Amsterdam Business School
Internal investments:
Advertising expenditures and expectations on future company
growth
Name: Walter Zeegers
Student number: 9905847
Date: 21st of July 2014
First supervisor: prof.dr. L.R.T. van der Goot Second supervisor: dr. ir. S.P. van Triest
MSc Accountancy & Control, specialization Control
Abstract
This thesis is an attempt to acquire a deeper understanding of the causal relationship between the amount of spending on advertising, relative to a firm’s net sales, and the market’s expectation for future growth of a firm. As a proxy for the expectation of growth of a corporation, the market-to-book ratios are selected. Furthermore, the influence of a firm’s size on the aforementioned relationship is investigated. It turns out that indeed advertising expenses are significant and positively related to market-to-book ratios. On the other hand, no evidence is found that supports the hypothesis that large companies are more efficient in their advertising efforts and that large firms have higher market-to-book ratios than smaller firms.
Table of contents 1 Introduction ... 4 1.1 Background... 4 1.2 Research Question ... 5 1.3 Contribution ... 5 2 Theory ... 6 2.1 Literature Review ... 6 2.2 Hypotheses ... 9
3 Data and Methodology ... 10
3.1 Data and sample selection ... 10
3.2 Regression model and estimation method ... 15
3.3 Correlation matrix ... 18
4 Results ... 20
5 Conclusions and Discussion ... 31
1 Introduction
1.1 Background
The purpose of advertising expenditures is often to increase a company’s sales in the same period as when the costs are incurred, but, in addition, these expenditures also represent some sort of an investment that gives rise to future economic benefits, lasting more than one year. This implies that spending on advertising in a current period could increase expectations for a company’s future financial performance. This spending could possibly enhance the value of intangible assets, for instance, the value of the brands a company owns. As a consequence, this may lead to higher share prices of the firm.
The shareholder value principle advocates that a business should be run to maximize the return on shareholders’ investment (Joshi et al., 2010). Nowadays, a great portion of market values of companies is formed by the market’s appreciation of intangible assets. To illustrate this, market-to-book ratios, ratios that are calculated by dividing market values of companies by their book values, were around 5 on average for S&P 500 companies in 2001 (Lev, 2001, page 9). As it is the management’s implicit obligation to maximize shareholder value, they should understand how the market appreciates expenses that affect the perceived value of intangible assets. An
understanding of the market’s responses to chosen spending levels on intangible assets, for instance on advertising and promotion, could help management to maximize shareholder value. In 2001 Baruch Lev developed a framework in which intangible assets are classified into the following four groups (Lev, Radhakrishan and Zhang, 2009):
1. Discovery/learning intangibles—technology, know-how, patents and other assets emanating from the discovery (R&D) and learning (e.g., reverse engineering) processes of business
enterprises, universities and national laboratories.
2. Customer related intangibles—brands, trademarks and unique distribution channels (e.g., internet-based sales), which create abnormal (above cost of capital) earnings.
3. Human-resource intangibles—specific human resource practices such as training and compensation systems, which enhance employee productivity and reduce turnover.
generating sustainable competitive advantages.
This framework classifies brands as part of the customer related intangibles. Amongst other factors, brand value itself is affected by a company’s current decisions on the level of spending on advertising. This research aims to investigate the relation between the level of spending on advertising and the market value of companies.
1.2 Research Question
The central research question in this thesis is: Does a relation exist between current spending levels on advertising and expectations for future financial performances of corporations? More specifically formulated, does a relation exist between relative advertising expenses (advertising spending, relative to companies’ net sales) and market-to-book ratios of corporations?
1.3 Contribution
This paper provides a deeper understanding of the relation between advertising expenditures and companies’ market values (or the relation between the level of advertising expenditures and the market-to-book ratios of companies). Therefore, this paper could provide managers of companies some theoretical background to base their practical spending and investment decisions on. From a scientific perspective, this paper builds on prior research. However, where prior research tried to capture a more comprehensive model, with many variables explaining market-to-book ratios or alternatives like Tobin’s Q, this paper specifically focuses on one independent variable, the level of spending on advertising.
2 Theory
2.1 Literature Review
Under most commonly applied accounting standards, IAS 38 (issued by the International Accounting Standards Board) and US GAAP 340-20 (issued by the Financial Accounting Standards Board), advertising expenditures usually are expensed and not capitalized in a
company’s financial statements. This implies that, from an accountancy point of view, the book value of a company would not increase by spending on advertising. On the contrary, expensing leads to a decrease of the book value of a company’s assets. On the other hand, Chauvin and Hirschey (1993) found evidence suggesting that stock market investors evaluate advertising efforts of firms within a long-term perspective. This implies that the market value of a company could actually increase as a reaction to advertising spending decisions, because of the long-term perspective of investors. The combination of the two effects described above would result in an increase of the market-to-book ratio, since the numerator of this ratio (market value) would increase and the denominator (book value) would decrease.
Brand value is a substantial component of a firm’s market value: According to a 2010 estimate, market value of brands accounts for over 30% of the market capitalization of S&P 500 firms, and even exceeds the book value of equity of those firms (Larkin 2013). Larkin et al. (2013) established that a favorable perception of a brand is associated with lower cash flow volatility and lower bankruptcy risk. A strong brand generates a clientele of loyal consumers who have a high subjective value for the firm’s products and are willing to stick with them over time. As a result, firms with favorable brand perceptions should enjoy a more stable stream of future profits and lower riskiness (Larkin, 2013). This illustrates that investing in brand reputation and
spending on advertising indeed creates value for a company. The contemporary accounting rules do not include internally generated brand value in financial statements; however, the effect of investing in brand value is expected to be visible in market-to-book ratios, as firms’ market values are expected to anticipate on these types of value adding investments.
In his book “Intangible assets: management, measurement and reporting” (2001) Baruch Lev illustrates that the average market-to-book ratios of the top 500 largest US listed companies (S&P 500) increased from just above 1 in 1977 to over 5 in 2001. In other words, in 2001 the market values of companies were five times as high as the book values of companies, on average. This implies that in 2001 only about 20 per cent of corporate market value was reflected in financial statements. Lev and Zarawin (1999) argue that the current accounting system fails to
reflect enterprise value and performance, due to the mismatching of costs (for instance the costs for advertising) and the revenues it will eventually generate. They explicitly state that this is indeed the case for investments in intangible assets such as brands. These activities are usually immediately expensed, while the benefits are recorded later and therefore are not matched with the previously expensed investments (Lev and Zarawin, 1999). In addition, investments in spending on advertising could be viewed as a form of investment in intangible assets with predictably positive effects on future cash flows (Chauvin and Hirschey, 1993). It may be clear that the financial markets appreciate intangible assets like brands in a different way than that they are accounted for in a company’s financial statements. The question throughout this thesis is, to what extend does the market, in general, anticipate on the size of company spending on
advertising. How is the firm’s market value, relative to the book value, impacted by the amount spent on advertising?
The gold standard metric for assessing branding’s impact on the firm is shareholder value, which is determined by levels of stock returns and the volatility associated with those returns
(Srinivasan and Hanssens 2009). Commonly used metrics of shareholder value include stock returns, market capitalization, Tobin’s Q and the market-to-book ratio (Srinivasan, 2011). Market-to-book ratios are constructed by dividing the current share price by the book value per share. An alternative approach is to divide current total market capitalization by the total book value of the firm. Ratios larger than one, signal that firms will create value for its shareholders. In that sense, this ratio can be regarded as a forward-looking measure, providing investors the expectations of the firm’s future profits. As the market-to-book ratio includes all available information on expected future profits, it also includes the expectation of future benefits generated by current spending on advertising. Therefore the market-to-book ratio could be included as the dependent variable in our research, which is (for a part) influenced by the spending decisions on advertising.
In their paper Srinivasan and Hanssens (2009) prefer to use the Tobin’s Q over the market-to-book ratio, in assessing the impact of marketing on firm value. They state that the Tobin’s Q ratio avoids accounting complications, because the Tobin’s Q is constructed by dividing the market value of a firm by the replacement costs of its assets, instead of dividing the market value by a firm’s book value. In this paper the market-to-book ratio is used, because of practical implications. Data on market-to-book values are relatively easily obtainable, compared to the
principles.
Under the efficient markets model, stock prices at any point in time fully reflect all available information and provide an expectation of discounted future cash flows. For example, when announcements on advertising campaigns occur, investors will decide to buy or sell company shares, based on expectations of how the advertising campaign will affect future cash flows (Lane and Jacobson 1993). If investors appreciate such advertising expenditures in a positive way, the advertising efforts will result in increases of the stock price. As a consequence, advertising expenditures will change market-to-book ratios.
Several recent studies have suggested that a firm’s advertising efforts directly affect stock returns beyond the indirect effect of advertising through lifting sales revenues and profits (Srinivasan and Hanssens, 2009). The intangible equity that advertising attempts to create, ostensibly for customer marketing purposes, can spill over onto investors and increase a firm’s salience with individual investors, who typically prefer holding stocks that are well known or familiar to them. These findings help explain why several firms advertise at levels beyond those justified by sales response alone. Indeed, recent studies have confirmed that advertising expenditures create an intangible asset (Srinivasan and Hanssens, 2009).
Like current cash flow, growth, risk and market share, advertising and research and development expenditures constitute key determinants of the market value of the firm, according to Chauvin and Hirschey (1993). In their article, they made a distinction between manufacturing and non-manufacturing companies. They concluded that the market value effects of advertising are broadly operative throughout both manufacturing and non-manufacturing sectors. In this thesis we will not focus on the same distinction, but by making use of dummies for industries, industry specific impact could be tested and controlled. Dummies are selected for a few industries, for instance for high technology, pharmaceuticals and financial services. These industries are on forehand expected to have a different relation between relative advertising budgets and market-to-book ratios, compared to other industries. Whether this is indeed true, will be tested when analyzing the acquired data.
According to the research of Chauvin and Hirschey (1993), the market valuation effect of advertising (and R&D) spending is greater for larger companies. A dollar spent on advertising for larger firms tends to have a greater market value effect, than for smaller firms. Market value differences across smaller firms may simply be tied more closely to the fortunes of specific industries, whereas large diversified firms are more broadly positioned to take advantage of scale
advantages in advertising (Chauvin and Hirschey, 1993). Still evidence suggests that expenditures by smaller, specialized firms can be highly effective (Chauvin and Hirschey, 1993). These findings suggest that size advantages make advertising more profitable for larger firms. In order to connect to the findings of prior literature, this research will include the size of companies as a variable. In this paper the firm size is measured by the natural logarithm taken from the number of employees of the company, which is treated as an independent variable.
2.2 Hypotheses
As discussed before, under accounting standards IAS38 and GAAP340-20, advertising expenditures usually are expensed and not capitalized in a company’s records. This means that these costs immediately decrease a company’s assets in book value, for instance its cash positions. On the other hand, these “internal investments” usually do increase expectations for future benefits as well. Spending on advertising is expected to increase a company’s future sales, because of an improvement of brand reputation and an increase in brand value. Hence, these expenditures will increase a company’s market capitalization, and therefore will lead to an increase of the market-to-book ratio.
The central hypotheses that will be tested in this research are:
Hypothesis 1:
In general, higher company spending on advertising, relative to a company’s net sales, leads to higher market-to-book ratios.
According to prior literature, for instance research of Chauvin (1993), company size has a positive effect on the effectiveness of spending on advertising. Larger companies seem to be more efficient in reaching potential customers. Advertising activities of one division could support other divisions and products of the firm at the same time.
Hypothesis 2:
The size of a company, measured by the (logarithm of the) number of employees of a company has a positive effect on the effectiveness of advertising spending and hence leads to higher market-to-book ratios of companies.
3 Data and Methodology
3.1 Data and sample selection
To acquire the desired data, the website http://wrds-web.wharton.upenn.edu was used. This site functions as an interface for users to be able to obtain data from a number of databases. One of those databases is COMPUSTAT, the source of all the data used in this research.
When approaching the COMPUSTAT database this way, directly some selections and filters have to be chosen. For this dataset only active corporations listed on the New York Stock Exchange (NYSE) are selected. This selection is chosen, because the NYSE represents a liquid market with sufficient firms listed for this research. Only active firms are selected, because inactive firms could include special cases with exceptional circumstances like bankruptcies. These cases could have great deviations in market-to-book ratios and have a disturbing influence on the relation we want to investigate. Hence, the decision is made to leave inactive firms out of this scope. Furthermore, only companies with quotations in US dollars are selected, leaving out the data of firms listed in other currencies on the NYSE, for example quotations in Canadian dollars. The period of time for investigation is January 2003 until December 2013, eleven full years of observation. This procedure gives an initial dataset of 23,685 records of a total of 2,577 companies. The variables and data fields extracted from the database are:
-Company name; -Ticker symbol; -CUSIP code;
-Accounting Standard type; -Actual period end date; -Data Year (fiscal);
-Fiscal year-end month, number of the month;
-Book value per share (in USD), at the end of the accounting period;
-Common/Ordinary equity total (in millions USD), including reserves and retained earnings; -Common shares outstanding (in millions of shares), at the end of the accounting period; -Employees (in thousands), at the end of the accounting period;
-Sales/Turnover net (in millions USD), of the full fiscal year; -Advertising expense (in millions USD), of the full fiscal year;
-Total market value of the corporation at the end of the fiscal period (in millions USD); -Share price at closing date at the end of fiscal period (in USD);
The initial dataset requires some further adjustments, because some ratios have to be constructed from the attained variables. For instance, the market-to-book ratio is constructed out of two data fields. This ratio is obtained by dividing the share price by the book value per share, both measured at the final date of the fiscal period. This exact same timing of measurement of both variables is chosen, because advertising activities are visible throughout the year they are conducted. When financial statements and reports are publicized, usually around three months after the final day of the fiscal period, the amount spent and effort made on advertising should not be a complete surprise anymore.
The denominator of the market-to-book ratio is the book value per share. If this variable is equal to zero, the ratio could not be calculated. To avoid these errors, the initial dataset is adjusted. All records with no data, or value zero for the book value per share are left out of the final dataset. The same counts for records with no positive figure for net sales (turnover), since this variable is used to calculate the relative advertising expenses (advertising expenses relative to the net sales of a company). Other records that are left out of the final dataset are entries with value zero or negative values for share prices, for advertising expenses and for the number of employees. After these steps from the initial 23,685 records, 5,239 records remained suitable for analyzing.
However, a final check is made on the outcomes of calculated market-to-book values. As discussed before, from prior literature we know the average of market-to-book values was around 5 in 2001 for S&P 500 listed companies. A quick glance at the constructed book ratios indeed shows values around this number. For 23 of the 5,239 records the market-to-book ratios are above 50, ten times the aforementioned average of 5. These 23 special cases are treated as extreme outliers and left out of the final dataset. The final dataset includes a total of 5,216 records of 713 firms.
Data is collected from the period January 2003 until December 2013. In this research, we want to compare market-to-book ratios of individual firms, without (or with as less presence as possible of) bias from general, external economic developments. Appreciations of market values of firms differ from time to time, for some part as a reaction to changes in the perception of external general market conditions and worldwide economic development. Bias from changes in general economic conditions should be avoided. To make a correction for this effect, dummies are constructed for each separate fiscal year. Of course, expectations on economic development could change every day, or at an even faster rate. By constructing dummies for each fiscal year, this effect is at least taken into account to some extent.
The dummy variables for the fiscal years included in the final dataset are: Dummy 2003, Dummy 2004, Dummy 2005, Dummy 2006, Dummy 2007, Dummy 2008, Dummy 2009, Dummy 2010, Dummy 2011, Dummy 2012 and Dummy 2013. Obviously, a dummy is assigned value 1 if the fiscal year of the data equals the year of the dummy, and value 0 otherwise.
Furthermore, dummies are constructed for some special industries and sectors, which are expected to have deviating average market-to-book ratios. Whether this is indeed the case and relevant, will be tested when analyzing regression results. Dummies are constructed based on the GIC Industries variable obtained from COMPUSTAT, dummies in the final dataset are:
-Dummy Pharma, with value 1 for all firms with a Genaral Industry Code (“GIC”) starting with “3520” (Pharmaceuticals & Biotechnology). This is the case for a total of 154 records of 20 firms.
-Dummy Financial, with value 1 for firms with a GIC starting with “40” (Financials). This is the case for 881 records of 124 firms.
-Dummy High Tech, with value 1 for firms with a GIC starting with “4520” (Technology Hardware & Equipment). This is the case for 210 records of 25 firms.
Initially, the intention was to include a dummy for the industry of utility corporations as well. However, the final dataset included only one corporation in the utilities industry, the company “Star Gas Partners”. Obviously, this does not provide enough data to justify the creation of a distinct dummy for the utilities industry.
To incorporate the effect of the size of a firm, data is collected on the number of employees of a firm. An alternative approach could have been to consider the market values to assess the firm size, however, this variable is closely related to the dependent variable under investigation, the market-to-book ratio. Recall that the numerator of the market-to-book ratio itself actually is the market value. Therefore this research could provide more valuable information if we include variables for firm size, based on the number of employees instead. Two new variables are constructed. The first variable is formed by taking the logarithm of the number of employees of the firm, the result is a scalar type variable. The second variable that is constructed is a dummy variable called “Dummy Firm Size”. This variable assigns value 1 to all records with more than 8,800 employees, the median number of employees for all the 5,216 records. The value 0 is attached to the lower half of the data set.
The book values of a company, the denominator of market-to-book ratios, may differ for a company, depending on the accounting standards that are used to estimate the book value. In
order to be able to make correct comparisons between companies, it is desirable for this research to use data of companies that are audited by similar sets of accounting standards. Of our dataset of 5,216 records, as obtained from COMPUSTAT, 4,647 records have an Accounting Standard Type variable called “DS”. According to the website http://wrds-web.wharton.upenn.edu, the source of the data, the value “DS” stands for “Domestic Standards”, without any specification whether these standards are consistent with IFRS, with GAAP or with any other set of accounting principles. For 398 records the Accounting Standard Type is “DU”, which means “Domestic Standards generally in accordance with United States GAAP”. For 168 records the Accounting Standard Type is “DI”, “Domestic Standards generally in accordance with or fully compliant with International Accounting Standards (IFRS)”. And 3 of the 5,216 records in total have no value for this variable. In order to maintain a large dataset and because for the vast majority of the records it is not specified which accounting principles exactly are used, it is decided to keep all 5,216 records and not to make a further selection based on the variable “Accounting Standard Type”. The implicit risk of deviations in book values (and as a consequence deviations in market-to-book ratios), originating from applying different accounting principles, is accepted in this case.
Some checks are done on the available final dataset. For instance, the book value per share indeed equals total book value (including reserves and retained earnings) divided by the number of outstanding shares. Also the share price at the fiscal end date is consistent with the total market value at fiscal end date, divided by the number of outstanding shares. To construct the market-to-book ratios, the market and book values per share are used.
The table below summarizes some descriptive statistics of the most important variables of the data set. It shows the mean, standard deviation, median, minimum and maximum of variables obtained from the COMPUSTAT database and also of the two constructed variables the market-to-book ratio and the relative advertising expenses, both of these two variables are constructed from variables initially obtained from the database.
Table 1: Descriptive statistics of observed data (5,216 records) from the COMPUSTAT database
The table above shows descriptive statistics of the data used in the regression models, definitions of the variables: Share price: The price of a share in U.S. dollars, measured at the closing date of the end of the fiscal year; Book value per share: The book value per share in U.S. dollars, measured at the end of the fiscal year; Market-to-Book ratio: This variable is constructed by dividing the Share Price by the Book value per share;
Advertising Expenses: The total amount of money spent on advertising by a firm in a fiscal year, measured in millions of U.S. dollars;
Net Sales: The realized net sales of a firm in a fiscal year, measured in millions of U.S. dollars;
Advertising / Net Sales: This variable is constructed by dividing the Advertising Expenses by Net Sales forms; Dummy 2003-2013: Dummy variables with value one if the data is from the same fiscal year, zero otherwise; (*) Dummy 2003 will eventually not be included in the regression models. Since dummies are included for all other possible years, including this dummy would be superfluous. Running the regression including the variable “Dummy 2003” would lead to sub optimal results;
Dummy Pharma: Dummy variable with value one if the data is from a firm operating in the pharmaceutical industry, zero otherwise;
Dummy Financial: Dummy variable with value one if the data is from a firm operating in the financial industry, zero otherwise;
Dummy High Tech: Dummy variable with value one if the data is from a firm operating in the high technology industry, zero otherwise;
Dummy Firm Size: Dummy variable with value one if the data is from the top 50% largest firms measured by the number of employees, zero otherwise.
Descriptive Statistics MODEL I Minimum Maximum Mean Std. Deviation Share Price 0.21 337.74 33.30 26.47 Book value per share 0.06 625,395.00 134.53 8,659.16 Market-to-Book ratio 0.00 42.77 3.27 3.74 Advertising Expenses 0.00 9,729.00 252.00 695.64 Net Sales 10.51 474,259.00 11,398.02 29,386.64 Advertising / Net Sales 0.00 0.49 0.03 0.04 Dummy 2003 (*) 0.00 1.00 0.07 0.25 Dummy 2004 0.00 1.00 0.08 0.27 Dummy 2005 0.00 1.00 0.08 0.28 Dummy 2006 0.00 1.00 0.09 0.28 Dummy 2007 0.00 1.00 0.09 0.29 Dummy 2008 0.00 1.00 0.09 0.29 Dummy 2009 0.00 1.00 0.09 0.29 Dummy 2010 0.00 1.00 0.10 0.30 Dummy 2011 0.00 1.00 0.10 0.30 Dummy 2012 0.00 1.00 0.11 0.31 Dummy 2013 0.00 1.00 0.09 0.29 Dummy Pharma 0.00 1.00 0.03 0.17 Dummy Financial 0.00 1.00 0.17 0.37 Dummy High Tech 0.00 1.00 0.04 0.20 Dummy Firm Size 0.00 1.00 0.50 0.50
This table shows that the average level of advertising spending is 0.03. Considered in more decimals, the average advertising expenses are 2.619% of a firm’s net sales. The maximum level of advertising expenses is 49% of net sales for this data set. Remind that the data set consists of mature companies, which are listed on the New York Stock Exchange. Intuitively the percentage of advertising expenses relative to net sales could be higher than 49% in reality, but most likely only in rare cases of fast growing start up firms. The average market-to-book ratio is 3.27 in this data set, which is lower than the average market-to-book ratio of 5.0 of the S&P 500 corporations in 2001, which was found by Lev (2001). The average values of the dummy variables simply show the rate of occurrence of the particular event. The average values of the dummy variables for the years 2003 until 2013 add up to exactly 100%, since each record originates from one of these years.
3.2 Regression model and estimation method
The basic regression model that is estimated has one dependent and fifteen independent variables. The equation below shows the variables and coefficients that form regression model I, the initial full regression model.
Equation 1: Regression “Model I”
M-to-B = c + 1 * RelativeAdvertisingExpenses + 2 * DummyFirmSize +
3 * DummyPharma + 4 * DummyFinancial + 5 * DummyHighTech +
6 * Dummy2004 + 7 * Dummy2005 + 8 * Dummy2006 + 9 * Dummy2007 +
10 * Dummy2008 + 11 * Dummy2009 + 12 * Dummy2010 + 13 * Dummy2011 +
14 * Dummy2012 + 15 * Dummy2013
M-to-B:
The dependent variable in the regression model is the variable called “M-to-B”, the market-to-book ratio of a company. The market-to-market-to-book ratio is the share price of a company, divided by its book value per share, both measured exactly at the end of the company’s fiscal year.
c:
The regression model contains a constant “c”, estimated with the help of Ordinary Least Squares (OLS) estimates.
1-15:
The Betas are the regression coefficients, estimated with the help of Ordinary Least Squares (OLS) estimates.
RelativeAdvertisingExpenses:
The “Relative Advertising Expenses” variable is the only independent variable of the scalar type. This variable is constructed by dividing the total incurred expenses on advertising of a firm in a fiscal year, by the firm’s realized net sales of the same fiscal year. The other fourteen independent variables are dummy variables.
DummyPharma:
This dummy variable has value one if the firm is operating in the Pharmaceutical industry and zero otherwise.
DummyFinancial:
This dummy variable has value one if the firm is operating in the Financial industry and zero otherwise.
DummyHighTech:
This dummy variable has value one if the firm is operating in the High Technology industry and zero otherwise.
Dummy2004-2013:
The dummy variables for the years have value one if the data originates from the year of the dummy and zero otherwise. All of the collected data is from the period 2003 until 2013. No dummy for the year 2003 is included in the regression model. The data of 2003 can be distinguished from data of other years, because it has no single year dummy with value one. Including an extra dummy for the year 2003 would be superfluous.
Figure 1 below shows the relationship as described in the regression model and in line with the developed hypotheses 1 and 2 (see section 2.2 “Hypotheses”):
Figure 1: Mechanism of independent variables in relation to the dependent variable, the market-to-book ratio
Firstly, the two hypotheses that are constructed in section 2.2 “Hypotheses” are tested for the full regression “Model I”. This regression includes all the aforementioned fifteen independent variables (see “equation 1” above). As is done for all other models, this regression is carried out on the complete data set, which includes all 5,216 records.
After having run the first regression model, a second model is tested, “Model II”. “Model II” is a variant on “Model I”, with a modification for the variable of the firm size. Instead of the
inclusion of the original dummy variable, a variable is included that is formed by the logarithm of the number of employees of a firm. In other words, in “Model II” a scalar type variable called “Log Employees” is included, instead of the dummy variable “Dummy Firm Size”.
The third regression model, “Model III”, is another variant of model I. The difference from “Model I” is that the dummies for the years 2004 until 2013 are omitted from the regression model. Instead, a dummy is included that separates the years before the financial crisis from the years during (or possibly after) the financial crisis. The new variable is called “Dummy crisis”. This dummy simply attaches a value of one to all records of the years 2008 and after, and a value of zero is attached to the years 2003 until 2007. This division of years is chosen since the
financial crisis, also called the “credit crunch”, began to dominate financial markets in the year 2008. For example, in September 2008 the financial services firm Lehman Brothers filed for bankruptcy protection. It can be argued that the financial climate started to improve around the year 2013 again. For this analysis, however, no further distinction is made. The adopted variable “Dummy Crisis” just separates the years 2003 until 2007 from the period as of the year 2008.
Relative
Advertising
Expenses
Market-to-BookRatio
Dummy Firm Size
(top 50% records with most employees)
+
not include the variables for the three specific industries, the financial, pharmaceuticals and high technology sector. Besides the dummies for the separate years, “Model IV” includes the
independent variables “Relative advertising expenses” and “Dummy firm size”.
“Model V” includes the three dummy variables for the sectors again, but excludes the dummies for the fiscal years. The model has five independent variables in total, being the three sector dummies, the variable “Relative advertising expenses” and the dummy variable for the firm size. The final model that is analyzed is “Model VI”. This model excludes the dummies for the fiscal years as well as the dummies for the three sectors. The only two independent variables remaining are the “Relative advertising expenses” and the “Dummy Firm Size”.
Of all six regression models, the goodness of fit is tested. This is a test for the usefulness and suitability of the model in general and carried out with an (adjusted) R-squared test and an F-test. Also the residuals are analyzed, the residuals being the discrepancies between the predicted values for the dependent variable and the observed values. The descriptive statistics of the residuals are evaluated. To assess the estimations of the coefficients of the independent variables, t-tests are conducted for all variables separately and the results are interpreted. Hypotheses 1 and 2 will be rejected at a significance critical level of alpha = 0.05 (5%). The regression analysis is performed with SPSS software package.
3.3 Correlation matrix
The table below shows the correlation matrix of all the 15 independent variables that are included in the regression of “Model I”. These results are obtained by conducting the Pearson method within the SPSS software package, in order to calculate the correlation values.
Table 2: Correlation matrix of independent variables of “Model I”
The variable “Advertising / Net Sales” represents the relative advertising expenses variable. See section 3.1 “Data and sample selection” for more information on the variables included in the table above.
Pearson Correlation Matrix Advertising / Net Sales Dummy 2004 Dummy 2005 Dummy 2006 Dummy 2007 Dummy 2008 Dummy 2009 Dummy 2010 Dummy 2011 Dummy 2012 Dummy 2013 Dummy Pharma Dummy Financial Dummy High Tech Dummy Firm Size
Advertising / Net Sales 1.00
Dummy 2004 0.01 1.00 Dummy 2005 0.00 -0.09 1.00 Dummy 2006 0.00 -0.09 -0.09 1.00 Dummy 2007 0.00 -0.09 -0.10 -0.10 1.00 Dummy 2008 -0.01 -0.09 -0.10 -0.10 -0.10 1.00 Dummy 2009 -0.01 -0.09 -0.10 -0.10 -0.10 -0.10 1.00 Dummy 2010 0.00 -0.10 -0.10 -0.10 -0.11 -0.10 -0.11 1.00 Dummy 2011 0.00 -0.10 -0.10 -0.11 -0.11 -0.11 -0.11 -0.11 1.00 Dummy 2012 0.01 -0.10 -0.10 -0.11 -0.11 -0.11 -0.11 -0.11 -0.12 1.00 Dummy 2013 -0.01 -0.09 -0.10 -0.10 -0.10 -0.10 -0.10 -0.11 -0.11 -0.11 1.00 Dummy Pharma 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 1.00 Dummy Financial -0.10 0.00 0.01 0.00 0.00 0.00 -0.01 0.00 -0.01 0.00 0.01 -0.08 1.00
Dummy High Tech -0.09 0.00 0.01 0.00 0.01 0.00 0.00 0.00 0.00 0.00 -0.01 -0.04 -0.09 1.00
This Pearson correlation matrix shows the correlation coefficients of the independent variables that are included in the model. The largest absolute value of all the calculated correlation coefficients is 0.13, which is the correlation coefficient of the dummy that indicates whether firms are active in the financial sector and the dummy for firm size. This value of (minus) 0.13 is relatively small and since the coefficients for all other correlations are even smaller, in absolute terms, the conclusion can be drawn that only rather low correlation exists between the independent variables of this regression model. Based on this correlation matrix we do not have to exclude any independent variables from the regression model (“Model I”).
The maximum absolute values of the correlation coefficients of the other regression models are, 0.163 for “Model II”, 0.131 for “Model III”, 0.117 for “Model IV”, 0.129 for “Model V” and 0.001 for “Model VI”. There is no indication to exclude any independent variables of any of the regression models, “Model I” to “Model VI”.
4 Results
Model I
The first analysis is performed on a general level, where the relevance of the model as a whole is assessed. This analysis includes an evaluation of the value of the adjusted R squared, and an F test is performed. The adjusted R squared value always varies in the range from zero to one and provides insight into the strength of the linear relation of the model. It gives insight into the rate at which the model explains the variances of the dependent variable.
The regression model turns out to have an adjusted R squared value of 0.034. This means that only 3.4% of the variance of the dependent variable, the market-to-book ratio, is explained by the independent variables included in this model. The low value of the adjusted R squared implies that the adopted regression model does not provide proper predictions for the value of the dependent variable. It seems that other factors and variables that influence the market-to-book ratios are not included in the model.
As indicated in the introduction, brand value, influenced by a firm’s chosen level of spending on advertising, is only one of a range of intangible assets. For instance, the internal investments and efforts made with regards to Research & Development is another example of an intangible asset. In addition, human resource intangibles and organizational capital are considered as other forms of intangible assets. All of these assets are separate drivers of the market value of a firm, and as a consequence, these assets all affect a firm’s market-to-book ratio. In this regression analysis only one of these intangibles is included, the level of spending on advertising. Hence, the finding that the adopted regression model explains variances in the market-to-book ratios for only a small portion does make sense. The finding of this research that this portion is only 3.4%, under these specific circumstances, is a new insight.
The value of the performed F-test for the general regression model is 13.422, with a significance level (or p-value) of 0.000. This p-value is less than the critical significance level of 0.05. That means that there is strong evidence to reject the null hypothesis, where the null hypothesis says that the model has no explanatory power. This finding seems to suggest that the model is suitable for predicting the value of the dependent variable, the market-to-book ratio.
The next step is to investigate the results on the coefficients of the independent variables separately. T tests on the coefficients of the independent variables are conducted and results are interpreted.
The null hypotheses in our analysis say that the independent variables are irrelevant and the coefficients of the variables are zero. If values of the t-tests are, in absolute terms, exceeding 1.96, than the null-hypotheses are rejected. In that case the variables are assumed to actually be relevant. T-values above 1.96 correspond to significance levels below the critical level of 0.05. In that case we conclude that the assumptions of the null hypothesis are not significantly verified, and hence we do not accept the null hypothesis. Therefore, we would assume that the estimated coefficient for the independent variable is significantly different from zero and indeed should be included in the regression model.
In our model we included a constant term, “c”. The regression analysis provides a value for this constant of 3.589, with a t-value of 17.209 and a p-value of 0.000. This means that the null-hypothesis, which says that the constant term has a value of zero, will be rejected. In other words, we should include this constant term in regression “model I” with a coefficient of 3.589. The first independent variable to consider is the “Relative advertising expenses” variable, which is the total amount spent on advertising divided by the total net sales of a firm in a fiscal period. The t-value of this coefficient is 4.812 and the corresponding p-value is 0.000. This means that there is strong empirical evidence to reject the null hypothesis, where the null hypothesis says that the coefficient of the variable is equal to zero. There is strong empirical evidence to include this variable in the regression model, with the estimated value for the coefficient. In other words, this independent variable is significantly relevant and should be included into our regression model.
The value of the coefficient of the “Relative advertising expenses” variable that is estimated by the model is 6.989. This means that an increase of one unit of the relative advertising expense will increase the market-to-book ratio by a predicted 6.989. In this case an increase of one unit of relative advertising expense means an increase of the advertising budget of the size of the total net sales. The magnitude of advertising expenditures will in reality probably never reach that level of 100% of net sales, at least not for mature firms that are listed on the Dow Jones Stock Exchange. Maybe only in rare situations of fast growing start up firms this level could be reached. From the descriptive statistics of the data set we know that the average level of advertising spending is 2.619% of net sales (see section 3.1 “Data and sample selection”).
If we interpret the coefficient of the relative advertising expense on a percentage level, it can be concluded that an increase of the advertising expenses of the size of 1.0% of a firm’s yearly net sales, leads to an increase of the market-to-book ratio of 0.06989 point (or rounded in three
decimals 0.070). This gives a more realistic and relevant interpretation of the estimated coefficient for the relative advertising expenses variable.
This result supports the hypothesis 1, which is developed in section 2.2. There is strong empirical evidence from this research to conclude that advertising expenses have a significantly positive effect on market-to-book ratios of corporations. Also, this result is in line with findings from prior literature, for instance the findings of Lane and Jacobson (1995), Chauvin and Hirschey (1993) and Srinivasan and Hanssens (2009). That literature formed the foundation of the development and formulation of hypothesis 1 of this thesis.
The remaining independent variables concern dummy variables, which take the value one or zero. At a 0.05 critical significance level (5%), the dummies for the years 2008, 2009, 2011 and 2013 are significantly different from zero. In other words, there is no evidence to include dummies for the other years 2004 until 2007, 2010 and 2012 in the regression model. The coefficient for the Dummy 2008 is –1.025, for Dummy 2009 the coefficient is -0.682 and for Dummy 2011 -0.564. The sign of these coefficients indicates that market-to-book ratios were, in general, lower for these years. This finding seems to be in line with the “bearish”, negative sentiment on the stock markets in those years. The years as of 2008 are known for the financial crisis and the credit crunch, with associated decreasing share prices. In these years the general expectations for the worldwide economic development were more pessimistic than in other years, resulting in, on average, lower market-to-book ratios of firms than in other years. This statement is supported by the development of the Dow Jones Industrial Average (DJIA), an index of the stock market performance which is based on the share prices of 30 large publicly traded corporations. This index was at a height of 13,264 points at closing time of the 31st of December of 2007. Exactly one year later the index closed at 8,776 points. This means that the thirty large firms that form the DJIA lost 33.8% of their market value in the year 2008. This magnitude of decline is exceptional and exemplifies the negative economic sentiment that prevailed throughout 2008.
The estimated coefficient for the dummy for the year 2013 is 0.585, a positive number. This finding is in line with the “bullish”, positive sentiment, which prevailed on the New York Stock Exchange in 2013, and especially at the end of 2013. Since most firms included in the data set report their financial results at the end of the calendar year, most market-to-book ratios of the data set are measured at the end of the calendar year 2013. To give and illustration of the positive sentiment on stock markets that year, the DJIA index went up from 13,104 points at the end of 2012 to 16,577 at the end of 2013, a sound increase of 26.5%.
The estimated coefficients of the dummy variables for the three specific industries: The financial, pharmaceuticals and high technology industry, are all significantly different from zero and therefore should indeed be included in the regression model.
The estimated coefficient for Dummy Financial is -1.292, this indicates that firms operating in the financial industry have lower market-to-book ratios on average than companies of other industries. This effect could have been enlarged by the financial crisis that evolved during some years of the period under investigation, 2003 until 2013. This financial crisis, or sometimes referred to as the “credit crunch”, started to take a flight during the year 2008. In this year and for some years after, financial institutions faced heavy financial risks and lower stock prices. Intuitively, this could be one of the drivers of the lower average market-to-book ratios of firms operating in the financial industry, than the average market-to-book ratios that are common in other industries.
The estimated coefficient for Dummy Pharma is 0.766, indicating higher average market-to-book ratios. The coefficient for the Dummy High Tech is estimated at -0.721, indicating lower market-to-book ratios for firms active in the high technology industry than average market-market-to-book ratios for companies of all sectors together.
The estimated coefficient for the dummy for firm size is not significantly different from zero. The t-statistic of the estimated coefficient for the variable “Dummy Firm Size” is -0.75 and, with a corresponding p-value of 0.453. Clearly, there is not enough empirical evidence to reject the null hypothesis and to include this variable in our regression model and with this coefficient. As a consequence, we have to reject the second hypothesis that we adopted in this research (see section 2.2 “Hypotheses”) which says that larger firms have higher market-to-book ratios. Also the opposite, that smaller firms have higher market-to-book ratios, is not proven by this research.
The table below provides an overview of the regression results of “Model I”. The table also gives the results for regression Models II, III, IV, V and VI. Those regressions are conducted on the same total data set of 5,216 observations. The difference between the models is within the set of independent variables that are used for the regression analysis. See more detailed information on the several regression models in section 3.2 “Regression model and estimation method”.
Table 3: OLS regressions of the various models (I, II, III, IV, V and VI). The dependent variable of all regression models is the market-to-book ratio. Data is collected from the COMPUSTAT database and includes 5,216 records of firms listed on the NYSE of the period 2003 until 2013.
*, **, *** Denote significance at the 10% level, 5% level, and 1% level, respectively; a) Because the variable “Dummy 2003”is omitted, it is included in the “Constant”.
Prediction coefficient t-statistic coefficient t-statistic coefficient t-statistic coefficient t-statistic coefficient t-statistic coefficient t-statistic
Constant 3.589 17.209*** 3.680 10.827*** 3.521 31.674*** 3.251 15.735*** 3.344 36.829*** 3.013 36.591***
Relative advertising expenses + 6.989 4.812*** 6.952 4.774*** 7.043 4.829*** 8.786 6.055*** 7.097 4.863*** 8.879 6.091***
Dummy Firm Size + -0.077 -0.750 - - -0.09 -0.869 0.050 0.491 -0.077 -0.743 0.049 0.477
Log Employees + - - -0.140 -0.494 - - - -Dummy 2004 -0.115 -0.434 -0.114 -0.432 - - -0.118 -0.441 - - - -Dummy 2005 -0.070 -0.269 -0.071 -0.271 - - -0.086 -0.325 - - - -Dummy 2006 -0.035 -0.135 -0.036 -0.139 - - -0.042 -0.163 - - - -Dummy 2007 -0.125 -0.488 -0.125 -0.488 - - -0.126 -0.487 - - - -Dummy 2008 -1.025 -4.008*** -1.025 -4.004*** - - -1.017 -3.938*** - - - -Dummy 2009 -0.682 -2.676*** -0.681 -2.671*** - - -0.665 -2.586*** - - - -Dummy 2010 -0.321 -1.272 -0.320 -1.269 - - -0.314 -1.236 - - - -Dummy 2011 -0.564 -2.254** -0.563 -2.249** - - -0.550 -2.177** - - - -Dummy 2012 -0.160 -0.648 -0.159 -0.642 - - -0.160 -0.641 - - - -Dummy 2013 0.585 2.298** 0.586 2.300** - - 0.589 2.289** - - - -Dummy Crisis - - - - -0.285 -2.750*** - - - -Dummy Pharma 0.766 2.540** 0.763 2.528** 0.781 2.579*** - - 0.782 2.582*** - -Dummy Financial -1.292 -9.312*** -1.290 -9.239*** -1.283 -9.203*** - - -1.278 -9.164*** -
-Dummy High Tech -0.721 -2.759*** -0.721 -2.757*** -0.732 -2.788*** - - -0.723 -2.752*** -
-Adjusted R-squared 0.034 0.034 0.026 0.019 0.025 0.007
F-statistic 13.422*** 13.400*** 24.083*** 8.187*** 27.352*** 18.666***
Total observations 5,216 5,216 5,216 5,216 5,216 5,216
Model III OLS Regressions
The table below shows descriptive statistics of the predicted values of the dependent variable, the market-to-book ratio, and of the residuals of regression “Model I”.
Table 4: Descriptive statistics of the Predicted Value and Residuals (regression “Model I”)
Descriptive statistics are shown of the predicted value for market-to-book ratios by regression “Model I”. In addition, the table shows descriptive statistics of the residuals, the difference between predicted and actual observed values of market-to-book ratios.
The mean of the predicted values of the market-to-book ratios is 3.27. This value is indeed equal to the actual mean of market-to-book ratios as observed in the original data set (see “table 1” for descriptive statistics of the actual observed data). The standard deviation of the predicted value is 0.72. This implies that around 95.5% of the predicted values for market-to-book ratios are in the range of 1.83 to 4.71. The residuals are the differences between actual observed market-to-book ratios and the predicted values by the model. The mean of the residuals is zero, obviously, and the standard deviation of the residuals is 3.67.
In order to provide some visual support, a two dimensional scatterplot can be drawn if the market-to-book ratio, the dependent variable, is regressed on one independent variable only. In this case, that independent variable is the scalar variable “Relative advertising expenses”. This situation clearly does not represent the full regression model of “Model I”. Actually, regression “Model I” did include fifteen independent variables, rather than one. However, the scatterplot below (figure 2) does give a first glance of the relation between the only two scalar variables of the full regression model.
Descriptive Statistics MODEL I Minimum Maximum Mean Std. Deviation
Predicted Value 1.21 6.97 3.27 0.72
Figure 2: Scatterplot of the relation between the variables “Relative advertising expenses” (independent variable) and “market-to-book ratio” (dependent variable). The graph includes an automatically generated regression line and all of the 5,216 observations.
This scatterplot contains all 5,216 observations and includes an automatically generated regression line. Because of the great number of observations, the relation is not clearly visible. A second scatterplot below (figure 3) zooms in on a part of the scatterplot above, the region where most observations lie. This is in the range 0 to 0.1 for the relative advertising expenses and in the range 0 to 10 for the market-to-book ratio. Again, the scatterplot below just shows a part of the plot above.
Figure 3: Zoomed version of figure 2. The graph includes an automatically generated regression line.
Also, in this zoomed version of the scatterplot, no clear relation between the two variables seems to be apparent. Intuitively, this finding is in line with the low value of the adjusted R-squared of 0.034, for regression “Model I” considered as a whole.
Model II
Similar to the first regression model, the second regression model regresses the market-to-book ratio against fifteen independent variables. The only modification in Model II is that it includes the scalar type variable “Log Employees”, which is the logarithm of the number of employees of a firm. The model leaves out the original variable “Dummy Firm Size”. As can be seen in table 3, the results of this regression analysis are close to the results of Model I. The adjusted R-squared exactly remains at a level of 0.034. The F-statistic of the model is 13.400, almost equal to the
variables are significant as in the first model. Again, the independent variable “Relative advertising expenses” is significant and the coefficient is close to the value estimated in “Model I”. The newly adopted variable for the size of firms, “Log Employees”, is not significantly different from zero. Similar to “Model I”, in “Model II” we accept hypothesis 1; there is a positive relation between spending on advertising and the market-to-book ratios of companies. Also we reject hypothesis 2, no evidence is found that the size of a firm, measured by its employees, has a significant impact on market-to-book ratios. As a consequence, we cannot conclude that larger firms are able to spend their advertising budgets in a more efficient way than smaller firms.
Model III
As described in section 3.2 “Regression model and estimation method”, “Model III” is another variant of “Model I”. The difference with “Model I” is that the dummies for the years 2004 until 2013 are omitted. Instead, the dummy variable “Dummy crisis” is included. As described before, this dummy variable separates the records of the years 2003 until 2007 from the records of the years 2008 until 2013. Now, we see the adjusted R-squared rate drop from 0.034 to 0.026, compared to “Model I”. The F-statistic of the model is again significant at a 1% significance level, the F-statistic even increases to a level of 24.083. Qualitative similar results appear for this regression model as was seen for “Model I”. The variable “Relative advertising expenses” is significant at a one percent significance level and the coefficient is positive and close to the value estimated in “Model I”. The dummy for firm size is not significant, not even at a ten percent significance level. This means that, again, hypothesis I is accepted and hypothesis 2 is rejected. The three dummies for the sectors are all significant at a one percent significance level. The signs and magnitude of its coefficients are in line as the results of “Model I”. The new variable “Dummy Crisis” is significantly different from zero and should be included in the model with a coefficient of -0.285. This finding does intuitively make sense; in times of crisis, market values of firms, and as a consequence market-to-book ratios, are appreciated at lower levels than in “regular” market conditions where expectations on general economic development are higher.
Model IV
In the fourth regression model, “Model IV”, dummies are omitted for the three specific sectors, the financial, pharmaceutical and high technology industry. This regression model turns out to be
useful since the F-test statistic has a value of 8.187. However, the adjusted R squared drops to a value of 0.019. That means that this model is only able to explain 1.9% of the variances of the dependent variable, the market-to-book ratio. Similar to the results of “Model I”, again the “Relative advertising expenses” variable is significant at a 1% significance level. This implies that, according to the results of the regression analysis, an increase of the advertising expenses of 1% of net sales leads to an estimated increase in the market-to-book ratio of 0.088. In accordance of the results of “Model I”, hypothesis 1 is accepted; higher spending on advertising leads to higher market-to-book ratios.
Also similar to “Model I”, the variable “Dummy Firm Size” is not significant and should not be included in the regression model. This means that, again, the second hypothesis is rejected; larger firms do not have higher market-to-book ratios.
At a five percent significance level, exactly the same year dummies are significant in “Model IV” as they were in “Model I”, Not surprisingly, the coefficients of these variables all have the same signs and are more or less of the same size, as was the case for the first regression model.
Model V
The fifth regression model includes only five independent variables, which are the variable called “Relative advertising expenses”, the dummy for the firm size and three dummies for the aforementioned specific industries. The results of this analysis are in line with the models that are investigated before. As can be seen in table 3, the only variable that is not significantly different from zero at a five percent significance level is the variable “Dummy firm size”. All other four independent variables are significant, even at a one percent significance level. This means that also for regression “Model V” hypothesis 1 is accepted and hypothesis 2 is rejected.
Model VI
The sixth model that is evaluated is a regression model, which includes only the two independent variables “Relative advertising expenses” and “Dummy firm size”. The difference with “Model I” is that all the dummies for the fiscal years are excluded as well as the dummies for the three specific industries (financial, pharmaceutical and high technology). The data set under investigation is the same as for all other models, which is the original dataset that includes all 5,216 records.
The adjusted R-squared now is 0.007. This indicates that the model has hardly any explanation power of the dependent variable, the market-to-book ratio, at all. Still the model is useful, regarding the F-test statistic of 18.666. The relative advertising expenses variable is the only significant independent variable of this model (see table 3 for more details on the regression coefficients).
The results of regression “Model VI” are consistent with the findings from all the former regression models. Relative advertising expenses are significant and positively related with market-to-book ratios. On the other hand, the size of a firm, measured by the number of employees, does not seem to be related to market-to-book ratios.
Even though the fact that all the regression models that we have developed in this thesis do not have strong explanatory power, as can be seen by the rather low values of the adjusted R-squared tests, all models are significant and useful. This statement is supported by the fact that all of the six models have F-statistics that are significant at a one percent significance level.
All six variants of the regression models showed similar results: Hypothesis 1 is accepted in all occasions and hypothesis 2 is rejected. In other words, this research has shown that a positive relation exists between the level of spending on advertising and the value of market-to-book ratios. Higher spending levels lead to higher expectations on the future financial performances of companies. At all investigated models no evidence has been found to conclude that the size of companies affects the value of market-to-book ratios. No evidence is found that suggests that larger firms are able to spend their advertising budgets more efficiently than smaller firms.
5 Conclusions and Discussion
The effectiveness of advertising campaigns cannot always be measured by the amount of dollars spent in reality. However, spending on advertising does seem to positively influence market valuations and market-to-book ratios of firms. The empirical results seem to prove that investors appreciate advertising expenditures as a sort of internal investments, which is expected to generate future financial benefits and not only boost sales in the current period.
The size of a company does not seem to be related with market-to-book ratios. The assumption that larger firms are more efficient in their advertising efforts than smaller firms, is not apparent. In this research a positive linear relationship between advertising expenditures and market-to-book ratios is investigated, the possibility for a non-linear relationship is left out of the scope of this thesis. For instance, beyond a certain level of spending on advertising, additional expenditures could be considered inefficient and could have a different impact on market-to-book ratios. The existence of this effect could be investigated in further research.
Further research could also try to decrease the level of bias from external factors. For instance, the bias from measuring market-to-book ratios at different moments in time could be controlled more strictly. In this research dummies are included for each fiscal year of investigation, however, market conditions change rapidly and are not constant over a year. Including dummies for shorter periods of time could decrease the risk and bias from changes in general market conditions even further. This procedure could result in more accurate and valuable information to assess the relation between advertising expenses and market values of firms.
Literature
Chauvin, K.W. and Hirschey, M. (1993), “Advertising, R&D expenditures and the market value of the firm”, Financial Management, Vol. 22 No. 4, pp. 128-40
Joshi, A. M. and D. M. Hanssens (2009), "Movie Advertising and the Stock Market Valuation of Studios: A Case Of "Great Expectations?", Marketing Science, Vol 28 (2), 239-50.
Joshi, A. M. and D. M. Hanssens (2010), "The direct and indirect effects of advertising spending on firm value", Journal of Marketing, Vol. 74, 20-33.
Lane, V. and Jacobson, R. (1993), “Stock market reactions to brand extension announcements: The effect of Brand attitude and familiarity”, Journal of Marketing, Vol. 59 January 1995, p.63-77 Larkin, Y. (2013), “Brand perception, cash flow stability and financial policy”, Journal of
Financial Economics (forthcoming)
Lev, B., Radhakrishnan, S. and Zhang W. (2009), Organizational Capital, Abacus a Journal of Accounting, Finance and Business Studies, Vol. 45, No. 3 pp. 275-298.
Lev, B. and Sougiannis, T. (1996), “The capitalization, amortization, and value-relevance of R&D”, Journal of Accounting and Economics, Vol. 21, pp. 107-38.
Lev, B. and Zarowin, P. (1999), “The boundaries of financial reporting and how to extend them”, Journal of Accounting Research, Vol. 37, No. 2, pp. 353-385.
Lev, B. (2001), “Intangibles: Management, and Reporting”, Brookings Institution Press, Washington, DC.
Srinivasan, S. and Hanssens, D. (2009), “Marketing and firm value: Metrics, methods, findings and future direction”, Journal of Marketing Research, Vol. 46, p 293-312
Srinivasan, S., Hsu, L. and Fournier, S. (2011), “Branding and firm value”, Handbook of Marketing and Finance, Edward Elgar Publishing.
