Chapter 5. Research Design and Methodology – Quantitative and Qualitative analysis
5.6. Multiple regression models
The empirical analysis involves pooled OLS regressions with standard errors robust to heteroskedasticity (White’s robust errors). The issue of outliers is addressed by winsorising at the 1st and 99th percentile for all continuous variables. Two versions of the model were run, one where the dependent variables was forecast accuracy (𝑴𝑭𝑨𝒊𝒕) and one where the dependent variable was forecast dispersion (𝑴𝑭𝑫𝒊𝒕). The basic regression model is:
𝑫𝒆𝒑𝒆𝒏𝒅𝒆𝒏𝒕 𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆 = 𝜶𝟎+ 𝜶𝟏𝒍𝒏(𝑵𝑶𝑨)𝒊𝒕+ 𝜶𝟐𝒍𝒏(𝑺𝑰𝒁𝑬)𝒊𝒕+ 𝜶𝟑𝒍𝒏(𝑨𝑮𝑬)𝒊𝒕 + 𝜶𝟒𝑩𝑴𝒊𝒕+𝜶𝟓𝑫𝑬𝒊𝒕+ 𝜶𝟔𝑹𝑶𝑨𝒊𝒕+ 𝜶𝟕𝑹𝑬𝑻𝑼𝑹𝑵𝑺𝒊𝒕 + 𝜶𝟖𝑽𝑶𝑳𝑨𝑻𝑰𝑳𝑰𝑻𝒀𝒊𝒕+ 𝜶𝟗𝑰𝑮𝑾𝒊𝒕+ 𝜶𝟏𝟎𝑵𝑰𝑨𝑰𝒊𝒕+ 𝜶𝟏𝟏𝑬𝑰𝑮𝑫𝒊𝒕 + 𝜶𝟏𝟐𝑵𝑨𝑺𝑨𝑪𝑸𝒊𝒕+ 𝜶𝟏𝟑𝒍𝒏(𝑴𝑹𝑲𝑻)𝒊𝒕+ 𝜶𝟏𝟒𝒍𝒏(𝑭𝑹𝑴𝑺𝑻𝑹)𝒊𝒕 + 𝜶𝟏𝟓𝒍𝒏(𝑪𝑶𝑹𝑷𝑮𝑶𝑽)𝒊𝒕+ 𝜶𝟏𝟔𝒍𝒏(𝑩𝑹𝑵𝑫)𝒊𝒕+ 𝜶𝟏𝟕𝒍𝒏(𝑭𝑰𝑵𝑷𝑶𝑺)𝒊𝒕 + 𝜶𝟏𝟖𝒍𝒏(𝑹𝑬𝑮𝑨𝑪𝑪)𝒊𝒕+ 𝜶𝟏𝟗𝑰𝑭𝑹𝑺𝒊𝒕+ 𝜺𝒊𝒕
The dependent variables are:
(MFE Mean) is the percentage difference between the analysts’ mean earnings per
share forecast each month and the reported earnings per share at the end of the financial year.
(MFA Mean) is the absolute value of (MFE Mean).
(MFE Median) is the percentage difference between the analysts’ median earnings per share forecast each month and the reported earnings per share at the end of the financial year.
(MFA Median) is the absolute value of (MFE Median).
(MFD) is the standard deviation of all analysts’ earnings forecasts in each month. (NOA) is the number of analyst estimations each month for EPS FY1.
The model is ran one time for each dependent variable using the following independent variables.
The independent variables are:
(ln(NOA)) is the number of analyst estimations each month for EPS FY1.
(ln(SIZE)) is the log of firm i’s market capitalisation at the end of each financial year.
(ln(AGE)) is the firm’s age measured as the natural logarithm of the number of valid annual return observations from Datastream.
(BM) is the market to book ratio from Datastream measured as market value over net assets.
(DE) is the debt to equity ratio from Datastream measured as total debt over common equity.
(ROA) is the return on assets ratio from Datastream measured as net profit over total assets.
(RETURNS) is the percentage change of the stock price at the end of each financial year.
(VOLATILITY) is the stock price volatility over the company’s financial year.
(IGW) is the goodwill intensity measured as gross goodwill over total assets.
(NIAΙ) is the intangible assets intensity, measured as net intangible assets over total assets.
(EIGD) is the goodwill impairments scaled by EBITDA.
(NASACQ) measures net assets from acquisitions scaled by total assets.
(LN(MRKT)) = the natural logarithm of the cumulative disclosure proxies for market risk, industry analysis and competitive forces, over the company’s fiscal year.
(LN(FRMSTR)) = the natural logarithm of the cumulative disclosure proxies for firm strategy, product market performance, performance of business strategy model, over the company’s fiscal year.
(LN(CORPGOV)) = the natural logarithm of the cumulative disclosure proxies for human and organisational capital, management performance, corporate governance and leadership, over the company’s fiscal year.
(LN(BRND)) = the natural logarithm of the cumulative disclosure proxies for market recognition, power and consistency of brand, over the company’s fiscal year.
(LN(FINPOS)) = the natural logarithm of the cumulative disclosure proxies for corporate and business performance and financial position, over the company’s fiscal year.
(LN(REGACC)) = the natural logarithm of the cumulative disclosure proxies for government regulation, accounting regulation, disclosure practices, over the company’s fiscal year.
(IFRS) is an indicator variable = 1 if firm i reports under IFRS in year t, and 0 otherwise. For voluntary adopters in Germany it represents the financial year of 2005; when IFRS was mandated in the EU.
For German companies only:
(MANDG) is an indicator variable that takes the value of 1 if the company is a German mandatory adopter and 0 otherwise.
(VOLUG) is an indicator variable that takes the value of 1 if the company is a German voluntary adopter and 0 otherwise.
To construct the above independent variables, financial data were obtained from Thomson Reuters Datastream, Thomson One Banker and the published annual reports of individual companies. From Datastream we obtained: market capitalisation at the financial year end (Worldscope item WC08001); PE ratio (Datastream item PE); gross goodwill (Worldscope itemWC02502); amortisation and impairment of goodwill (Worldscope item WC18224); base date (since when Datastream had available data for the firm) (Datastream item BDATE); market to book value (Datastream item MTBV); stock price volatility (Worldscope item WC08806); stock price at financial year end (Worldscope item WC05001), Net assets from acquisitions, (WC04355Total Assets worldscope item WC02999), Total Debt as percentage of Common Equity (Worldscope item WC08231), Return On Assets (Worldscope item WC08326), Total Net Intangibles from Thomson One Banker, EBITDA (Worldscope item WC18198). Also, information for the IFRS voluntary adoption, IFRS financial year adoption and the IFRS transition restatements is obtained from the annual reports. Any 0/0 calculations are replaced by 0 such as intangibles intensity and goodwill impairments that lead to DIV0.
Additional checks for missing gross goodwill values were completed by subtracting amortisation and impairment of goodwill from gross goodwill values. Where needed values were replaced with the correct values obtained from the annual reports. Also, other missing values were replaced such as market capitalisation by multiplying the Market to Book ratio with book values. Additionally, missing Return On Assets values were replaced by obtaining the values from Morningstar.com. Also, a number of companies had in some cases less or more than 12 month forecasts for a given financial year. To deal with this issue the last 12 monthly forecasts for each financial year were obtained and the missing values were filled by substituting with the average forecast error across all years.
5.7. Summary
This chapter provided an explanation of how this research project aims to make a contribution to the academic literature in the area of the impact of IFRS adoption on analysts’ earnings forecasts. Based on the previous literature, we outline the development of the hypotheses of this project. Also, we explain in detail the development of a new custom dictionary to analyse the narratives of the company reports. Also, the chapter includes the detailed procedure for the data collection, sample selection and the development of the econometric models employed in the empirical analysis. The next two chapters contain the empirical analysis, investigating a) the impact of IFRS adoption on the analysts’ earnings forecasts across the whole sample period and b) the impact of IFRS standards’ revisions on the analysts’ earnings forecasts during the post IFRS adoption.
Chapter 6. IFRS Adoption, Corporate Disclosure and Analysts’ Information