# «Leif Atle Beisland University of Agder Dissertation submitted to the Department of Accounting, Auditing and Law at the Norwegian School of Economics ...»

Companies not reporting consolidated statements can postpone the transition until 2007, at the latest. Companies not reporting consolidated financial statements because they have no subsidiaries may continue to report according to NGAAP in their company reports.

Definition of Variables in the Price and Return Regression, Including Control

**Panel B:**

Variables PRICE The stock price of firm i = 1, 2, …, N at the end of the accounting year t = 2003, 2004, 2005, 2006.

N, the number of firms in the sample, equals 251.

BOOK Firm i’s book value of equity per share at the end of year t (including provision for proposed dividend not reported as equity).

RET The excess logarithmic stock return of firm i in year t, where excess means return in excess of the estimated risk free rate. The risk free rate of return is taken to be the one month effective Nibor rate (which is an interbank borrowing rate), adjusted for 28% tax and a risk premium of 10%, i.e. the risk free rate = Nibor01M · (1 - 0.28) · (1 - 0.1).

EARN Firm i’s earnings per share during year t divided by the previous year’s stock price, at the end of that year. In the price regression, EARN’ is earnings divided by the number of outstanding shares at year end. Thus, EARN = EARN’/PRICEt-1.

IFRS Indicator or dummy variable which equals 1 if the firm reports according to IFRS, or 0 if the firm reports according to NGAAP in a particular year.

LOSS Indicator variable which equals 1 if EARN 0, and 0 if EARN 0.

INTAN Indicator variable for a firm belonging to an industry with presumably high intangible asset intensity, particularly biotechnology, information technology and communication industries. The industry classification of the companies has been performed by the authors.

TRAN Indicator variable for a firm with transitory earnings (deflated by the prior year’s stock price) below the lower quartile or above the upper quartile. Transitory items are comprised of impairments, other unusual operating items – e.g. large gains from the sale of operational assets and restructuring charges, special income from associated companies, special or non-recurring financial items – e.g.

gains and losses on financial instruments and currency gains and losses, and other unusual items reported in the income statement, including the net result from discontinued operations and extraordinary items. The transitory component of earnings has been classified by the authors based on information given in the notes of the financial statements.

BETA Beta is an estimate of systematic risk, where BETA is estimated from the time series 60 months before the year end. If less data is available, BETA is estimated from shorter time series.

SIZE Firm size is a proxy risk factor and is measured by the logarithm of previous year’s market value of equity.

BTM Book-to-market ratio is a proxy risk factor. The book-value of equity is reduced by earnings (which is represented through the variable EARN) and the market value equals the previous year’s market value.

MOM Momentum is a proxy risk factor. In the return regression it is measured as the previous year’s excess return, i.e. MOM equals RET lagged by one year. In the price regression MOM is the lagged stock price.

The truncation of the stock price and corresponding book value leads to a reduction in the sample of 16 or 2.2%, which is lower than 4.0% (equals 2 tails each of 1% · 2 variables) due to overlapping observations. Similarly,

Panel A of Table 1 presents the sample selection. To avoid outliers having unreasonably large influence on the empirical results, the sample is truncated. In the price regression, the upper and lower percentile of stock price and book value of equity are deleted. Equivalently, the upper and lower percentile of stock return and price-deflated earnings are deleted in the return regression. The truncation is performed separately for the NGAAP and the IFRS sample. As there is some overlap among extreme observations, the final sample for the price regression consists of 725 firm-year observations – a reduction of 2.2%; 383 are IFRS and 342 are NGAAP observations. The final sample for the return regression is 629 firm-year observations – a reduction of 3.4%, 319 observations are according to IFRS and 310 are according to NGAAP.

3.2 Distributional Characteristics Table 2 lists distributional statistics for the variables entering into the price and return regression, main variables as well as control variables; see Panel B of Table 1 for a list of variable definitions. Data is displayed for the selected sample as well as for the IFRS and NGAAP subsamples.

P/B is the price/book ratio, i.e. PRICE/BOOK in which BOOK 0, P/E is the price/earnings ratio, i.e. PRICE/ EARN’ in which the earnings per share EARN’ 0. All other variables are defined in Panel B of Table 1. Obs means the number of observations, mean is the sample average of the variable, st.dev. is the corresponding standard deviation, 25-percent is the first quartile, median is the second quartile, and 75-percent is the third quartile.

stock price and the two key accounting numbers have distributions skewed to the right. The mean price, book value and earnings are in the interval between the median and the third quartile for the total sample, as well as for the two subsamples. The standard deviations are generally very high.

Focusing on the median, which is a better representation of the middle of skewed distributions than the mean, the price/book ratio is 2.024. The median price/book ratio is 2.368 for the IFRS and 1.715 for the NGAAP observations. This is somewhat surprising. Since IFRS embraces more recognition and measurement at fair value, we would expect that NGAAP would yield the highest price/book ratio. The median price/earnings ratio is 14.876 and very similar in the two subsamples, 14.719 for the IFRS versus 14.979 for the NGAAP sample. Especially in the IFRS sample, there are a few extreme observations, making the mean relatively high.

Notice that we have not truncated the sample on the basis of the price/book or price/earnings ratios.

Panel B reveals that we are examining a period of extraordinarily high stock return. The average stock market return is 31.4%, measured by logarithms and in excess of our proxy for the The average book value is 60.144 according to NGAAP and 40.688 according to IFRS. This might seem strange as far as there are many observations from the same firms reporting according to NGAAP in the years 2003-2004 and IFRS in the years 2005-2006. For these firms, we would expect the BOOK to be higher for the IFRS sample due to more recognition and measurement at fair value. Explanations for the opposite may be stock splits and changes of reporting currency from NOK to EUR. In addition, newly listed firms tend to have lower BOOK than firms delisted from the OSE.

Proposed dividend is included in the book value. According to NGAAP, proposed dividend is accounted for as a short term provision under short term debt. Thus, we add such provisions back to equity, as dividends to investors are certainly not debt in their view. According to IFRS, proposed dividend is only considered as ‘debt’ when it is decided by the general assembly. Therefore proposed dividend is almost always reported as equity at the end of the accounting year according to IFRS.

viate considerably from their long-term means. Note for instance the relatively poor earnings yield for the companies; 2.5% on average. The median is higher; 7.4 % of the market value of equity. The earnings distribution appears to be skewed to the left, especially for the NGAAP sample. Thus, there is a tendency of more loss reporting in the NGAAP sample, i.e. in 2003 and 2004.

Panel C presents the distributional statistic of the control variables entering into the return regression.30 The first variable IFRS is an indicator or dummy variable which equals 1 if the observation comes from the IFRS sample, and 0 if it comes from the NGAAP sample. The average value of this variable is 0.507, suggesting that the two subsamples are almost equally large.

The next three variables are related to properties of earnings and book values of equity - and may function as moderators for their response coefficients; see Hayn (1995), Lev and Zarowin (1999) and Elliot and Hanna (1996). LOSS is an indicator variable for negative earnings, i.e. losses. Panel C reveals that 22.6% of the observations are losses. INTAN is equal to 1 if the firm belong to industries with an a priori high intensity of intangible assets, for example biotechnology, information technology or communications, and zero otherwise.31 The percentage of ‘new economy’ observations in the total sample is estimated at 25.1. TRAN is In terms of plain stock market return the average is 50.8% and the median is 33.2%. The return of the valueweighted stock market index on the OSE, i.e. the OSEBX, was 48.4 %, 38.4 %, 40.5 %, and 32.4 % in the years 2003 to 2006. OSEBX consists of a representative selection of exchange listed companies at the OSE with high liquidity. Note that the high stock return might be an explanation for the higher price/book ratio under IFRS than under NGAAP. When the return is high, stock prices typically grow faster than the book equity does.

We apply control variables in the price regression as well. However, as the number of observations differs between the two regression specifications, the descriptive statistics for the control variables will not be completely identical in the two sets of analyses. This difference is, though, minor, and we present descriptive statistics only for the control variables that enter the return regression.

Note that the BTM ratio is also related to intangible assets in terms of non-capitalized intangibles. In general, conservative accounting will depress the BTM ratio.

upper quartile, i.e. the most extreme observations, and zero otherwise.32 Thus, the proportion of observations related to extensive transitory earnings is constructed to be approximately 50.0%; the mean turns out to be 49.4%.

The next four variables are risk factors or proxy risk factors expected to influence expected returns. BETA is the beta from the Capital Asset Pricing Model, and it is estimated as the market model beta from time series of monthly stock market returns; see also Panel B of Table 1. The average beta is 1.008. SIZE is the logarithm of the stock market value of the firm at the beginning of the year – and is a measure of firm size. Average value of SIZE is 6.671. The average market value of equity is slightly above NOK 7 billion. BTM is the adjusted book-tomarket ratio; the average is 0.878. According to Fama and French (1992), both SIZE and BTM are proxy risk factors. MOM is the previous year’s stock return – and functions as a measure of return momentum; see e.g. Carhart (1997). The average momentum value is 0.180.

3.3 Simple Correlations Panels A and B of Table 2 display the correlation coefficients between the variables applied in the price regression and the return regression, respectively. Correlations are presented both for the IFRS and for the NGAAP sample, below and above the diagonal in the two matrices.

The transitory component of earnings has been classified by the authors based on information given in the notes of the financial statements, see Table 1.

Correlation coefficients for the NGAAP sample are presented above the diagonal; the correlations for the IFRS sample are presented below the diagonal. All variables are defined in Panel B of Table 1; EARN’ is an additional control variable in the price regression; see Easton and Harris (1991); BTM is not a valid control variable in the price regression. Statistical significance at the 10% level is indicated by one asterisk * (weakly significant), at the 5% level by two asterisks ** (significant) and at the 1% level by three asterisks *** (highly significant), tested two-sided.

Panel A reveals that stock price is highly correlated with the per share book value of equity in both subsamples. The correlation coefficient is 0.863 for the IFRS and 0.884 for the NGAAP sample. The squared correlation coefficient equals the explanatory power, the R2, in a regression of one of the variables on the other, suggesting that the book value explains 74.4% in the IFRS and 78.1% in the NGAAP sample. The difference of about 3.7 percentage points (= 0.88392 - 0.86282) in favour of reporting according to NGAAP is not ‘significant’ (by the Cramer (1987) test). The R2 from a regression of stock price on earnings per share EARN’ is

0.564 and 0.632 for the IFRS and NGAAP observations, respectively. Nevertheless, the difference of 6.8 percentage points in favour of NGAAP is not ‘significant’. Panel A also presents binary correlation with and between the other control variables.

under IFRS as well as under NGAAP. The correlation coefficients between return and earnings are 0.294 and 0.159, respectively. This implies a difference in R2 of 6.1% – though it is not ‘significant’. Notice also that stock market returns are highly correlated with most of the control variables – both earnings moderators and risk factors. For example, RET is negatively correlated with LOSS and positively correlated with BETA.

4. Empirical Findings In this section, we formally test Hypothesis 1 and 2 by employing the test methodology suggested by the response coefficients (3) and (6). In the next section, we perform tests in which the book value of equity and earnings have been disaggregated into underlying accounting items to find out more about the underlying sources of the main results obtained in subsections 4.1 and 4.2. Some robustness tests are presented along the main analyses, related to estimation technique, outlier effects and constant sample.

4.1 Test of Hypothesis 1 Hypothesis 1 says that the book response coefficient BRC is different when financial reports are prepared according to IFRS than when prepared according to NGAAP. The hypothesis is tested using the test methodology (1) - (3), focusing on whether α4 ≠ 0 and statistical significant. The results from running regression (1) and its counterpart with control variables are presented in Table 4.