this was a group assessment, and we got 20 out of 30. the main problem was the purpose of our study was unclear, so the marks were just so-so
Abstract
The Ohlson (1995) and Feltham and Ohlson (1995) papers are landmark works in financial accounting. But when we employ Ohlson model for Australian market, it shows that the book value of equity is not as significant as we expected. So, we develop our new models. We decompose book value of equity per share into assets per share and liabilities per share to explore whether the disaggregated accounting items can provide better value relevance to share price. And then we decompose assets and liabilities into current/ non-current assets and liabilities. Including Ohlson model, we totally employ three models in our paper. Finally, after conduct regression test, we found that using current assets per share, non-current assets per share, current liabilities per share, non-current liabilities per share and earnings per share can provide more and better information of value relevance.
Introduction
Prior research has documented how disaggregated accounting data explain return (Ohlson & Penman, 1995). Ohlson presented a formulation derived from the classical conceptions that used accounts variables in the function of evaluation. The structuring was baptized Ohlson Model (MO) and had great impact in the academic research about capital markets. Also, Kothari and Zimmerman (1995) states that price models is better than return as price models have arguably provided better estimates for the coefficient and the profit figure than the returns model. Price models have also been found to be intuitively appealing (Rees, 1999) as they suggest that the current market price reflects the future estimated earnings power of the firm. Barth (1992) examines whether pension cost components result in an increase in explanatory power and found that the disaggregation of earnings does provide additional information to investors.
In this report, we first use the Ohlson model to check whether the book value of equity and earnings are good indicators of share price for the Australian market. Then, we hold earning unchanged, use total assets and total liabilities to substitute the book value of equity. After that we further decompose the total assets into current assets and non-current assets, and the total liabilities into current liabilities and non-current liabilities to examine whether these two developed model can provide more and better information of value relevance.
Models and Hypotheses
Models
The following version of the valuation model is the simplified Ohlson model which is used and consistent with which employed by Collins et al. (1997) for a US sample, Harris et al. (1994) using a German sample, Joos and Lang (1994) using German, French and British samples and Rees (1997) using a British sample.
Pit=a0+a1BVit+a2Eit+eit (1)
Where Pit is the year end price per share, BVit is the book value of equity per share and Eit is the earnings per share, for firm i year t. Generally, firms with higher book value of equity and earnings per share will have relatively higher share prices. Also, refer to the results from previous study (Giner & Rees, 1999), we expect there is a positive relation between the share price and the two variables. In other words, we expect that a1 and a2 are positive.
Previous studies have adjusted the simplified Ohlson model to examine whether the disaggregation of earnings or disaggregation of balance sheet items impacts on the valuation model. Lipe (1996) investigates whether gross profit and the breakdown of various expenses provides additional information to the model and finds that there is a statistically significant difference in the model compared with the consolidated model. Based on the result of Ohlson model in our case which shows that book value of equity per share in our case is not as significant as we expected, we decompose the variable Eit into Ait and Lit, which Ait is the total assets per share and Lit is total liabilities per share, for firm i year t. So we develop our new model:
Pit=a0 +a1Ait –a2Lit+a3Eit + eit.. (2)
Since assets and liabilities together compose the equity, we aim to use this model to test whether disaggregated equity provides additional information.
For our developed model, because generally when the company has more assets, it will likely have higher share price. Also, the ratio of asset-to-liability is usually at a particular level for big company, when the company has more assets, it usually means the company has more liabilities. So we expect there will be a positive relation between liabilities/assets per share and share price.
Furthermore, we decompose the Ait into CAit and NCAit and decompose Lit into CLit and NCLit. CAit means current asset per share, NCAit means non-current asset per share, CLit equal to current liabilities per share, and NCLit equal to non- current liabilities per share, for firm i year t. So, we have Model (3), that is
Pit=a0 +a1CAit +a2NCAit–a3CLit – a4NCLit +a3Eit + eit (3)
Model (3) is the further exploration of Model (2). Similarly, we also expect there will be a positive relation between the share price and our variables in Model (3). By using these three models, we want to find out which model can explain the relationship between share price and variables better.
Hypotheses
As mentioned above, there are three models used during our research, correspondingly, there are three hypotheses.
As the previous research conducted by Giner and Rees (1999), they found that the three basic variables, market price, book value per share and earnings per share are high correlated in Spanish. But according to the research of Lopes (2002), it states that earnings provide low value relevance when compared with book values in Brazil. But compared Spanish with Brazil market, we believe that, Australian market has more common with Spanish market. Thus, our first hypothesise is that:
H1 Market price, book value of equity per share and earnings per share are high correlated in Australia.
According to the development of our new models, we further hypothesise that:
H2 : The disaggregation of book value of equity provides additional information to the model.
H3: The disaggregation of assets and liabilities into current and non-current provides additional information to the model.
Data
Sample selection
We selected 50 companies which ranked in the top 100 listed Australian Companies colleting their data from Aspect Fin Analysis. The reason why we choose Australian top 100 is that it includes various industries, such as banks, financial services firms, retailers, mining and so on, which enable our results to be more explanatory. Since our developed model is based on current and non-current financial items (e.g. current asset and non-current asset), we first exclude all banks from our data, as it is hard to distinguish current and non-current financial items of a bank. Then, considering our relatively small sample size, to prevent our sample from any industry level bias, we randomly select 50 companies from the remaining top 100 companies to form our sample.
All of our data collected are base on the financial information of 2008, because, currently, some companies’ financial information of 2009 is still not available. Also, on 3rd July 2002, FRC (Financial Reporting Council) announced that Australia should adopt international accounting standards by 1st January 2005, and AASB changed a lot in 2007. So the accounting data of 2008 is more reliable.
To insure our data collected against errors, we don’t only collect the data which we need, but also collect relevant data to test the accuracy of our data. For example, we collect total equity for Ohlson model, also, we collect total assets and total liabilities, use their difference to double check the value of equity. Same approach also being used to insure the accuracy of other variables.
Another thing need to mention is that our input data of liabilities are all negative considering their nature.
Sample characteristic
We select 50 companies from the top 100 companies of Australia. And from Model 1 to Model 3, there are total nine variables involved. We select total asset, total liabilities, total equity, current assets, non-current assets, current liabilities, and non-current liabilities from each company, and using these accounting number divided by number of shares (weighted average number of shares). After that, we conduct the descriptive statistics test, the results are shown as followed:
These tables provide the descriptive statistics concerning the variables used. The data exhibits a considerable amount of skewness and kurtosis, as is normal for cross-sectional valuation models but there is no evidence that this affects the reliability of the results. Because all the companies we selected are from the top 100 companies in Australia, we are not surprising all of these figures are high. (During our research, we compared our results to other groups, we found our descriptive statistic results were a little higher than others.) We also noticed that, the range of “share price” and “asset per share” are very high, they are 106.63 and 100.54, and the minimum figures in these two columns are 0.24 and 0.245. These two figures are from Centro Properties Group and Publishing & Broadcasting. In Centro Properties Group’s annual report, it shows its financial situation has been changing dramatically, for example in 2004, its reported total assets are 488.9 million, but in 2005 it increase to 6564.6 million. Similarly, in 2007 its total assets are 8165.1 million, but in 2008, it reaches to 20576.1 million. Within 5 years, its total assets increase 42 times. We thought this company may influence our results, but we did another descriptive test which excludes Centro Properties Group and Publishing & Broadcasting, the results are similar with the original test. (The mean of share price is 16.5118 and the mean for assets per share is 15.94). So, we still believe that our data is reliable.
Results & Discussion
There are three models involved in our research. And for each model there are two levels of analysis are conducted. Initially, we conduct correlation test to explore the relationship among our variables. Then, we used regression to test the performance of the three models. In the rest of the section, we will discuss about these models one by one.
Ohlson Model (Model 1)
Firstly, we conduct the correlation test, the results are:
Table 2
Secondly, we conduct the regression test. The results are as followed:
The table above provides the regression analysis concerning the relationship between dependent variables and independent variables used. The positive coefficients (0.46 and 5.49) confirm the positive relation between share price and book value of equity per share (BVE), and same conclusion for share price and earnings per share (eps n). Adjusted R2 reports the explanatory power of the equity and earnings model. A 0.789 adjusted R2 shows a very high explanatory power of this model. A 2.67*10-12 P-value of earnings per share also shows that the result is very significant at the 0.05 level. However, the P-value of equity per share is 0.09, which indicates that equity per share is still significant at the 0.1 level, but not significant at the 0.05 level. Thus, we ran another test. This time we excluded the earnings per share, only analyse the relation between the share price and book value of equity per share. The result shows substantial erosion in the value of adjusted R2 which is 0.41 compared with 0.79, it also parallels the insignificant result of equity per share.
The results of Ohlson model which shows that book value of equity per share is actually not significant at the 0.05 level as we expected, thus we conduct the same tests for our new model: Pit=a0 +a1Ait –a2Lit+a3Eit + eit.. Since asset and liability together compose the equity, we aim to use this model to test whether disaggregated equity provides further information.
Model 2
We also did the correlation test at first. The results are:
For our Model (2), the absolute value of correlation is almost the same with Model (1). And the correlations of these two variables with share price are 0.67 and -0.62. The number here, is -0.62, is a negative number, as we expected, because our input data of liabilities are all negative. So, in fact, the relation between share price and liabilities per share is actually positive. And from the earnings per share (eps) perspective, same as that in Model (1), also shows high correlation with share price, (0.89 in the table). So model (2) shows there is a high positive correlation between asset per share, liabilities per share, earnings per share and share price.
Next table shows the result for regression test for Model 2
The positive coefficients confirm with our expectation which is the positive relation between assets per share, liabilities per share and share price. The value of adjusted R2 is 0.787 which is nearly the same of Ohlson model. That indicates that explanatory power of this model is also high. The P-value of earnings per share is 5.12E-11, still significant. However, the P-values of assets and liabilities are 0.06 and 0.08, still not significant at the 0.05 level. But, compared with the P-value for book value of equity (0.09) before disaggregation, the results we have in this table get a little improved.
Next, we further develop our model 3: Pit=a0 +a1CAit +a2NCAit–a3CLit – a4NCLit +a3Eit + eit. We disaggregate asset into current asset and non-current asset, and disaggregate liability into current liability and non-current liability. Still, we aim to use this model to test whether further disaggregated equity provides further information
Model 3
As the first two models, it also presents the results correlation test first.
This time our results change a little compared with our first two models. The correlation in Model (3) are 0.30 for current assets per share (ca/s) and share price; 0.73 for non-current assets per share (nca/s) and share price; -0.58 for current liabilities per share (cl/s) and share price; and -0.53 for non-current per share (ncl/s) and share price. From the assets perspective, the correlation between non-current assets per share and share price is higher than that with current assets. Then we calculate the proportion of current asset in total asset, the result shows that, the average is 30%, which means non-current asset makes up 40% more than current asset in total asset. We also notice that, all the insurance companies have more current asset than non-current, because the characteristics of insurance industry. For example, 97% of AMP’s total assets are current asset, and for IAG is 71%. That is to say, in the other industries, the current asset takes less than 30% of its total asset. Non-current asset makes up most part of total assets, that can explain why the correlation between current assets per share is less than non-current asset per share. From the liabilities perspective, we can see that the correlation for share price and current liabilities per share (cl/s) is -0.58, similar with share price and non-current liabilities (ncl/s) which is -0.53. These two figures are almost the same, because current and non-current liabilities make up almost same proportion of total liabilities. We find that, on average, there are 40% of companies’ liabilities are current liabilities. But when we eliminate some extreme samples, (for example, ASX’s 97% liabilities is current liabilities, and Fortescue’s 8% liabilities is current liabilities.) the proportion of both current and non-current liabilities are almost the same.
The correlation between current or non-current liabilities per share and share price shows there is a positive relation between current or non-current liabilities per share and share price. As explained in our Model (2), the values of the correlations for current/ non-current liabilities are negative, but in fact, they have positive relation with share price. To sum up, share price, current asset per share, non-current asset per share, current asset per share, non-current liabilities per share and earnings per share are positively correlated in Australia.
The results for regression test are listed below:
Robustness Test
To perform a test of the robustness of our results, we rerun the correlation analysis and regression analysis for each of our three models, using the share price at 90 days after the financial report issued to replace our original share price data. The tenor of the results nearly remains the same for all three models. The correlation analysis in the robustness test parallels the results in our original tests. In addition, the adjusted R2 for Ohlson model and our first developed model are both approximate 0.59 which is still high. The adjusted R2 for our second developed model is 0.65 which is comparatively higher than 0.59. This also confirms our expectation which is our second model provides better value relevance in explaining the share price. However, we can see that there is a little erosion in the adjusted R2. We are not surprising with this result. Given the financial crisis all around the world, nearly every single company’s share price was going through a hard pressure, especially in the year 2008. Consequently, the erosion in the adjusted R2 is considered reasonable and acceptable. In other words, our results are still reliable.
Conclusions
This study suggests using disaggregated accounting item can provide more and better information of value relevance. During our study, we first conduct descriptive statistic test to check the characteristics of our samples. The results show that although the range, kewness and kurtosis are little high, our data is still reliable. Then we conduct the correlation test. And correlation results show there are only some minor differences among our three models. But we find more differences when we conduct the regression test. We find that the value of adjusted R2 increased a little from Ohlson model (Model 1) to Model 3. So the explanatory power increased a little when we decomposed our variables. (We first decomposed book value of equity per share into assets and liabilities per share, so generate our Model 2. Then we decomposed assets and liabilities by their characteristic- current and non-current to get our Model 3) But the P-value of our sampling change essentially. In Ohlson model, the P- value of book value of equity per share is over 0.05 which indicates that this variable is not significant at the 0.05 level. But in our Model 3, all of our variables are significant at the 0.1 level, four of our five variables are significant at the 0.05 level, and the one left is 0.051 which is very close to 0.05. So, the Model 3 can provide more and better information of value relevance.
Limitations and further research
Our research takes only 1 month, the time is too short, so there are some shortcomings in our research.
Firstly, the sample size is too small. We only collect 50 companies’ data for 2008. Compared with the previous research which related to this topic, they normally collect hundreds of samples for their studies. For example, Ohlson collected 391 companies’ accounting data from SP500. Lynn at el (2006), their final sample consists of 1628 firm year observations.
Secondly, the entire sample comes from top 100 companies in Australia. Weather the sample can demonstrate the real situation in Australia is questionable. We should notice that there are only a few industries involved in the top 100 companies, energy, financial, insurance, material and mining take up a large proportion in these companies. But there are lots of other industries in Australia, the industries included in our sample may not reflect the reality. Also, there are seven banks in the top 50 companies, but when we select our sample, we exclude all of them. It is because the unique characteristic of bank industry, in the bank’s financial reports, we can not classify their current and non- current assets or liabilities in their financial reports. Because that, our results may biased and noisy.
To improve our research, we should choose more companies of different industries in order to enlarge our research sample. Also, it is important to use other variables for further research if suitable data are available. According to the paper of Barth and Clinch (1998) - Revalued Financial, Tangible, and Intangible Assets: Associations with Share Prices and Non-Market-Based Value Estimates, we can use the tangible assets and intangible assets instead of the current assets and non-current assets in the future to find out which variables are more relevant to share price.
Reference
Barth, M. E., and G. Clinch. (Spring 1996): "International Differences in Accounting Standards: Evi- dence from U.K., Australian, and Canadian Firms." Contemporary Accounting Research 135-70.
Barth, M. E. Beaver, W and Landsman W. (March 1992): "The Market Valuation Implications of Net Periodic Pension Cost." Journal of Accounting and Economics 27-62.
Collins, D., E. Maydew and I. Weiss: 1997, “Changes in the Value-Relevance of Earnings and Book Values over the Past Forty Years”, Journal of Accounting and Economics 24: 39–67.
Giner, B.: 1997, “The Influence of Company Characteristics and Accounting Regulation on Information Disclosed by Spanish Firms”, The European Accounting Review 6(1): 45–68.
Giner, B.: 1999, “A Valuation Based Analysis of the Spanish
Accounting Reforms”, Journal of Management and Governance 3: 31–48, 1999.
Harris, T., M. Lang and H. Moller: 1994, “The Value Relevance of German Accounting Measures: An Empirical Analysis”, Journal of Accounting Research 32(2): 187–209.
Joos, P. and M. Lang: 1994, “The Effects of Accounting Diversity: Evidence from the European Union”, Journal of Accounting Research 32(Supp): 141–168.
Kothari, S and J. Zimmerman: 1995, “Price and Return Models”, Journal of Accounting and
Economics 20: 155–192
La Porta, R., Lopes-de-Silanez F. Shleifer, A. Government Ownership of Banks. Journal of
Finance, Vol. 57, No1, 2002
Stephen L. (2006), Incremental Value Relevance of Unrecognized Deferred Taxes: Evidence from the United Kingdom.
Ohlson, J.: 1995, “Earnings, Book Values and Dividends in Equity Valuation”, Contemporary Accounting Research (Spring) 11(2): 661–687.
Posts
No comments:
Post a Comment