Project Topics
Get Complete Project Material File(s) Now! »
Method
Research process
In order to conduct the research, prior scholars (Jennings et al., 2001, Mehrani and Mehrani 2003, Hamberg and Beisland 2014, Riedl 2004) has helped laid the foundation of the method. The mentioned authors have stated a framework for which variables and proxies might be useful in a quantitative approach of investigating goodwill and its effects. That framework has been used by the thesis to define which data to collect and how to improve the hypothesis.
In cases where firms included in the sample presents their financial reports in other currencies than SEK (EUR) the thesis has converted this to SEK for comparability reasons. The thesis has used the currency exchange rate provided by Riksbanken for every year. The thesis used the average exchange rate in each year.
The analysis will be conducted as follows. The sample will be run through the regression model where the sample is divided into two periods, one for the amortization regime (AMORT) and one for the impairment regime (IMP). The allocation of the data to each period is done by using dummy variables.
The multiple regression will be run on the sample which contains of 152 individual firm-year observations for 19 firms. After the regression is run the result will be analysed. The p-values will be examined to see whether or not any statistical significance exists on each independent variable.
Data collection and Sample
Table 1 show how the sampling of the thesis has been conducted. The Nasdaq OMXS30 index which the thesis examines, contains 30 firms for which the thesis collects 8 years each. The period for which the thesis collects data is from 1999-20091. This summarizes to a total 240 firm-year observations. The thesis collects data manually from the annual report provided. The thesis excludes firms that do not carry any goodwill in the balance sheet, hence 16 firm-year observations are excluded. Further, the thesis removes 32 firm-year observations from years with incomplete data in terms of lack of financial data and share prices. Thirdly, 40 firm years are excluded due to not being traded on the public stock market for the complete period of 1999-2009. Companies that had their initial public offering later than the 1st January 1999 have been removed.
The final sample consists of 152 firm-year observations (which is 63,33% out of possible initial observations) from 19 firms (63,33%) during the period of 1999-2009.
The thesis selects the OMX30 index for the period 1999-2009 as the sample which contains a total of 30 companies. Companies that lack any of the data that follows below are excluded from the sample. In this case, 11 companies were excluded due to the lack of data. The sample were thus reduced to 19 companies in the end. For each company, the thesis selects the following data: total assets, total debt, earnings, goodwill balance and goodwill costs. In addition, the thesis collects share price data as well.
From each company four years were collected from the amortization and impairment regime respectively. Some firms implemented IFRS 3 in 2004 while some implemented IFRS 3 in 2005 which meant that the years differs for the different periods between the firms. For the firms that adopted IFRS 3 in 2004, the amortization period is 2000-2003 and the impairment period is 2005-2008. For the firms that adopted IFRS 3 2005, the amortization period is 2001-2004 and the impairment period is 2006-2009.
Hypothesis
To be able to answer the research question, the thesis has developed a hypothesis.
Similar to Hamberg et al. (2014) the thesis examines the effect of goodwill costs since previous authors mentioned found goodwill costs to be statistically significant. The thesis explains the term associated as if there exists a significant statistical correlation between the variables, in this case if there exist a significant correlation between goodwill costs and a firm’s share price.
Since Hamberg et al. (2014) found goodwill costs to be significant in their research in a similar setting, it will be examined if this statement is applicable to this setting as well. The difference in their sample is that it investigates all firms in OMX Nasdaq for the period 2001-2010, while the sample of this thesis is OMXS30 during the period of 1999-2009.
The thesis examines the association between share prices but adds the perspective if there is a difference after the implementation of IFRS 3 to be able to see if the implementation of the impairment regime made a difference in the reaction of the investors to the treatment of goodwill.
Model
In the regression model, SPRICE is the dependent variable and there are nine different independent variables to test which variables explain the variation in share price. The regression has data that has a span from 1999-2009. As such, the regression for the thesis is as follows.
Similar to Hamberg et al. (2014) the thesis examines the effect of goodwill costs since previous authors mentioned found goodwill costs to be statistically significant. The thesis explains the term associated as if there exists a significant statistical correlation between the variables, in this case if there exist a significant correlation between goodwill costs and a firm’s share price.
Since Hamberg et al. (2014) found goodwill costs to be significant in their research in a similar setting, it will be examined if this statement is applicable to this setting as well. The difference in their sample is that it investigates all firms in OMX Nasdaq for the period 2001-2010, while the sample of this thesis is OMXS30 during the period of 1999-2009.
The thesis examines the association between share prices but adds the perspective if there is a difference after the implementation of IFRS 3 to be able to see if the implementation of the impairment regime made a difference in the reaction of the investors to the treatment of goodwill.
Variables
The thesis uses the following variables in the regression model:
In the model, SPRICE is used as the dependent variable. SPRICE is the value of the share the third month after the end of the company’s fiscal year. The third month is chosen to make sure that the financial statement has been published before the share estimation, in line with Jennings et al. (2001).
The variable total assets represent the firm’s size in terms of total assets. It has been studied previously that size has an effect on a firm’s share price. Zaheri and Barkhordary (2010), found that total assets have a positive relationship with share price, i.e. the bigger the firm is the higher the firms share price. Therefore, total assets are included in the model, to see if the amount of total assets has an effect on the share price.
Total liabilities are also included in the model. When the association between leverage (debt in relation to total assets) and share return has been tested in the past, there have been mixed results. Ho, Strange and Piesse (2008) and Zaheri and Barkhordary (2015) found no evidence of a statistically significant correlation between leverage and share price while Bhandari (1988) found a positive correlation between share price and leverage. Since there are conflicting evidence, total liabilities are added to the model instead of leverage to test if that variable has more relation to the share price.
ROE is a performance-based variable. ROE is included since previous research, e.g. Mehrani and Mehrani (2003), have found that the variable has a significant relationship with share price. ROE measures the firm’s use of its equity and how it turns the equity that it has into returns.
Balance of the goodwill post, goodwill costs and the change between each year of each of the two variables are added to the regression in accordance with the research provided from Hamberg (2014) since they were statistically correlated with the share price. Goodwill costs were divided into two different variables, goodwill impairment and goodwill amortization, but in this thesis, they are merged in to one variable – goodwill cost.
The earnings of the firms, and the change in earnings are added as two separate independent variables. Earnings and change in earnings have a history of being included in return models in prior research such as Easton and Harris (1991) and Lev and Zarowin (1999). Both variables have been proven in similar return models to have an impact on share price which is shown by Easton and Harris (1991).
The thesis expects firms reported earnings to increase in the impairment regime since the need for annual amortization has been removed. Hence, the expectation of future cash flow should be higher for a firm which carries goodwill in their balance sheet after the adoption of IFRS 3 since the removal of amortizations enable the firm to make a choice to make an impairment or not, which could lower the costs of the firm. This might affect the value of equity in the market. The history of amortizations of goodwill had no correlation with the performance of the firm, which is why this accounting information about goodwill under the amortization regime was ignored (Jennings, Robinson, Thompson and Duvall, 1996).
The price of a share may vary a lot and there are many dependent variables. Jennings et al. (1996) explains the treatment of goodwill as one of these dependent variables. The thesis expects to gain results explaining that there is an association between how the goodwill costs are treated and the share price, hence that impairments do affect the share price. The thesis argues that goodwill costs under an amortization regime does not provide much information to investors, instead this simply adds more noise in an annual report. Due to previous research which the thesis refers to in the literature review the thesis expects that there will be a correlation between the goodwill cost and the share price as well (Hamberg et al., 2011).
Results
Descriptive statistics
The table below presents the descriptive statistics for the thesis. The mean value of goodwill for the sample is 11 155 million SEK. The maximum value of goodwill is 60 102 million SEK which shows a significant difference between the firms. The mean cost of goodwill is 728 million SEK, where the largest goodwill cost equals 9 795 million which was recognized by Stora Enso in 2002.
There is a big variation in size of total assets within the sample. The average firm reports total assets of 498 851 million SEK while the firm with the most assets recognized in their balance sheet reports a total of 5 390 777 million SEK. There is a variance in size in the thesis because of the inclusion of Swedish banks. The four banks which is included in the sample is significantly larger than the average firm. The banks are the biggest four firms in the sample. The firm with largest size in terms of assets, except the banks, is Nokia in 2008 which recorded a total of 380 billion SEK.
Pearson correlation
Table 4 presents the Pearson correlation coefficients for the variables of the regression for both periods. The Pearson correlation coefficient measures the association between to variables in a linear regression and the value of the coefficient can be from -1 to 0 and 0 to 1 (Oxford Reference, n.d.). The Pearson correlation table is included for transparency reasons since there is a large correlation between the two variables TA and TL, 0,99 Pearson correlation for both regressions. Thus, TL is excluded from the final regression to avoid any problem with multicollinearity.
The thesis also recognizes the correlation between other variables and its close relation to the change of the same variable. As an example, there is a correlation that is significant at the 0,05 level between GW and ∆GW. This concern is identified with GWC and ∆GWC and EARN and ∆EARN as well. Since all these variables are closely related in nature, e.g. ∆GW is based on GW, the subsequent correlation between the variables is expected. This could result in a problem with multicollinearity. However, the thesis still argues that all variables except TL are important to test against the dependent variable and that the multicollinearity is not as big as for TA and TL.
Linear regression
Table 5 displays the results from the linear regression. The model had an explanatory power of 8,8% for the amortization regime and 14,6% for the impairment regime. That means that the independent variables in the regression explains 8,8% and 14,6% of the variation in the dependent variable.
In the amortization regime, the only variable of the ones tested against share price that has a statistically significant correlation with the dependent variable is EARN. EARN has a positive correlation with share price which means that an increase in the earnings of the company increases the share price as well. The rest of the variables does not have a statistically significant association with the share price, including the variable goodwill cost which represent the amortization cost of goodwill.
For the impairment regime, there are several variables that do have a statistically significant correlation with share price. Variables GW, ∆GW and GWC are all statistically significant at the 10%-level, with both GW and ∆GW also being statistically significant at the 5%-level. The correlation with share price is positive for ∆GW and GWC while GW has a negative correlation with share price.
The comparison between the two samples shows that there is a difference between them. The three variables (GW, ∆GW and GWC) that do have a statistical significance for the impairment regime are not statistically significant for the amortization regime while the only variable (EARN) that is statistically significant for the amortization regime is not statistically significant in the impairment regime.
Table 5 displays the results from the linear regression. The model had an explanatory power of 8,8% for the amortization regime and 14,6% for the impairment regime. That means that the independent variables in the regression explains 8,8% and 14,6% of the variation in the dependent variable. In the amortization regime, the only variable of the ones tested against share price that has a statistically significant correlation with the dependent variable is EARN. EARN has a positive correlation with share price which means that an increase in the earnings of the company increases the share price as well. The rest of the variables does not have a statistically significant association with the share price, including the variable goodwill cost which represent the amortization cost of goodwill.
For the impairment regime, there are several variables that do have a statistically significant correlation with share price. Variables GW, ∆GW and GWC are all statistically significant at the 10%-level, with both GW and ∆GW also being statistically significant at the 5%-level. The correlation with share price is positive for ∆GW and GWC while GW has a negative correlation with share price.
The comparison between the two samples shows that there is a difference between them. The three variables (GW, ∆GW and GWC) that do have a statistical significance for the impairment regime are not statistically significant for the amortization regime while the only variable (EARN) that is statistically significant for the amortization regime is not statistically significant in the impairment regime.
The results of the regression do as is mentioned above reveal the statistically significant relationship between GW, GWC and ∆GW in the impairment regime. However, the sign of the correlation can seem to be counter-intuitive. GW has a negative correlation with share price and GWC has a positive correlation with share price. The correlation would therefore mean that when the size of the goodwill post increases, the share price is decreasing and when there is an increase in the impairment cost the share price is increasing. This is counter-intuitive since if the goodwill post is increasing the profit/loss-account increases, while an increase of the impairment cost decreases the profit/loss-account. It is difficult to determine a specific reason for this result, but a possible explanation is that investors are of the opinion that a bigger goodwill post is a result of earnings management and manipulation of the result, as has been argued by AbuGhazaleh et al. (2011). If managers do not record an impairment even though it would be reasonable to do so, the goodwill post will increase, and the goodwill cost will not increase. But if they instead record an impairment, investors could be of the opinion that the firm is revealing numbers that reflects reality better.
As well as GW and ∆GW, EARN is significant in one period but not significant in the other period. In contrast to the other variables however, EARN is significant in the amortization regime and not significant in the impairment regime. The statistically significant correlation between EARN and share price during the amortization regime is positive, thus meaning that an increase in EARN leads to an increase in the share price. The variable is however not significant in the impairment regime. As such, in the amortization regime goodwill was not affecting share prices but the earnings were the variable from this set of variables that were affecting the investment decisions.
Table of content
1.Introduction
1.1Background
1.2Problem
1.3Purpose
1.4 Definitions
2.Frame of reference
2.1Accounting standards
2.2Literature review
2.3Theories
3Method
3.1Research process
3.2Data collection and Sample
3.3 Hypothesis
3.4Model
3.5Variables
4Results
4.1 Descriptive statistics
4.2 Pearson correlation
4.3 Linear regression
5.Analysis
6.Conclusion
7.Discussion
7.1Limitations
7.2Future research
7.3Social and ethical implications
8.Reference list
GET THE COMPLETE PROJECT
The Association Between Goodwill Costs and Share Prices
