate whether raw and risk-adjusted returns of the High–Low (H-L) portfolios are significantly different across the good and the bad news subset of observations (which would indicate that predictability is mitigated by the type of news considered). Our expectation is that predictability is more pronounced in the bad news subsample. Next, to investigate whether the change in segment reporting has any influence on this asymmetry, we further separate the good/bad subsamples into good/bad subsamples in the pre and post SFAS 131 period. The idea is to document whether the change in predictability driven by SFAS 131 is more pronounced within the universe of good news or the universe of bad news (H2). We therefore compare the pre and post good news raw and risk-adjusted returns of the H-L portfolios and the pre and post bad news raw and risk-adjusted returns of the H-L portfolios. Moreover, to test H2a, we compare the θ coefficient for PCRET_t₋₁ from Equation 2 across the different groups (pre/post, good/bad).

Empirical Results

Complicated Information Processing Before and After SFAS 131

As a first step in our analysis, we attempt to replicate the finding of CL for our sample period from 1977 to 2016. When sorting firms are based on the lagged returns of their pseudo-conglomerate portfolios, we should observe that the portfolio with high lagged pseudo-returns records significantly higher raw and abnormal returns relative to the portfolio with low lagged pseudo-returns. Untabulated results show that over the entire sample period (1977-2016), this is indeed the case when one looks at equal-weighted portfolios. However, results are weaker than those presented by CL and are insignificant when considering weighted portfolios.¹⁸ One reason for this mixed evidence may be that SFAS 131 significantly changed the speed with which industry information gets impounded in conglomerates’ stock prices. Table 2 compares the predictability of conglomerate returns between the sample period prior and after the adoption of SFAS 131.¹⁹ Column 1 in Panel A shows a significantly positive equal-weighted hedge return (H-L) of 116 basis points per month, or 13.92% per year, for the pre period. In contrast, Column 1 in Panel B shows no significant hedge return for the post period. As shown below Panel B, the difference between the hedge returns from the pre and the post period (the coefficient on a dummy variable identifying the post period is −1.04 with a t-statistic of −3.01) is highly statistically significant. Columns 2 to 5 show that the results reported in Panels A and B are robust to controlling for several known risk factors. Panels C and D show that the inferences based on equal-weighted returns also hold for VW returns.

Table 2.

Complicated Information Processing Portfolios, Abnormal Returns: Pre Versus Post SFAS 131 Implementation.

Portfolio		(1)Excess Returns	(2)1-Factor Alpha	(3)3-Factor Alpha	(4)4-Factor Alpha	(5)5-Factor Alpha
Panel A: EW, 1977-1996
High	param t stat	1.35 (4.03)	0.64 (3.99)	0.50 (3.17)	0.37 (2.57)	0.38 (2.60)
Low	param t stat	0.20 (0.53)	−0.57 (–3.08)	−0.71 (–4.72)	−0.63 (–4.12)	−0.63 (–3.96)
H-L (Pre)	param t-stat	1.16*** (5.68)	1.21*** (5.81)	1.21*** (5.54)	1.00*** (4.93)	1.01*** (4.83)
Panel B: EW, 2000–2016
High	param t stat	0.72 (1.67)	0.32 (1.34)	0.00 (0.02)	0.02 (0.07)	−0.03 (–0.12)
Low	param t stat	0.69 (1.39)	0.19 (0.83)	−0.08 (–0.44)	−0.03 (–0.15)	−0.08 (–0.43)
H-L (Post)	param t-stat	0.03 (0.12)	0.13 (0.48)	0.08 (0.31)	0.04 (0.16)	0.05 (0.20)
H-L (Post vs. Pre)	param t stat	−1.04*** (–3.01)	−1.08*** (–3.21)	−1.08*** (–3.21)	−0.93*** (–2.79)	−1.00*** (–3.00)
Panel C: VW, 1977-1996
High	param t stat	1.26 (3.75)	0.54 (2.74)	0.57 (2.81)	0.40 (1.92)	0.39 (1.87)
Low	param t stat	0.26 (0.73)	−0.49 (–2.54)	−0.48 (–2.50)	−0.29 (–1.50)	−0.29 (–1.49)
H-L (Pre)	param t-stat	1.00*** (3.41)	1.03*** (3.53)	1.05*** (3.48)	0.69** (2.27)	0.69** (2.22)
Panel D: VW, 2000–2016
High	param t stat	0.36 (0.89)	−0.01 (–0.04)	−0.09 (–0.36)	−0.10 (–0.38)	−0.20 (–0.71)
Low	param t stat	0.70 (1.49)	0.21 (0.81)	0.15 (0.61)	0.18 (0.74)	0.09 (0.38)
H-L (Post)	param t-stat	−0.33 (–0.89)	−0.22 (–0.59)	−0.24 (–0.63)	−0.28 (–0.74)	−0.29 (–0.76)
H-L (Post vs. Pre)	param t stat	−1.30*** (–2.74)	−1.34*** (–2.86)	−1.34*** (–2.87)	−1.17** (–2.50)	−1.17** (–2.49)

Note. This table shows calendar-time portfolio abnormal returns for the pre SFAS 131 period 1977 to 1996 (covering 237 months) and the post SFAS 131 period 2000 to 2016 (covering 186 months). Returns and Alphas are in monthly percentage, Newey–West t-statistics with 3 lags are shown below the coefficient estimates in parentheses, and *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively. Panels A and B (C and D) are based on equal-weighted (value-weighted) returns. For value-weighted results, we winsorize the size variable at 1% and 99% to mitigate the effect of outliers. High (low) shows the average returns for the firms in the top (bottom) decile ranked by pseudo-conglomerate returns in the previous month. H-L is the alpha of a zero-cost portfolio which holds (sell short) the firms in the high (low) portfolio. For both EW and VW results, we compute the difference in (risk-adjusted) returns between the post and pre periods and report it under Panels B and D, respectively. To estimate the difference H-L (post vs. pre), we run the same models over the full sample, but we incorporate a dummy variable that identifies the observations in the post period (we separate the intermediate years using another dummy variable). Excess Returns reports the portfolio return above the risk-free rate (1-month T-Bill). The Alphas refer to the intercepts for regressions of monthly excess returns on various risk factors. The explanatory variables used in the alpha-estimations are the monthly returns from the Fama and French (1993) factor mimicking portfolios, the Carhart (1997) momentum factor, and the Pástor and Stambaugh (2003) liquidity factor. SFAS = Statement of Financial Reporting Standards; EW = equally weighted; VW = value weighted.

Together, the results presented in Table 2 indicate that while focused firms predict conglomerate returns prior to the adoption of SFAS 131, this predictability has disappeared after the adoption of SFAS 131. This result supports Hypothesis H1 and is consistent with the idea that SFAS 131 has improved segment reporting, such that investors can process segment information more efficiently, thereby largely eliminating the delay with which the same piece of information is captured by stock prices for standalones and conglomerates.²⁰

Next, we move our attention from portfolio tests to cross-sectional Fama–MacBeth regressions. Table 3 is based on a cross-sectional regression framework as described in Equation 2.²¹ This approach complements our prior analysis in the sense that it allows us to control for other determinants of conglomerate returns and thereby helps isolate the marginal predictive ability of lagged pseudo-conglomerate returns (PCRET_t₋₁), the main variable of interest.

Table 3.

Complicated Information Processing Returns, Cross-Sectional Regressions.

$Y_{t} = θ PCRE T_{t - 1} + λ_{1} RE T_{t - 1} + λ_{2} IRE T_{t - 1} + λ_{3} SIZE + λ_{4} B / M + λ_{5} MOM + λ_{6} TURN + ε_{t}$
	RET_t		RET_t– PCRET_t
Y× 100 =	(1)	(2)	(3)	(4)
Panel A: 1977-2016
PCRET_t₋₁	0.195*** (4.73)	0.177*** (4.723)	0.081** (2.184)	0.075* (1.686)
RET_t₋₁		−0.428***(–8.711)		−0.457***(–9.903)
IRET_t₋₁		0.201***(5.739)		0.132***(4.03)
Controls	Yes	Yes	Yes	Yes
Adjusted R²	2.2%	3.2%	2.0%	3.0%
Panel B: 1977-1996
PCRET_t₋₁	0.318***(5.948)	0.291***(5.977)	0.174***(5.192)	0.192***(4.76)
RET_t₋₁		−0.534***(–8.863)		−0.571***(–10.14)
IRET_t₋₁		0.225***(4.473)		0.094**(2.241)
Controls	Yes	Yes	Yes	Yes
Adjusted R²	2.1%	3.0%	1.7%	2.8%
Panel C: 2000-2016
PCRET_t₋₁	0.046(0.879)	0.048(1.013)	−0.013(–0.21)	−0.041(–0.531)
RET_t₋₁		−0.294***(–3.894)		−0.302***(–4.44)
IRET_t₋₁		0.138***(2.939)		0.149***(2.84)
Controls	Yes	Yes	Yes	Yes
Adjusted R²	3.0%	4.2%	3.0%	4.2%

Note. This table reports Fama–MacBeth cross-sectional regressions of stock returns. The dependent variable in Columns 1 and 2 is the monthly conglomerate excess return (RET_t), while in Columns 3 and 4, the dependent variable is the excess conglomerate return minus its paired pseudo-conglomerate (RET_t– PCRET_t). A pseudo-conglomerate is a portfolio of the conglomerates industry segments constructed using solely the standalones from any given industry. The explanatory variables are the lagged pseudo-conglomerate return (PCRET), the firm’s own lagged return (RET), and the lagged return of the corresponding industry portfolio to the conglomerate’s main industry (IRET). All regressions also include SIZE, book-to-market (B/M), momentum (MOM), and turnover (TURN), all of which are measured at the end of June of each year. All variables are standardized before running the Fama–MacBeth regressions. Cross-sectional regressions are run every calendar month, and the time-series standard errors are adjusted for heteroskedasticity and autocorrelation (up to 12 lags; Newey & West, 1987). Fama–MacBeth t-statistics are reported below the coefficient estimates in parentheses, and *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Consistent with the results documented in Table 2, the statistically significantly positive θ coefficient on PCRET_t₋₁ in Column 1 of Panel A in Table 3 shows that for the entire sample period, the pseudo-conglomerate return in the previous months is a strong predictor of next month’s returns for the corresponding conglomerate. As shown in Column 2, this finding is robust to including the lagged conglomerate return (RET_t₋₁) and the lagged industry return (IRET_t₋₁). This suggests that the predictive ability of PCRET_t₋₁ is not driven by the short-term reversal effect documented by Jegadeesh (1990) or by the conglomerate’s primary industry. Our subsample analysis shows that the results in the overall sample (Panel A) are mainly driven by the period pre SFAS 131 (Panel B). Panels B and C reveal that the predictive ability of PCRET_t₋₁ has declined substantially after the implementation of SFAS 131. In Column 1, the coefficient estimate on PCRET_t₋₁ decreases by 85% from 0.318 (t = 5.948) in the pre period to 0.046 (t = 0.879) in the post period, and the difference is statistically significant at the 1% level (untabulated). Controlling for lagged RET and IRET (Column 2) does not meaningfully affect this decrease.

One potential explanation for the results shown in Columns 1 and 2 is the previously documented industry-momentum effect (e.g., Moskowitz & Grinblatt, 1999). To the extent that RET_t is simply capturing the momentum of industry returns, the positive PCRET_t₋₁ coefficient would not indicate return predictability but simply the autocorrelation of lagged industry returns with current industry returns. We address this concern by subtracting the contemporaneous pseudo-conglomerate return (PCRET_t) from the conglomerate return (RET_t). This approach helps us separate the predictive ability of PCRET_t₋₁ that is driven by delayed information processing due to information complexity from the industry-momentum effect. The results in Columns 3 and 4 of Panel A indicate that even after removing industry momentum, the lagged PCRET_t₋₁ remains a significant predictor of RET_t (albeit weaker, as expected), supporting the hypothesis that the delayed information processing for conglomerates at least partly drives the predictive ability of lagged pseudo-conglomerate returns.

Consistent with results in Columns 1 and 2, PCRET_t₋₁ displays predictive ability in the pre period but not in the post period. In Column 3, the PCRET_t₋₁ coefficient estimate drops from 0.174 (t = 5.192) in the pre period to a coefficient that is statistically indistinguishable from zero (–0.013, t = −0.21) in the post period. In Column 4, the same result holds after we control for lagged RET and IRET. This finding suggests that any remaining predictive ability of lagged PCRET after the implementation of SFAS 131 is not due to information complexity of conglomerates but rather due to industry momentum. Put differently, this result is consistent with the view that SFAS 131 has substantially increased the speed with which information about conglomerates is captured by prices. This provides further support for H1.

Cross-Sectional Effects of SFAS 131 on Information Processing

Complexity

In Table 4, we investigate whether the results documented in Table 3 are influenced by a conglomerate’s complexity.²² The logic is that the more complex a conglomerate, the more difficult it is for investors to interpret and incorporate the available information into the conglomerate’s stock price and thus the stronger the return predictability. As described in “Level of complexity and information processing” section, we estimate the θ coefficient for PCRET_t₋₁ from Equation 2 separately for each group to obtain $θ_{Simple}^{Full}$ , $θ_{Complex}^{Full}$ , $θ_{Simple}^{Pre}$ , $θ_{Complex}^{Pre}$ , $θ_{Simple}^{Post}$ , and $θ_{Complex}^{Post}$ .

Table 4.

Does the Level of Complexity Matter?

$RE T_{t} = θ PCRE T_{t - 1} + λ_{1} RE T_{t - 1} + λ_{2} IRE T_{t - 1} + λ_{3} SIZE + λ_{4} B / M + λ_{5} MOM + λ_{6} TURN + ε_{t}$
	(1)	(2)	(3)	(4)	(5)	(6)
	1977-2016	1977-2016	1977-1996	1977-1996	2000-2016	2000-2016
Independent Variable	$θ_{Simple}^{Full}$	$θ_{Complex}^{Full}$	$θ_{Simple}^{Pre}$	$θ_{Complex}^{Pre}$	$θ_{Simple}^{Post}$	$θ_{Complex}^{Post}$
PCRET_t ₋₁	0.208***(4.407)	0.143***(3.18)	0.289***(4.382)	0.292***(5.077)	0.117*(1.775)	−0.034(–0.595)
RET_t ₋₁	−0.380***(–7.132)	−0.483***(–7.774)	−0.447***(–6.764)	−0.630***(–7.707)	−0.237***(–3.041)	−0.363***(–3.867)
IRET_t ₋₁	0.181***(3.836)	0.211***(5.548)	0.226***(3.246)	0.217***(4.011)	0.099(1.614)	0.168***(2.894)
SIZE	−0.004(–0.133)	−0.015(–0.41)	0.057(1.517)	0.037(0.734)	−0.097**(–2.011)	−0.123**(–2.571)
B/M	0.117***(2.667)	0.051(1.318)	0.173***(2.877)	0.038(0.73)	0.087(1.292)	0.037(0.607)
MOM	0.180***(3.103)	0.121***(2.626)	0.221***(3.449)	0.107*(1.876)	0.071(0.686)	0.045(0.687)
TURN	−0.173**(–2.466)	−0.164***(–3.048)	−0.226***(–2.819)	−0.178***(–2.98)	−0.210(–1.617)	−0.164*(–1.742)
Adjusted R²	3.5%	3.3%	3.3%	3.1%	3.4%	3.4%
Number of months	474	474	234	234	198	198
Number of observations	134,599	146,924	63,775	69,652	59,877	64,987
Differences in the $θ$ coefficients across columns
	(1) – (2)	(3) – (4)	(5) – (6)	(3) – (5)	(4) – (6)	(3) – (5) vs. (4) – (6)
	−0.065*(1.399)	0.004(0.056)	−0.151**2.145	0.171**1.840	0.326***4.360	−0.154*1.950

Note. This table reports Fama–MacBeth cross-sectional regressions of stock returns. The dependent variable (RET_t) is the monthly conglomerate return. A pseudo-conglomerate is a portfolio of the conglomerate firm’s industry segments constructed using solely the standalone firms from any given industry. The explanatory variables are the lagged pseudo-conglomerate return (PCRET), the firm’s own lagged return (RET), and the lagged return of the corresponding industry portfolio to the conglomerate’s main industry (IRET). All regressions also include SIZE, book-to-market (B/M), momentum (MOM), and turnover (TURN), all of which are measured at the end of June of each year. All variables are standardized before running the Fama–MacBeth regressions in each sample. We use the Herfindahl index calculated as the sum of the squared ratios of segment sales over total sales to assign firms to the simple or complex group. A higher level of diversification is captured by a lower Herfindahl index. Observations with a Herfindahl index above (below) its annually estimated median are assigned to the simple (complex) group. Cross-sectional regressions are run every calendar month, and the time-series standard errors are adjusted for heteroskedasticity and autocorrelation (up to 12 lags; Newey & West, 1987). Fama–MacBeth t-statistics are reported below the coefficient estimates in parentheses, and *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

As expected, the results presented in Column 1 of Table 4 show that $θ_{Simple}^{Full}$ and $θ_{Complex}^{Full}$ are both significantly positive. From the results presented in Table 3, we know that SFAS 131 has reduced the predictive ability of PCRET_t₋₁. The set up in Table 4 allows us to learn more about the mechanism, via which SFAS 131 has reduced the predictive ability of PCRET_t₋₁. To the extent that SFAS 131 has mitigated the pricing lag between conglomerates and standalones by reducing the complexity of segment information, one would expect the impact of SFAS 131 to be more pronounced for more complex conglomerates, as predicted in H1a. In other words, the more complex the conglomerate’s operations, the higher the potential for SFAS 131 to help investors process information more quickly.

Columns 3 and 4 display a statistically significant $θ_{Simple}^{Pre}$ and $θ_{Complex}^{Pre}$ , thereby showing predictability for both simple and complex conglomerates in the pre period. For the post period, Column 5 shows that predictability remains for the simple conglomerates (the $θ_{Simple}^{Post}$ coefficient is weakly significant), while in contrast, predictability is no longer present for the complex conglomerates (Column 6 shows a $θ_{Complex}^{Post}$ that is insignificantly different from zero). The evidence so far indicates that predictability after SFAS 131 declines more for complex conglomerates. As our interest is in the differential effect of SFAS 131 for simple versus complex conglomerates, we present the differences in $θ$ coefficients between the pre and post periods for each respective group in the bottom of Table 4. We observe a significant decline in predictability between the pre and post period for both simple and complex conglomerates. Specifically, we find a significant decline from $θ_{Simple}^{Pre}$ to $θ_{Simple}^{Post}$ (0.171**) and from $θ_{Complex}^{Pre}$ to $θ_{Complex}^{Post}$ (0.326***).

A comparison of conglomerate return predictability across our pre (1977-1996) and post (2000-2016) sample period is potentially liable to confounding factors other than the introduction of SFAS 131. As discussed above, the impact of SFAS 131 is likely to be relatively more (less) pronounced for complex (simple) conglomerates. Using the simple conglomerates as a proxy for a control group and the complex conglomerates as a treatment group allows us to compare the changes in return predictability between the pre and post sample periods across simple and complex conglomerates.²³ This difference-in-difference design provides a cleaner look at the impact of SFAS 131 on the predictability of conglomerate returns. Consistent with the idea that the implementation of SFAS 131, rather than other confounding factors, has significantly decreased the predictability of conglomerate returns, we find that the difference across the changes, that is, $(θ_{Simple}^{Pre} - θ_{Simple}^{Post}) - (θ_{Complex}^{Pre} - θ_{Complex}^{Post})$ , is statistically significant (–0.154, t = 1.950).

Together, these results support our H1a and show that while it was possible to predict conglomerate returns in the pre period, after the implementation of SFAS 131, it is not possible to predict conglomerate returns in the post period for complex conglomerates. This result implies that SFAS 131 has increased pricing efficiency for conglomerates to such an extent that the pricing lag between simple and complex conglomerates cannot be gainfully exploited anymore. Moreover, we have documented that the impact of SFAS 131 was more pronounced for more complex conglomerates.

We also consider what is the specific mechanism driving the differential effect of SFAS 131 for simple/complex conglomerates. When regulation such as SFAS 131 modifies the way in which firms report their financial information, it has the potential to differentially increase the speed of information flow for complex, as compared to less complex, conglomerates. For example, the increase in disaggregation following SFAS 131 provides a potential mechanism for the differential effect, where more complex conglomerates increased their disaggregation more relative to less complex conglomerates.²⁴ We therefore investigate whether the impact of increased disaggregation on information processing is driving the firm complexity results presented above. We focus on changes in disclosure related to the number of business segments reported, as operational segments are not available in the pre window. Our approach is to sort firms based on how much they changed their reporting around SFAS 131 and investigate whether the magnitude of changes in disclosure/disaggregation around SFAS 131 affects the change in predictability of conglomerates’ returns. We repeat the Fama–MacBeth tests described in Table 4 within various subsamples and explore whether firms that increased disaggregation the most exhibited the most reduction in return predictability. The detailed results of this analysis are presented in the Online Appendix.²⁵ In short, when splitting the sample into high/low disaggregation-increasing firms based on business segments, we observe that the decrease in predictability is stronger for firms that increased disaggregation more after the passing of SFAS 131. Overall, these results support the idea that increasing disaggregation (from a business segment disclosure perspective) helps to make information easier to process, which in turn reduces the predictability of conglomerates’ returns.

Differences in processing good and bad news

As good and bad news disseminate in capital market differently (Hou, 2007; Skinner & Sloan, 2002), we consider whether return predictability varies with the sign of industry news and whether SFAS 131 has a differential effect on that phenomenon. Good (bad) news is defined as described in the “Differences in processing good and bad news” section. We start by confirming that good and bad news are indeed processed differently,²⁶ which leads to the natural question as to whether SFAS 131 changed the predictive ability of PCRET_t₋₁ for good and/or bad news differently and whether the substantial reduction in predictability in our post sample is not simply driven by having relatively less bad industry news for that particular time period. To investigate these questions, we next provide a test of H2 and H2a and examine our results relative to good and bad news separately in the pre and post SFAS 131 time periods. We provide these results in Table 5.²⁷

Table 5.

Predictive Ability for Good Versus Bad News (Pre vs. Post SFAS 131).

Portfolio	Excess Returns	1-Factor Alpha	3-Factor Alpha	4-Factor Alpha	5-Factor Alpha
Panel A: good news (1977-1996)
High	1.46(4.20)	0.59(3.11)	0.54(2.91)	0.43(2.42)	0.42(2.35)
Low	0.88(2.52)	0.02(0.12)	−0.01(–0.08)	−0.01(–0.06)	−0.02(–0.16)
H-L	0.58***(2.98)	0.57***(2.77)	0.55**(2.45)	0.44*(1.94)	0.44*(1.97)
Panel B: good news (2000-2016)
High	0.81(1.55)	0.18(0.68)	0.00(0.00)	0.00(0.02)	−0.06(–0.26)
Low	0.90(1.96)	0.29(1.47)	0.17(1.02)	0.20(1.14)	0.14(0.87)
H-L	−0.09(–0.35)	−0.11(–0.43)	−0.17(–0.66)	−0.19(–0.72)	−0.20(–0.75)
Panel C: bad news (1977-1996)
High	0.97(3.05)	0.14(0.90)	0.05(0.36)	0.07(0.47)	0.03(0.23)
Low	0.06(0.13)	−0.86(–3.37)	−0.92(–4.07)	−0.91(–3.87)	−0.93(–3.82)
H-L	0.92***(3.77)	1.00***(4.08)	0.97***(4.03)	0.98***(4.00)	0.96***(3.83)
Panel D: bad news (2000-2016)
High	1.22(2.69)	0.62(3.59)	0.51(3.04)	0.55(3.14)	0.52(3.01)
Low	0.73(1.32)	0.02(0.08)	−0.16(–0.62)	−0.06(–0.22)	−0.09(–0.35)
H-L	0.48*(1.69)	0.60**(2.13)	0.67**(2.45)	0.61**(2.19)	0.61**(2.19)

Note. This table shows the H-L alphas of a zero-cost equal-weighted portfolio which holds (sells short) the firms in the top (bottom) decile ranked by pseudo-conglomerate returns in the previous month based on calendar-time portfolio raw excess and risk-adjusted returns for the pre SFAS 131 period 1977-1996 (covering 237 months) and the post SFAS 131 period 2000-2016 (covering 186 months). Each time period is further separated into a good/bad news subsample (good/bad news is defined as to where the pseudo-conglomerate return at time t– 1, PCRET_t₋₁, is higher/lower than the market return in t– 1, respectively). Returns and Alphas are in monthly percentage, Newey–West t-statistics with 3 lags are shown below the coefficient estimates in parentheses, and *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively. Excess Returns reports the portfolio return above the risk-free rate (1-month T-Bill). The Alphas refer to the intercepts for regressions of monthly excess returns on various risk factors. The explanatory variables used in the alpha-estimations are the monthly returns of the market (1-Factor Alpha); the monthly returns for the Fama and French (1993) factor mimicking portfolios for market, size, and book to market (3-Factor Alpha); as well as the Fama and French (1993) factors and the additional Carhart (1997) momentum factor (4-Factor Alpha). SFAS = Statement of Financial Reporting Standards.

Overall results show that, while predictability exists for both good and bad news in the pre period (albeit weaker for good news than for bad), only bad news continue to exhibit predictability post SFAS 131. These results provide support for our H2 and suggest that the implementation of SFAS 131 has not been effective in fully mitigating the pricing lag for complex firms with bad news, thereby failing to fully remove the predictive ability of PCRET_t₋₁ for bad news. That said, an interesting pattern emerges from Panels C and D in Table 5—specifically, in Panel C, the significant results for the H-L portfolio are driven by the low portfolio (for the pre period); in contrast, Panel D shows that the significant returns from the H-L portfolio are driven by the high portfolio in the post period. One potential explanation is that bad news is processed differently by the market in the pre and post periods (and our argument is that the disclosure change had something to do with that). However, alternative explanations are also possible—for example, this pattern could be related to the switch in patterns of exposure to risk factors (in particular, momentum).²⁸

In Table 6, we investigate whether the predictability of conglomerate returns is affected by the type of news, that is, good versus bad. The logic is that as bad news travels slowly, it is more difficult for investors to interpret and incorporate bad news into the conglomerate’s stock price, and thus, one would expect a different impact of SFAS 131 for good and bad news, as predicted in H2a. As described in the “Level of complexity and information processing” section, we estimate the θ coefficient for PCRET_t₋₁ from Equation 2 separately for each group to obtain $θ_{Good}^{Full}$ , $θ_{Bad}^{Full}$ , $θ_{Good}^{Pre}$ , $θ_{Bad}^{Pre}$ , $θ_{Good}^{Post}$ , and $θ_{Bad}^{Post}$ .

Table 6.

Good/Bad News and the Impact of SFAS 131 on Information Processing.

$RE T_{t} = θ PCRE T_{t - 1} + λ_{1} RE T_{t - 1} + λ_{2} IRE T_{t - 1} + λ_{3} SIZE + λ_{4} B / M + λ_{5} MOM + λ_{6} TURN + ε_{t}$
	(1)	(2)	(3)	(4)	(5)	(6)
	1977-2016	1977-2016	1977-1996	1977-1996	2000-2016	2000-2016
Independent variable	$θ_{Good}^{Full}$	$θ_{Bad}^{Full}$	$θ_{Good}^{Pre}$	$θ_{Bad}^{Pre}$	$θ_{Good}^{Post}$	$θ_{Bad}^{Post}$
PCRET_t ₋₁	0.224**(2.274)	0.531***(3.954)	0.452**(2.435)	0.796***(3.83)	0.022(0.111)	0.497**(2.03)
RET_t ₋₁	−0.566***(–7.338)	−0.492***(–7.635)	−0.637***(–7.077)	−0.559***(–6.502)	−0.382***(–3.124)	−0.369***(–4.289)
IRET_t ₋₁	0.381***(2.705)	0.594***(4.35)	0.664***(3.837)	0.672***(3.486)	0.230(0.859)	0.697**(2.475)
SIZE	−0.620**(–1.984)	−0.211(–0.599)	−0.347*(–1.934)	−0.104(–0.438)	−0.199*(–1.961)	−0.135*(–1.774)
B/M	0.073(0.828)	0.096(1.449)	0.134(1.586)	0.132*(1.671)	0.123(0.752)	0.018(0.174)
MOM	0.179**(2.37)	0.162**(2.58)	0.031(0.262)	0.006(0.054)	0.062(0.493)	0.118(1.32)
TURN	−0.824***(–3.998)	−0.540***(–3.177)	−0.363***(–3.781)	−0.271***(–3.399)	−0.361**(–2.052)	−0.274*(–1.68)
Adjusted R²	15.5%	15.6%	15.0%	15.1%	14.1%	14.7%
Number of months	474	474	234	234	198	198
Number of observations	151,974	130,484	75,549	59,503	65,201	58,713
Differences in the $θ$ coefficients across columns
	(1) – (2)	(3) – (4)	(5) – (6)	(3) – (5)	(4) – (6)	(3) – (5) vs. (4) – (6)
	0.308**(2.102)	0.345**(1.699)	0.475**(1.707)	0.430**(1.688)	0.299(0.971)	0.130(0.932)

Note. This table reports Fama–MacBeth cross-sectional regressions of stock returns. The dependent variable (RET_t) is the monthly conglomerate return. A pseudo-conglomerate is simply a portfolio of the conglomerate firm’s industry segments constructed using solely the standalone firms from any given industry. The explanatory variables are the lagged pseudo-conglomerate return (PCRET), the firm’s own lagged return (RET), and the lagged return of the corresponding industry portfolio to the conglomerate’s main industry (IRET). All regressions also include SIZE, book-to-market (B/M), momentum (MOM), and turnover (TURN), all of which are measured at the end of June of each year. All variables are standardized before running the Fama–MacBeth regressions in each subsample. Each sample is split into firms with good and bad news, where good/bad news is defined as to where the pseudo-conglomerate return at time t– 1, PCRET_t₋₁, is higher/lower than the market return in t– 1, respectively. Cross-sectional regressions are run every calendar month, and the time-series standard errors are adjusted for heteroskedasticity and autocorrelation (up to 12 lags; Newey & West, 1987). Fama–MacBeth t-statistics are reported below the coefficient estimates in parentheses, and *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively. SFAS = Statement of Financial Reporting Standards.

As expected, the results presented in Columns 1 and 2 of Table 6 show that $θ_{Good}^{Full}$ and $θ_{Bad}^{Full}$ are both significantly positive. The finding that $θ_{Bad}^{Full}$ is significantly larger than $θ_{Good}^{Full}$ is consistent with the idea that bad news is incorporated less quickly into stock prices, thereby enhancing return predictability.

From the results discussed above, we know that SFAS 131 has reduced the predictive ability of PCRET_t₋₁. The set up in Table 6 allows us to learn more about the mechanism via which SFAS 131 has reduced the predictive ability of PCRET_t₋₁. To the extent that SFAS 131 has mitigated the pricing lag between conglomerates and standalones by enhancing the speed with which information is impounded into stock prices, one would expect the return predictability to be different post SFAS 131.

The results presented in Columns 3 and 4 (and 6) of Table 6 are based on the pre (post) sample period. Columns 3 and 4 show a strong and statistically significant predictability for both good and bad news in the pre period. As expected, $θ_{Bad}^{Pre}$ is significantly larger than $θ_{Good}^{Pre}$ . For the post period, Column 5 shows a statistically insignificant $θ_{Good}^{Post}$ but a statistically significant $θ_{Bad}^{Post}$ . The difference between these two coefficients is statistically significant, suggesting that while SFAS 131 has eliminated the predictability for good news, it has failed to eliminate return predictability for bad news. However, the difference-in-difference coefficient in Column 6 is statistically insignificant, and hence, there is no evidence that the regulation had a differential effect on good versus bad news.

Together, these results show that while it was possible to predict conglomerate returns for good as well as for bad news in the pre period, post SFAS 131, it is not possible to predict conglomerate returns for good news while predictability for bad news remains, as expected by H2. However, we do not find evidence to support H2a that the effect of SFAS 131 was stronger for bad news, relative to good news. Therefore, while SFAS131 increased the speed of information processing for good as well as bad news, as bad news were processed more slowly prior to the regulation, some predictability remains in the post period for bad news.

Discussion of the Results and Robustness Checks

Delayed Information Processing or Overreaction

So far, we argue that the predictive ability of pseudo-conglomerate returns with respect to conglomerate returns is the result of prices for conglomerates lagging behind the prices for standalones. If this is indeed the case, one would not observe a subsequent reversal of those returns. In contrast, if our findings are merely the result of a temporary overreaction, one would observe a subsequent reversal of the returns. To distinguish between these two alternative explanations, we investigate the cumulative return portfolios described in Tables 2 and 3 over 1- to 6-month horizons.

Figure 1a shows the cumulative return for the H-L hedge portfolio for the full (1977-2016), pre (1977-1996), and post (2000-2016) sample period. Consistent with the findings discussed above, the predictability of conglomerate returns is strongest (weakest) in the pre (post) sample period. Most importantly, the graphs in Figure 1 do not display any sign of return reversal. This finding is consistent with the explanation that prices for complicated firms lag behind prices for simple firms. Moreover, the result presented in Figure 1 also shows that the introduction of SFAS 131 may have eliminated this pricing lag, thereby making the pricing of conglomerates more efficient.²⁹ Figure 1b (1c) shows the cumulative return for the high (low) leg of hedge portfolio for the full, pre, and post sample period. Neither leg displays any signs of return reversals. These results also hold for VW returns (Figures 1d to 1f).

Figure 1.

Cumulative hedge portfolio returns for up to 6 months out.

Other Robustness Checks

One potential concern is that the results are driven by our method of defining good and bad news, and we define bad news when PCRET_t₋₁ is lower than the overall market return. As our focus is on cross-sectional variation in news across firms, we do not account for whether PCRET_t₋₁ (i.e., “the news”) is actually positive (good) or negative (bad)—we are only trying to capture whether it is better or worse, relative to other firms in the market. To confirm that our results are robust to various definitions of good/bad news, we consider two alternative proxies: (a) we use IRET_t₋₁ being higher/lower than the market to define good/bad news and (b) we use PCRET_t₋₁ being positive/negative to define good/bad news. Untabulated tests show that using these proxies leads to substantially the same conclusions as we obtain based on our main measure of good/bad news.³⁰

In terms of our empirical design, when comparing predictability across groups, we opt to present results where we run our Fama–MacBeth regressions within each group of interest and then compare standardized coefficients (see, for example, Tables 3 and 6). One alternative approach is to use dummy variables and interactions to capture the differences between two groups in the same cross-section. We obtain similar inferences when taking this approach, based on dummy variables.

Although we investigate the delay with which information is incorporated into prices for conglomerates, we do not claim that information frictions (i.e., speed of information processing) is the only mechanism that can generate the observed return predictability. We do not attempt to eliminate other potential channels that could generate serial correlation in returns (e.g., we do not attempt to disentangle the role of information frictions from liquidity frictions as discussed in Biais & Bossaerts, 1998, and Makarov & Rytchkov, 2012). Our key point remains that SFAS131 has helped mitigate the information frictions, thereby increasing the speed with which conglomerates’ prices capture information.

Conclusion

CL document that stock prices for complicated firms capture industry news with a lag, relative to focused firms. Specifically, they exploit information events that affect an entire industry, and they document that this information is incorporated faster for firms operating solely in that industry, relative to conglomerates operating in multiple industries. We identify a change in financial reporting policy (SFAS 131), which affects the way in which conglomerate firms report segment information to the investors. We then investigate whether this change in financial reporting alleviates the complicated processing constraints documented by CL and whether the effect varies cross-sectionally based on firm complexity, level of disaggregation, and the direction of news.

Our results are three-fold. First, we document that SFAS 131 has significantly altered the processing of industry-level information and has increased the speed with which industry information is impounded into prices. Specifically, we document that the return predictability from the set of easy-to-analyze firms to their more complicated peers, documented by CL, has significantly decreased and is almost nonexistent in the period post SFAS 131. Second, consistent with the complex information processing explanation, we show that the effect of SFAS 131 on the predictability of conglomerate firms’ returns is mitigated by the complexity of the conglomerates (i.e., the effect of SFAS 131 is stronger for more complex firms and weaker for simpler firms). Finally, we document that the return predictability derived from complex information processing is mostly driven by investors’ difficulty to process bad news. On this note, we show that SFAS 131 has significantly alleviated the lag with which good news is incorporated in the prices of complicated firms. However, the regulation has had a much weaker effect on the processing of bad news, which continues to drive significant predictability in the returns of conglomerate firms. Therefore, FASB may wish to consider this result when contemplating changes to the current segment reporting standard.

Supplemental Material

sj-docx-1-jaf-10.1177_0148558X221086248 – Supplemental material for Cutting Through Complexity: Segment Disclosure and Pricing Efficiency

Supplemental material, sj-docx-1-jaf-10.1177_0148558X221086248 for Cutting Through Complexity: Segment Disclosure and Pricing Efficiency by Doina C. Chichernea, Philipp D. Schaberl and Maya A. Thevenot in Journal of Accounting, Auditing & Finance

Footnotes

Appendix

Variable Definitions

RET	Monthly stock returns from CRSP.
IRET	The value-weighted primary industry return of the conglomerate, as used in Moskowitz and Grinblatt (1999).
PCRET	The pseudo-conglomerate return. A “pseudo-conglomerate” is a portfolio of the conglomerate firm’s industry segments (two-digit SIC) constructed using solely focused firms in each industry. The segment portfolios are then weighted by the percentage of sales contributed by each industry segment within the conglomerate.
SIZE	Market value of equity taken from CRSP as of the end of June. ABS (PRC) ×SHROUT× 1000 from CRSP
B/M	Book value of equity (I) divided by market value of equity (MVE). BVE is calculated based on Compustat. We prefer CEQ, but if it is unavailable, we use CEQL. Also, if short- and/or long-term deferred taxes are available (TXP, TXDITC), we add them to book equity. If both CEQ and CEQL are missing, we proxy book equity by the last period’s book equity plus earnings (IB) less dividends (DVT). We treat negative or zero book equity values as missing. MVE is taken from CRSP as of the end of June.
MOM	Momentum, calculated as the cumulative returns over the last 6 months.
TURN	TURN is the average monthly turnover calculated over the past 12 months. Monthly turnover is calculated as total shares traded (VOL) divided by total shares outstanding (SHROUT). TURN is calculated as of the end of June of each year.
Excess Returns	Excess return is the portfolio return above the risk-free rate.
1-Factor Alpha	The explanatory variable used in the alpha-estimations is the monthly return for the market.
3-Factor Alpha	The explanatory variables used in the alpha-estimations are the monthly returns for the Fama and French (1993) factor mimicking portfolios (3-Factor Alpha), which include the market, size, and book-to-market factor.
4-Factor Alpha	The explanatory variables used in the alpha-estimations are the monthly returns for the 4-factor factor mimicking portfolios, which include the market, size, book-to-market factor, and the additional Carhart (1997) momentum factor.
Alpha	The Alphas refer to the intercepts for regressions of monthly excess returns from the rolling strategy.
MKT	The market factor mimicking portfolio as in Fama and French (1993).
SMB	The size factor mimicking portfolio as in Fama and French (1993).
HML	The book-to-market factor mimicking portfolio as in Fama and French (1993).
UMD	The momentum factor mimicking portfolio as in Carhart (1997).
LIQ	The liquidity factor mimicking portfolio as in Pástor and Stambaugh (2003).

Note. CRSP = Center for Research in Security Prices; SIC = Standard Industrial Classification.

Acknowledgements

The authors thank Xiao-Jun Zhang (Editor-in-Chief), C. S. Agnes Cheng (Editor), Mike Weng (Associate Editor), Aaron Henrichsen, Jim Reardon, Axel Haller, seminar participants at the University of Northern Colorado, the University of Trieste, the University of Regensburg the 2020 FMA virtual conference, and two anonymous referees for their helpful comments and suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Philipp D. Schaberl

Maya A. Thevenot

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Akins

(2018). Financial reporting quality and uncertainty about credit risk among ratings agencies. The Accounting Review, 93(4), 1–22.

Barber

B. M.

Odean

(2008). All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial Studies, 21(2), 785–818.

Barinov

(2020, June). Firm complexity and conglomerate expected returns [Working Paper]. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3144717

Bens

D. A.

Berger

P. G.

Monahan

S. J.

(2011). Discretionary disclosure in financial reporting: An examination comparing internal firm data to externally reported segment data. The Accounting Review, 86(2), 417–449.

Bens

D. A.

Monahan

S. J.

Steele

L. B.

(2018). The effect of aggregation of accounting information via segment reporting on accounting conservatism. European Accounting Review, 27(2), 237–262.

Berger

P. G.

Hann

(2003). The impact of SFAS No. 131 on information and monitoring. Journal of Accounting Research, 41(2, The Effects of Regulation (Including Taxation) on Financial Reporting and Disclosure), 163–223.

Berger

P. G.

Hann

R. N.

(2007). Segment profitability and the proprietary and agency costs of disclosure. The Accounting Review, 82(4), 869–906. https://doi.org/10.2308/accr.2007.82.4.869

Berger

P. G.

Ofek

(1995.). Diversification’s effect on firm value. Journal of Financial Economics, 37, 39–65.

Bernard

V. L.

Thomas

J. K.

(1989). Post-earnings-announcement drift: Delayed price response or risk premium? Journal of Accounting Research, 27(3), 1.

10.

Bernard

V. L.

Thomas

J. K.

(1990). Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting & Economics, 13(4), 305–340.

11.

Biais

Bossaerts

(1998). Asset prices and trading volume in a beauty contest. Review of Economic Studies, 65(2), 307–340.

12.

Blankespoor

DeHaan

Wertz

Zhu

(2019). Why do individual investors disregard accounting information? The roles of information awareness and acquisition costs. Journal of Accounting Research, 57(1), 53–84.

13.

Botosan

C. A.

Stanford

(2005). Managers’ motives to withhold segment disclosures and the effect of SFAS No. 131 on analysts’ information environment. The Accounting Review, 80(3), 751–771.

14.

Carhart

M. M.

(1997). On persistence in mutual fund performance. The Journal of Finance, 52(1), 57–82.

15.

Chan

L. K. C.

Jegadeesh

Lakonishok

(1996). Momentum strategies. The Journal of Finance, 51(5), 1681–1713.

16.

Chen

T. K.

Liao

Y. P.

(2015). The economic consequences of disclosure quality under SFAS No. 131. Accounting Horizons, 29(1), 1–22.

17.

Cho

Y. J.

(2015). Segment disclosure transparency and internal capital market efficiency: Evidence from SFAS No. 131. Journal of Accounting Research, 53(4), 669–723.

18.

Cohen

Frazzini

(2008). Economic links and predictable returns. The Journal of Finance, 63(4), 1977–2011.

19.

Cohen

Lou

(2012). Complicated firms. Journal of Financial Economics, 104(2), 383–400.

20.

Dechow

P. M.

Lawrence

Ryans

J. P.

(2016). SEC comment letters and insider sales. Accounting Review, 91(2), 401–439.

21.

DeHaan

Madsen

Piotroski

J. D.

(2017). Do weather-induced moods affect the processing of earnings news? Journal of Accounting Research, 55(3), 509–550.

22.

DeHaan

Shevlin

Thornock

(2015). Market (in)attention and the strategic scheduling and timing of earnings announcements. Journal of Accounting & Economics, 60(1), 36–55.

23.

DellaVigna

Pollet

J. M.

(2009). Investor inattention and Friday earnings announcements. The Journal of Finance, 64(2), 709–749.

24.

Elliott

J. A.

Philbrick

D. R.

(1990). Accounting changes and earnings predictability. The Accounting Review, 65(1), 157–174.

25.

Epstein

M. J.

Palepu

K. G.

(1999). What financial analysts want. Strategic Finance, 80, 48–52.

26.

Ettredge

M. L.

Kwon

S. Y.

Smith

D. B.

Stone

M. S.

(2006). The effect of SFAS No. 131 on the cross-segment variability of profits reported by multiple segment firms. Review of Accounting Studies, 11(1), 91–117.

27.

Ettredge

M. L.

Kwon

S. Y.

Smith

D. B.

Zarowin

P. A.

(2005). The impact of SFAS No. 131 business segment data on the market’s ability to anticipate future earnings. Accounting Review, 80(3), 773–804.

28.

Fama

E. F.

French

K. R.

(1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3–56.

29.

Fama

E. F.

MacBeth

J. D.

(1973). Risk, return, and equilibrium: Empirical tests. Journal of Political Economy, 81(3), 607–636.

30.

Herrmann

Thomas

W. B.

(2000a). An analysis of segment disclosures under SFAS No. 131 and SFAS No. 14. Accounting Horizons, 14(3), 287–302.

31.

Herrmann

Thomas

W. B.

(2000b). A model of forecast precision using segment disclosures: Implications for SFAS No. 131. Journal of International Accounting, Auditing & Taxation, 9(1), 1–18.

32.

Hong

Lim

Stein

J. C.

(2000). Bad news travels slowly: Size, analyst coverage, and the profitability of momentum strategies. The Journal of Finance, 55(1), 265–295.

33.

Hong

Stein

J. C.

(1999). A unified theory of underreaction, momentum trading, and overreaction in asset markets. The Journal of Finance, 54(6), 2143–2184.

34.

Hope

O. K.

Kang

Thomas

W. B.

Vasvari

(2008). Pricing and mispricing effects of SFAS 131. Journal of Business Finance & Accounting, 35(3–4), 281–306.

35.

Hope

O.-K.

Thomas

W. B.

Winterbotham

(2006). The impact of nondisclosure of geographic segment earnings on earnings predictability. Journal of Accounting, Auditing & Finance, 21(3), 323–346.

36.

Hope

O. K.

Thomas

W. B.

Winterbotham

(2009). Geographic earnings disclosure and trading volume. Journal of Accounting and Public Policy, 28(3), 167–188.

37.

Hou

(2007). Industry information diffusion and the lead-lag effect in stock returns. Review of Financial Studies, 20(4), 1113–1138.

38.

Hui

K. W.

Yeung

P. E.

(2013). Underreaction to industry-wide earnings and the post-forecast revision drift. Journal of Accounting Research, 51(4), 701–737.

39.

Hund

Monk

Tice

(2021). The Berger-Ofek diversification discount is just poor firm matching. Critical Finance Review, forthcoming.

40.

Hwang

Rubesam

(2015). The disappearance of momentum. European Journal of Finance, 21(7), 584–607.

41.

Jegadeesh

(1990). Evidence of predictable behavior of security returns. The Journal of Finance, 45(3), 881–898.

42.

Kang

Khurana

I. K.

Wang

(2017). International diversification, SFAS 131 and post-earnings-announcement drift. Contemporary Accounting Research, 34(4), 2152–2178.

43.

Kim

Verrecchia

R. E.

(1997). Pre-announcement and event-period private information. Journal of Accounting & Economics, 24(3), 395–419.

44.

Kohlbeck

Warfield

(2010). Accounting standard attributes and accounting quality: Discussion and analysis. Research in Accounting Regulation, 22(2), 59–70.

45.

Kothari

S. P.

Shu

Wysocki

P. D.

(2009). Do managers withhold bad news? Journal of Accounting Research, 47(1), 241–276.

46.

Livnat

Mendenhall

R. R.

(2006). Comparing the post–earnings announcement drift for surprises calculated from analyst and time series forecasts. Journal of Accounting Research, 44(1), 177–205.

47.

A. W.

MacKinlay

A. C.

(1990). When are contrarian profits due to stock market overreaction? Review of Financial Studies, 3(2), 175–205.

48.

Louis

Sun

A. X.

(2016). Abnormal accruals and managerial intent: Evidence from the timing of merger announcements and completions. Contemporary Accounting Research, 33(3), 1101–1135.

49.

Makarov

Rytchkov

(2012). Forecasting the forecasts of others: Implications for asset pricing. Journal of Economic Theory, 147(3), 941–966.

50.

Michaely

Rubin

Vedrashko

(2016). Further evidence on the strategic timing of earnings news: Joint analysis of weekdays and times of day. Journal of Accounting & Economics, 62(1), 24–45.

51.

Moskowitz

T. J.

Grinblatt

(1999). Do industries explain momentum? The Journal of Finance, 54(4), 1249–1290.

52.

Newey

W. K.

West

K. D.

(1987). Hypothesis testing with efficient method of moments estimation. International Economic Review, 28(3), 777.

53.

Park

J. C.

(2011). The effect of SFAS 131 on the stock market’s ability to predict industry-wide and firm-specific components of future earnings. Accounting & Finance, 51(2), 575–607.

54.

Pástor

L’.

Stambaugh

R. F.

(2003). Liquidity risk and expected stock returns. The Journal of Political Economy, 111(3), 642–685.

55.

Schaberl

P. D.

(2014). The influence of disclosure policy on analyst behavior: The case of segment data. Advances in Accounting, 30(2), 440–451.

56.

Segal

(2016). Are managers strategic in reporting non-earnings news? Evidence on timing and news bundling. Review of Accounting Studies, 21, 1203–1244.

57.

Skinner

D. J.

Sloan

R. G.

(2002). Earnings surprises, growth expectations, and stock returns or don’t let an earnings torpedo sink your portfolio. Review of Accounting Studies, 7, 289–312.

58.

Street

D. L.

Nichols

N. B.

Gray

S. J.

(2000). Segment disclosures under SFAS No. 131: Has business segment reporting improved? Accounting Horizons, 14(3), 259–285.

59.

Zhang

X. F.

(2006). Information uncertainty and stock returns. Journal of Finance, 61(1), 105–137.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.06 MB