«Leif Atle Beisland University of Agder Dissertation submitted to the Department of Accounting, Auditing and Law at the Norwegian School of Economics ...»
Table description Table 6 describes the value relevance of accounting information for a sample of Swedish firms in the time-period 1979 to
2004. It summarizes the number of observations (N), regression coefficients (a1, a2, a3 and a4) and the total explanatory power (R2TOT) for the total sample, the traditional industries sample and the non-traditional industries sample. Firms are classified into industries following Table 1. Each Panel presents data for individual years, the mean for all years, pooled results for 5year periods and pooled results for the whole 25-year period. The highlighted years refer to the “IT-bubble” years. Data is
analyzed using a Price model specification:
where Pit is the share price of firm i in period t, BVS is the book value per share, and EPS is the net earnings per share. A
sustainable component of reported earnings is estimated using the model:
The sustainable earnings are divided by the number of outstanding shares at time t. The difference between reported and sustainable earnings is the transitory component of earnings (EPS - SEPS).
The annual means for the non-traditional industries are computed for the period 1985-2004 due to few observations in the first years of the sample period. Boldface denotes significance at a 10% level, two-sided test. R2TOT is set equal to zero if negative.
t-test for difference in the mean adjusted R2 between traditional and non-traditional industries:
0.941 for the Price model specification (p-value, in favour of the non-traditional industries) While the long term average value relevance does not differ between traditional and nontraditional industries, it seems as if the larger and more frequently occurring transitory components makes the associations more unstable in the non-traditional industries. We expect
industries varies more over time. The standard deviations of annual price and return regressions (from Tables 3 and 4) are presented in Panel A of Table 7. The standard deviation of the mean annual R2 is 0.15/0.16 for the complete sample when the price model specification is considered. When the sample is divided into traditional and non-traditional industries, the standard deviations amount to respectively 0.09/0.10 and 0.22/0.20. All price model specifications yield statistically significant differences (p-values of 0.003 and 0.000, respectively).1112 Non-traditional industries have a considerably higher variability in the association between accounting information and share prices than traditional industries. The return model specifications yields smaller and statistically insignificant differences (p-values of 0.186 and 0.714, respectively), suggesting a higher variance in the non-traditional industries. In addition, Panel B of Table 7 shows that the Mean Absolute Deviation is higher for non-traditional industries for both price and return model specifications, with and without an adjustment for negative earnings. As in Panel A, the differences are smaller for return model specifications.
In response to the critique made by Gu  that comparisons of value relevance (as measured by R2) across samples are biased we use his suggested alternative measure; a scaleadjusted RMSE (i.e., RMSE minus RMSE in the appropriate scale decile), as the metric of value relevance. This alternative measure does not alter any of the conclusions and we therefore present R2s throughout the paper as they can be related to past studies and have a more intuitive interpretation. When the RMSE and R2 are as highly correlated as in this study, the problems suggested by Gu  vanishes. In summary, the RMSE measure When applying a standard two-sample variance comparison F-test.
The models including sustainable earnings are no different. There is almost no decrease in standard deviation for the traditional industries (0.09 for the specifications in Tables 5 and 6) and a slight decrease for nontraditional industries (to 0.19 and 0.18 respectively). Differences between the two industry categories remain statistically significant. The same holds for our tests based on RMSE as suggested by Gu .
Table description Table 7 shows the variation in value relevance of accounting information for a sample of Swedish firms in the time-period 1979 to 2004 using both Price and Return model specifications. The model without a dummy is based on R2s from Panels A, B and C of Table 3, and the model with a dummy is based on R2s from Panels A, B and C of Table 4. Both panels are computed with information from the years 1985-2004 (due to few observations for the non-traditional industries in the first years of the sample). Panel A shows the standard deviations of the annual adjusted R2 for each model. These standard deviations are measures of the over-time variability in value relevance for the various sub-samples. Panel B shows an alternative measure; the mean absolute deviation (MAD), of the annual adjusted R2 for each model.
F-test for difference in the variance of adjusted R2 between traditional and non-traditional industries:
suggests that there is no difference in the average annual explanatory power when using a price model specification (p=0.909) and a return model specification (p=0.230, but the difference is in favour of non-traditional industries). The RMSE measure suggests that there is a difference in the variations (p=0.020 for both model specifications).13 As mentioned earlier, the period 1997-2000 – popularly referred to as the “IT-bubble” – was given much attention in media. It was a particularly dramatic period in the history of equity markets and during these years the “new economy” was often discussed, also by accounting researchers (see e.g. Core, Guay and Van Buskirk ). Table 4 shows that the mean R2 for Details on these tests can be obtained from the authors upon request. The p-values refer to the regression specifications applied in Table 4.
1997 to 2000. However, for non-traditional industries, the mean R2 is only 29%. While the explanatory power decreased for both industry categories in this turbulent period we note that both the period preceding and following the IT-bubble display much higher value relevance for both industry categories. In fact, the average for these years (1993-96 and 2001-04) is 64% for the traditional industries and 74 % for the non-traditional industries. While all industries suffered in the IT-bubble years the non-traditional industries suffered considerably more.
What is really surprising is that the surrounding years make the average 1993 to 2004 explanatory power higher in the non-traditional industries. The finding is unlikely to be caused by a sample selection bias as (1) the explanatory power is higher both before and after the IT-bubble, and (2) the number of firms in the non-traditional industries continued to increase in the years after the IT-bubble years, while the number of firms in the traditional industries decreased.
The return models display slightly different results. Both model specifications (with and without a consideration for negative earnings) show that the non-traditional industries provide more value relevant information in the last 12 years (1993-2004). During the IT-bubble years the value relevance is considerably higher for non-traditional industries and it is only in the years after the IT-bubble that firms in traditional industries experience higher value relevance.
Firms in the non-traditional industries are those that provide the most value relevant accounting information in the IT-bubble years.
In the price regressions, the incremental value relevance of book value decreases substantially in the years 1997 to 2000, while the incremental explanatory power for earnings remain at almost the same level. More specifically, it is the incremental value relevance of BVS that is
particularly during a boom, when stock prices are at their highest level, that the statistical association between the stock prices and conservative accounting values is likely to reach its lowest levels. Furthermore, the stock price increases were especially large for IT and hightech stocks during this period. As such, the combination of the extreme levels of share price for non-traditional firms and the fact that these industries are relatively more influenced by conservative accounting rules, may be the cause of their low value relevance in these years.
In accordance with the third hypothesis the value relevance of accounting information varies more over time in the non-traditional industries. To understand the variations better we display them in Figure 1. This figure shows that value relevance not only varies more for nontraditional industries, but also that it moves in a cyclical pattern around the, reasonably stable, value relevance of the traditional industries.
Next, we assess the extent to which these variations in value relevance are determined by the economic conditions and equity market sentiments. We use three crude measures based on annual information: the stock market return, the growth of the economy, and the equity market’s valuation. The equity market return captures primarily market sentiments, but also expectations of future economic growth. We measure it as the 12-month change in the AFGX, an index based on the largest firms at the Stockholm Stock Exchange. The economic growth disregards the stock market’s expectations and focuses purely on the state of the economy.
We measure it as the annual change in GDP-per-capita. The stock market’s valuation captures both expectations about the future economic growth and market sentiments. We measure it as the average equal-weighted market-to-book ratio for all firms listed at the Stockholm Stock Exchange. Table 8 provides correlation coefficients for the three measures and as expected all
0,90 0,70 0,50 0,30 Non-traditional industries Traditional industries 0,10 Figure description Figure 1 is based on an analysis of the value relevance over time for a sample of firms listed at the Swedish Stock Exchange
in the years 1983 to 2004 using a Price model specification with an adjustment for negative earnings:
Pit = a0 + a1 BVS it + a 2 EPS it + a3 EPS it ⋅ D + ε it where D = 1 when EPS 0, otherwise 0 where Pit is the share price of firm i in period t, BVS is the book value per share, and EPS is the net earnings per share. See Table 4 for details on the model specification. Figure 1 shows the explanatory power (adjusted R2) for the traditional and non-traditional industry categories as outlined in Panels B and C of Table 4.
measures are correlated. We find the highest correlation between stock market valuation and the economic growth. As shown the measures are correlated to each other, but they are far from perfect substitutes of each other.
To test the third hypothesis we rank all the years based on each variable and sort out the ten years with the highest/lowest values for each variable. Table 8 displays findings for the price and return model specifications with an adjustment for negative earnings.14 For the price model specification we expect that there is a greater difference in value relevance of nontraditional industries between strong and weak years. We expect that there is a negative Untabulated results show qualitatively similar results when applying price and return models without an adjustment for negative earnings, as well as when applying models with adjustment for sustainable earnings.
and strong market sentiments. We also expect book values of equity to be particularly affected by stock market sentiments and therefore that the price model specification is more affected.
Panel A presents the value relevance’s association to stock market performance. Tests based on the price model specification confirm our expectations. In good times the average difference is 9 percentage points (in favour of the traditional industries) whereas there is no difference in bad times. We also note that there is no difference at all in value relevance between high- and low-return years within the traditional industries. Thus the difference in relative value relevance between traditional and non-traditional industries is driven solely by inter-temporal variations in the non-traditional industry category. Panel B displays results of an analysis based on economic growth. Again, the price model specification confirms our expectations. In good times the average difference is 14 percentage points higher for the traditional industries whereas the non-traditional industries have a 5 percentage point higher relevance in bad years. The inter-temporal variation is almost completely due to the nontraditional industries.
Panel C displays the same analysis based on stock market valuation. When firms have high values relative to fundamentals the value relevance is 14 percentage points in favour of the traditional industries, whereas it is 5 percentage points in favour of the non-traditional industries when firms are valued low relative to fundamentals. As in the previous panels we see no variation within the traditional industries, but all changes in the relative value relevance is caused by inter-temporal variations within the non-traditional industries. To summarize, the three measures of economic conditions show differences in the relative value relevance of 9, 19 and 19 percentage points in support of hypotheses 3a and 3b.