challenges and opportunities in dis

challenges and opportunities in disclosure research—A discussion of ‘the financial reporting environment: Review of the
recent literature’

Beyer et al. (2010) review the financial reporting literature related to voluntary and mandatory firm disclosures, and sell-side analyst reports. The discussion summarizes their approach, highlights some of their main conclusions, and presents alternative ideas about promising avenues for future research.
1. Introduction

The review article by Beyer, Cohen, Lys, and Walther (BCLW, 2010), summarizes a large body of literature related to disclosure choices, disclosure regulation, and disclosure intermediation by sell-side analysts. The authors provide a useful framework for thinking about research questions in these areas, point out some of the more important contributions and limitations of papers written during the past decade, and briefly suggest some general ideas for promising areas of future research. This discussion summarizes the portions of BCLW’s review related to voluntary and mandatory disclosure, discusses some related literature, and offers thoughts about promising areas for future work.
Section 2 addresses the empirical work in the BCLW review and discusses alternative empirical approaches being introduced to study related research issues. Section 3 covers voluntary disclosure models with empirical voluntary disclosure research discussed in Section 4. Section 5 reviews some of the research on models of mandatory disclosure and Section 6 explores issues in the empirical research on mandated disclosure. Concluding remarks are offered in Section 7.
2. Empirical analysis by BCLW
The BCLW survey begins with a brief empirical assessment of the relative contribution to the information reflected in security prices of five accounting disclosures: management forecasts, analyst forecasts, SEC filings, earnings guidance, and earnings announcements.1 The assessment is performed with a decomposition of the quarterly stock return variance. The decomposition shows that about 28% of quarterly stock return variance occurs on days when accounting disclosures are made and that the majority (about 55%) of the accounting-based information is provided by management forecasts. The remaining 45% of the accounting-based information is provided about equally by analyst forecasts versus the combination of earnings pre-announcements, earnings announcements, and SEC filings.
I have two main concerns with this analysis. First, it provides evidence similar to that already in Ball and Shivakumar (2008), without fully acknowledging its debt to their paper. Second, Ball and Shivakumar note (p. 1010) that the higher informativeness of management forecasts over earnings announcements is likely due to managers self-selecting when they forecast. BCLW do not address or discuss the impact of the greater selectivity of management forecasts versus earnings announcements. If managers choose to issue forecasts when they think the forecast will be more informative, the less frequent management forecasts will be more informative than earnings announcements for selection reasons.
There are some minor measurement and econometric issues with the BCLW regression. These issues are of little concern as long as one realizes that the point of the return variance decomposition approach is to give a simple and descriptive illustration of well-known points, such as earnings explaining relatively little of a period’s total return variance. The minor econometric issues are as follows. First, there is likely to be measurement error in some of the short window CARs. For example, without a control for the expected earnings announcement date, the CARs for the earnings announcements are influenced by deviations from the expected report date. Measurement error in the announcement dates results in a biased estimate of the explained return variance. Second, in a variance decomposition, the covariance between items must be grouped into the portion of the variance explained by one of the items, which can be done only arbitrarily. Third, the attempt to decompose quarterly return variance is based on coding each information item as zero whenever the specific event did not occur during the quarter. Thus, none of the return variance being decomposed can be
assigned to an information item for the firm-quarters in which the item is coded as zero.

The econometric issues are of little concern given the descriptive goal of the variance decomposition approach, and the approach has merit. It is not, however, the only possible approach for use in studying questions such as how various information events contribute to the total information conveyed over a period of time. I therefore discuss an alternate approach that can be used to address questions like those in Ball and Shivakumar (2008) and BCLW, but that can also allow potentially more interesting and complex questions to be examined.
Decomposing stock return variance from a long window (such as a quarter or a year) to assess the relative contribution of various three-day-window information items is an example of studying how a low frequency item of interest (quarterly return) is related to higher frequency variables (the information events such as management forecasts and earnings announcements). Ball and Easton (2010) develop a method for regressing low frequency data on high frequency data. Their focus is on the timeliness of the incorporation of the value changes captured by daily stock returns into earnings. The method they develop permits the possibility that the earnings–return relation changes over the earnings measurement period. In other words, instead of regressing annual earnings on annual returns and estimating a single regression coefficient to capture the average relation, Ball and Easton (2010) regress annual earnings on 251 daily returns and estimate a separate coefficient for each trading day of the fiscal year.
With regard to the findings of Ball and Shivakumar (2008) and the related Table 1 findings in BCLW, Ball and Easton (2010) use their approach to show that a much higher proportion of the stock price change on earnings announcement days is incorporated into the year’s earnings compared to the proportion of stock price change incorporated into the year’s earnings from non-earnings-announcement days. These findings suggest earnings announcements may be more important information events than the descriptive variance decomposition approach indicates.
With regard to expanding the set of researchable questions with their method, the main focus of Ball and Easton (2010) is on examining the portion of stock price change that is recognized in the earnings of the period, which they refer to as earnings recognition timeliness (ERT). By allowing the earnings–return relation to change over the earnings measurement period, Ball and Easton are able to separate two elements of ERT that have been aggregated together in previous empirical approaches. The two elements are the recognition in the current period’s earnings of the effects of (1) news regarding sales and matched expenses for the current period and (2) news that changes expectations about future periods’ earnings. Ball and Easton (2010) provide evidence that separating these two elements can provide additional insights related to the many studies that examine differences in ERT across data partitions, with the best known example being partitions based on positive return (good news) versus negative return (bad news) samples following Basu (1997).
A useful related technique is the Mixed Data Sampling (MIDAS) regression approach used in Ball (2010). MIDAS regression models were recently developed by Ghysels et al. (2006, 2007) and provide a parsimonious and flexible class of regression models that allow the dependent variable to be sampled at a low frequency (e.g., earnings announcements) while explanatory variables are sampled at a high frequency (e.g., daily returns). MIDAS regressions resolve severe


1 I appreciate comments from John Core that helped considerably in improving the writing of this section.
multicollinearity by restricting the individual coefficient estimates on the high frequency explanatory variables to be a function of a small vector of parameters. This reduces the sampling variability of the coefficient estimates and counteracts increased variability from multicollinearity.

3. Voluntary disclosure models

BCLW’s review of voluntary disclosure research raises several interesting observations. With regard to recommenda-tions for future voluntary disclosure models, I agree that assessing the impact of incentive systems on management disclosure decisions may be misleading if it is assumed that the main goal of these systems is to influence disclosure choices. BCLW note that incentive systems are instead likely to be at least as concerned with influencing investment, effort, and competitive behavior choices (and, I would add, risk-taking). BCLW rightly argue that disclosure models would benefit from incorporating these real choices.
I think it is worthwhile to push further on this theme, which relates to the fact that most of the models of disclosure choice focus on a pure exchange economy (i.e., one with no real effects on production choices). It is not clear that further modeling of pure exchange economies is likely to provide significant additional insights for our understanding of discretionary disclosure decisions. Instead, the literature seems to have reached a point where incorporating real effects on production and investment choices needs to occur to provide substantial new insights into the causes and consequences of managers’ disclosure choices.
As discussed by BCLW (see their footnote 30), until recently theoretical literature on the real effects of voluntary disclosure has focused on the effects on product market competitors. Recent working papers that instead consider the imp

Hasil (Bahasa Indonesia) 1:

tantangan dan peluang dalam penelitian pengungkapan-diskusi tentang ' financial pelaporan lingkungan: Reviewliteratur terbaru 'Garasi et al. (2010) review financial pelaporan sastra berkaitan dengan sukarela dan wajib berlandaskan pengungkapan, dan sisi-Jual analis laporan. Diskusi meringkas pendekatan mereka, menyoroti beberapa kesimpulan utama mereka, dan menyajikan ide-ide alternatif tentang jalan yang menjanjikan untuk penelitian.1. PendahuluanReview Pasal oleh garasi, Cohen, Lys dan Walther (BCLW, 2010), meringkas tubuh besar sastra yang berkaitan dengan pengungkapan pilihan, pengungkapan peraturan dan pengungkapan intermediasi oleh sisi-Jual analis. Para penulis memberikan suatu kerangka yang berguna untuk berpikir tentang penelitian pertanyaan di wilayah ini, menunjukkan beberapa kontribusi lebih penting dan keterbatasan dari karya-karya yang ditulis selama dekade terakhir, dan briefly menyarankan beberapa ide umum untuk bidang penelitian masa depan yang menjanjikan. Diskusi ini meringkas bagian dari BCLW's review berkaitan dengan sukarela dan wajib pengungkapan, membahas beberapa literatur yang terkait, dan menawarkan pikiran tentang tempat-tempat yang menjanjikan untuk masa depan pekerjaan.Alamat Bagian 2 empiris bekerja di BCLW review dan mendiskusikan alternatif pendekatan empiris yang diperkenalkan untuk mempelajari masalah-masalah penelitian terkait. Bagian 3 mencakup sukarela pengungkapan model dengan penelitian empiris sukarela pengungkapan yang dibahas dalam Bagian 4. Bagian 5 ulasan beberapa penelitian pada model kewajiban untuk keterbukaan dan Bagian 6 mengeksplorasi masalah dalam penelitian empiris pada perintah pengungkapan. Kesimpulan yang ditawarkan di bagian 7. 2. empiris analisis oleh BCLWSurvei BCLW dimulai dengan penilaian empiris yang singkat kontribusi relatif untuk reflected informasi keamanan harga five akuntansi pengungkapan: manajemen prakiraan, analis prakiraan, SEC filings, penghasilan bimbingan dan penghasilan announcements.1 penilaian dilakukan dengan dekomposisi saham kuartalan kembali varians. Dekomposisi menunjukkan bahwa sekitar 28% dari kuartalan varians kembali saham terjadi pada hari-hari ketika membuat pengungkapan akuntansi dan bahwa sebagian besar (sekitar 55%) informasi berbasis akuntansi yang disediakan oleh manajemen Prakiraan. Sisa 45% informasi berbasis akuntansi yang disediakan tentang sama oleh analis Prakiraan versus kombinasi penghasilan pra-pengumuman, pengumuman laba dan SEC filings.I have two main concerns with this analysis. First, it provides evidence similar to that already in Ball and Shivakumar (2008), without fully acknowledging its debt to their paper. Second, Ball and Shivakumar note (p. 1010) that the higher informativeness of management forecasts over earnings announcements is likely due to managers self-selecting when they forecast. BCLW do not address or discuss the impact of the greater selectivity of management forecasts versus earnings announcements. If managers choose to issue forecasts when they think the forecast will be more informative, the less frequent management forecasts will be more informative than earnings announcements for selection reasons.There are some minor measurement and econometric issues with the BCLW regression. These issues are of little concern as long as one realizes that the point of the return variance decomposition approach is to give a simple and descriptive illustration of well-known points, such as earnings explaining relatively little of a period’s total return variance. The minor econometric issues are as follows. First, there is likely to be measurement error in some of the short window CARs. For example, without a control for the expected earnings announcement date, the CARs for the earnings announcements are influenced by deviations from the expected report date. Measurement error in the announcement dates results in a biased estimate of the explained return variance. Second, in a variance decomposition, the covariance between items must be grouped into the portion of the variance explained by one of the items, which can be done only arbitrarily. Third, the attempt to decompose quarterly return variance is based on coding each information item as zero whenever the specific event did not occur during the quarter. Thus, none of the return variance being decomposed can beassigned to an information item for the firm-quarters in which the item is coded as zero.The econometric issues are of little concern given the descriptive goal of the variance decomposition approach, and the approach has merit. It is not, however, the only possible approach for use in studying questions such as how various information events contribute to the total information conveyed over a period of time. I therefore discuss an alternate approach that can be used to address questions like those in Ball and Shivakumar (2008) and BCLW, but that can also allow potentially more interesting and complex questions to be examined.Decomposing stock return variance from a long window (such as a quarter or a year) to assess the relative contribution of various three-day-window information items is an example of studying how a low frequency item of interest (quarterly return) is related to higher frequency variables (the information events such as management forecasts and earnings announcements). Ball and Easton (2010) develop a method for regressing low frequency data on high frequency data. Their focus is on the timeliness of the incorporation of the value changes captured by daily stock returns into earnings. The method they develop permits the possibility that the earnings–return relation changes over the earnings measurement period. In other words, instead of regressing annual earnings on annual returns and estimating a single regression coefficient to capture the average relation, Ball and Easton (2010) regress annual earnings on 251 daily returns and estimate a separate coefficient for each trading day of the fiscal year.Berkaitan dengan findings bola dariyolanda (2008) dan terkait tabel 1 findings BCLW, bola dan Easton (2010) menggunakan pendekatan mereka untuk menunjukkan bahwa proporsi yang jauh lebih tinggi dari perubahan harga saham pada hari-hari pengumuman pendapatan dimasukkan ke dalam penghasilan tahun dibandingkan dengan proporsi perubahan harga saham dimasukkan ke dalam tahun pendapatan non-penghasilan-pengumuman hari. findings ini menyarankan pengumuman laba mungkin peristiwa informasi lebih penting daripada pendekatan dekomposisi varians deskriptif yang menunjukkan.Berkaitan dengan memperluas serangkaian pertanyaan yang diteliti dengan metode mereka, fokus utama bola dan Easton (2010) adalah memeriksa bagian perubahan harga saham yang diakui dalam pendapatan periode, yang mereka sebut sebagai penghasilan pengakuan ketepatan waktu (ERT). Dengan membiarkan hubungan laba-laba untuk mengubah selama periode pengukuran penghasilan, bola dan Easton yang mampu memisahkan dua elemen ERT yang telah dikumpulkan bersama-sama dalam pendekatan empiris sebelumnya. Dua elemen adalah pengakuan dalam periode saat ini penghasilan efek dari (1) berita mengenai penjualan dan cocok pengeluaran untuk periode saat ini dan (2) berita bahwa perubahan harapan tentang penghasilan masa depan periode. Bola dan Easton (2010) memberikan bukti yang memisahkan kedua unsur ini dapat memberikan tambahan wawasan yang berkaitan dengan banyak penelitian yang mengkaji perbedaan ERT di partisi data, dengan contoh terkenal menjadi partisi berdasarkan kembali positif (berita baik) versus negatif kembali (berita buruk) sampel mengikuti Basu (1997).Teknik terkait yang berguna adalah pendekatan regresi campuran Data Sampling (MIDAS) digunakan dalam bola (2010). Model regresi MIDAS baru-baru ini dikembangkan oleh Ghysels et al. (2006, 2007) dan menyediakan kelas terlalu kikir dan flexible model regresi yang memungkinkan variabel untuk dicicipi di frekuensi rendah (misalnya, pengumuman laba) sementara jelas variabel sampel pada frekuensi tinggi (misalnya, pengembalian harian). Regresi MIDAS menyelesaikan parah1 I appreciate comments from John Core that helped considerably in improving the writing of this section.multicollinearity by restricting the individual coefficient estimates on the high frequency explanatory variables to be a function of a small vector of parameters. This reduces the sampling variability of the coefficient estimates and counteracts increased variability from multicollinearity.3. Voluntary disclosure modelsBCLW’s review of voluntary disclosure research raises several interesting observations. With regard to recommenda-tions for future voluntary disclosure models, I agree that assessing the impact of incentive systems on management disclosure decisions may be misleading if it is assumed that the main goal of these systems is to influence disclosure choices. BCLW note that incentive systems are instead likely to be at least as concerned with influencing investment, effort, and competitive behavior choices (and, I would add, risk-taking). BCLW rightly argue that disclosure models would benefit from incorporating these real choices.I think it is worthwhile to push further on this theme, which relates to the fact that most of the models of disclosure choice focus on a pure exchange economy (i.e., one with no real effects on production choices). It is not clear that further modeling of pure exchange economies is likely to provide significant additional insights for our understanding of discretionary disclosure decisions. Instead, the literature seems to have reached a point where incorporating real effects on production and investment choices needs to occur to provide substantial new insights into the causes and consequences of managers’ disclosure choices.As discussed by BCLW (see their footnote 30), until recently theoretical literature on the real effects of voluntary disclosure has focused on the effects on product market competitors. Recent working papers that instead consider the imp