下载掌阅APP，畅读海量书库
立即打开
The primary source of the financial data is the Bureau van Dijk’s Amadeus database.This database provides standardized financial statement data for approximately 21 million firms across Europe，and is consistent with several well-established national information collections.The main advantage of the Amadeus database is that it includes privately-held firms，which allows us to focus on an economically important group of firms.
This book compares EM variation between private and public firms by using a large cross-country European dataset of financial statements.The comprehensive information on these firms is gathered across firms based in all the former 28 European Union member states（including the UK），and focuses primarily on private firms.It is based on business accounts filed in national registers and collected by national contact points before being fed into the database.The origin of the data explains that it contains mostly financial information，with a large number of variables available to assess the balance sheets and income statements of each firm.
Except for financial information，Amadeus encompasses additional information on auditing and accounting standards used for the preparation of financial statements.
There are three main advantages for using Amadeus database.Firstly，as mentioned above，it includes privately-held firms，which allows focus on an economically important group of firms.Secondly，it provides a broad coverage at both firm and country level，and enables their proprietary procedures to be studied to ensure consistency among the accounting items from different countries and across accounting standards.Thirdly，Amadeus is updated weekly，providing standardised annual accounts with up to ten years’archive.Data is collected and collated by Bureau van Dijk/Moody’s Analytics from company filings and reports.To analyse the impact of financial analysis，this book merges the 2010 version and 2014 version，so that the final sample spans 2001-2013.
The sample used in this book covers 11 European Union countries，for the period from 2001 to 2013.These countries are Belgium，Finland，France，Germany，Greece，Italy，the Netherlands，Spain，Portugal，Sweden and the United Kingdom.Following the study of Leuz et al.（2003）and Burgstahler et al.（2006），EM has been classified into four categories in the first empirical chaper：（1）hiding small losses，（2）the magnitude of total accrual，（3）smoothness of earnings and（4）the correlation of accounting accruals and operating cash flows.Each one of these proxies can identify a specific reason that may lead firms to manipulate earnings.Therefore，they might lead to different results in the increase or decrease in EM，and reduce potential measurement error in the individual scores.Most of the analysis is based on an aggregate measure of EM.
To further test the institutional factors influencing the level of EM，the following institutional factors were considered，i.e.，（1）country-level enforcement；（2）country-level tax rates；（3）country-level investor protection；（4）country-level accrual accounting rules and（5）absence index.Finally，T-tests are used here to re-estimate to examine whether the study’s results are robust in respect of these different measures.
The second empirical chapter focuses on accrual quality and conservatism.Employing alternative measures can help to ensure that the findings are not driven by measurement error in any one particular variable，and mitigate the possibility that the results captured by using one particular proxy are also driven by other factors.
For accrual quality，it followed the study of Hope et al.（2013）by using two of the most popular methods in the literature，which are the Dechow-Dichev model（modified Jones model with operation cash flow）in Dechow and Dichev（2002）and the Performance matched model（modified Jones model with return on assets）in Kothai et al.（2005）.In order to increase reliability of the test，revenue-accrual quality（McNichols and Stubben，2008）and the ratio of the magnitude of accruals to cash flow（Burgtahler et al.，2006）are employed.For conditional conservatism，the two models in Ball and Shivakumar（2005）are used.
Then，conditional analyses have been employed to test the firm-level factors affecting differential EM between private and public firms.These include measures associating with that（1）the greater accrual quality of public firms relative to private firms is mitigated when：（a）meeting earnings benchmarks，（b）obtaining external financing in following years and（c）employing no Big Four auditors；（2）the greater conservatism of public firms relative to private firms decreases when：（a）just beating earnings benchmarks and（b）having lower leverage.Finally，a propensity score matches（PSM）approach has been employed.
The PSM approach generates samples in which private and public firms are more similar，which helps mitigate concerns that omitted correlated variables were driving the results.
As we know，managers are able to influence reported earnings not only by accounting-based activities but also real activities.So the third empirical chapter employs REM to exam firms’earnings manipulation.The measurement of real earnings management was built from Roychowdhury（2006），and also followed Cohen et al.（2008）and Cohen and Zarowin（2010）for the composite measures.Three main categories of REM are used here：Sales manipulation（ACFO），abnormal production（APROD）and managing discretionary expenses（ADISEXP）.It then constructed three composite measures：REM1 represents the sum of the standardised APROD and ADISEXP ；REM2 includes the sum of the standardised ACFO and ADISEXP ；REM_SUM is all measures aggregated and used to examine the total effect of REM.All measures were constructed so that a positive figure is associated with income increasing REM，and vice versa with a negative number.Following prior research，several cross-sectional factors（such as the effects of benchmark beating，the need for external financing，leverage and big loss）that could affect differential incentives for EM between private and public firms are considered in this study.Finally，the regression models are re-estimated using a stricter PSM procedure which created a more closely matched sample，in order to correct for sample selection bias.