Literature Review

Our study is part of the literature that focuses on the association between advertising expenditures and outcome-based measures, such as earnings and cash flows. Various studies across the disciplines of marketing, accounting, and finance suggest that advertising spending can, directly and indirectly, affect sales, firm value, and market participants (Madsen and Niessner 2019; Srinivasan and Hanssens 2009). Research focusing on advertising primarily employs two approaches.

The first stream of research focuses on the ways that advertising influences consumer behavior from a psychological perspective. Specifically, these studies examine how advertising triggers cognition or affect to change consumers' behavioral actions (e.g., make purchases). The basic premise that advertising influences consumer behavior from a psychological perspective has general implications, such as greater market awareness, quality competitiveness, customer preference, and brand image (Koslow et al. 2006; Kulkarni et al. 2003; Tellis 2009). These benefits induced by advertising, in turn, boost the future sales and profits of the firm (Kirmani and Wright 1989; Leone 1995; Mela et al. 1997; Osinga et al. 2010). Further, Srinivasan and Hanssens (2009) suggest a link between advertising and instant awareness of new products, which would imply that firm advertising leads to more and faster cash flows.

The second stream of research focuses on the economic gains of advertising expenditures created by consumers. The focus of this stream is outcome-based accounting performance measures, such as sales. These studies collect observable constructs from annual reports, financial statements, or private data sets. The underlying assumption is that advertising produces immediate shifts in consumer purchase and product selection (Wang et al. 2009). These immediate shifts have a significant impact on measures of firm performance such as sales volume (Givon and Horsky 1990; Joshi and Hanssens 2010) and market share (Wang et al. 2009).

Studies in this second research stream find that advertising helps smooth out the variability in seasonal demand and reduces consumer risks with safer cash flows. Byzalov and Shachar (2004) note that advertising can help consumers to deal with risk aversion. Also, advertising can create a barrier to competition, provide bargaining power vis-à-vis suppliers, and achieve “greater dynamic efficiency and flexibility in adapting environmental changes than its competitors” (McAlister et al. 2007).

Chemmanur and Yan (2009) argue that advertising can signal quality to the product market and stock market investors on the true value of a firm's projects, thus allowing them to price the firm's equity correctly in equilibrium. Boyd and Schonfeld (1977) point out a positive impact of advertising on stock prices because the purchase of products and stocks is similar: individuals who respond to advertising messages in a product setting also respond to advertising messages in a financial setting. That is, investors may pick stocks based on their familiarity with the stocks. Indeed, investors often prefer firms that have higher brand visibility induced by advertising (Fehle et al. 2005). Further, advertising enables firms to enjoy more information channels to communicate with investors and obtain greater ownership breadth and investor attention (Grullon et al. 2004). Advertising reduces investors' search costs and signals firm-specific competitiveness regarding its existing products and new projects. Chauvin and Hirschey (1993, 128) conclude that “spending on advertising is a form of investment in intangible assets with predictably positive effects on the size and variability of future cash flows.”

As discussed above, prior studies look at the association between advertising spending and firm value by either examining how advertising triggers the cognition of consumers or focusing on immediate shifts in outcome-based accounting performance measures due to advertising. However, prior research focuses less on an important feature of advertising spending that sets this type of expenditure type apart from others. Advertising reaches consumers through media vehicles. Simple intuition indicates that different media vehicles bring different firm benefits. If they were all the same and the choice to use one vehicle over another was trivial, there would not be spending variations across these vehicles. Simply put, a firm would spend its entire advertising budget on one vehicle without any regard for other options. This is not the practice. For instance, a viewer of a commercial TV broadcast encounters frequent advertisements, and advertisements are also prominent in magazines, newspapers, and radio broadcasts. A firm is willing to pay only so much for a message, and its willingness to pay is driven by the number of potential consumers that the message might reach and changes in performance measures. No other expenditure type can be further categorized using clear boundaries.

While limited, there are some studies examining various advertising media vehicles independently. Prior research shows that television advertising affects a firm's profitability (Danaher and Dagger 2013; Notta and Oustapassidis 2001; Pedrick and Zufryden 1991). Studies investigating exposure through television shows about product development find a positive relationship between these shows and positive stock returns through enhanced attention (Takeda and Yamazaki 2006). The influence of online information on firm performance reveals that online media have a strong relationship with firm performance. Among new vehicles of advertising, emails and sponsored search influence sales and profits most, but others, such as online display advertising and social media, do not have significant associations (Danaher and Dagger 2013). These studies only focus on one type of media vehicle and exclude a firm's spending on other vehicles. The findings are important to argue the importance of various vehicles, but they do not paint a full picture. Our paper focuses on understanding the association between advertising expenditures categorized by media vehicles and sales. We discuss our motivation for our research question in the following section.

Motivation

In accrual accounting, the revenue recognition principle states that revenues should be recorded during the period they are earned, regardless of when the transfer of cash occurs. In addition, the matching principle instructs that an expense should be reported in the same period in which the corresponding revenue is earned, and is associated with accrual accounting. By recognizing costs in the period they are incurred, a business can see how much money was spent to generate revenue, reducing noise from the timing mismatch between when costs are incurred and when revenue is realized.

The topic of matching has a long history in accounting. Accounting earnings are shaped by both economic factors and the degree of matching success. However, matching considerations are absent for advertising because such advertising expenditures are expenses as they are incurred. IAS 38 (IFRS Foundation 2017) requires firms to write off all expenditures in the current period. Consequently, firms record all the expenditure on advertising in the current period when profit levels are known with more certainty, rather than risking carrying these costs over into future periods as capitalized costs. The accounting practice to charge advertising expenditure to current expenses produces an implicit amortization rate of 100%.

Another important reason for the lack of matching considerations for advertising is the lack of disclosure on the financial statements. Advertising expenditures are disclosed voluntarily. In most cases, these expenditures are bundled with SG&A expenses, and firms rarely report advertising expenditures separately. Further, firms never report how their advertising spending is allocated among various media vehicles in their quarterly or annual reports.

A marketing set which describes advertising spending on every medium (i.e., TV, Internet, press, outdoor) will generate a given level of revenue output. With perfect matching, all relevant expenses should be matched against the associated contemporaneous revenue. Because advertising is presumed to generate revenue, perfect matching should lead to a significant positive association between advertising expenditures and sales.

On the other hand, if there is poor matching and noise over economic benefits gained, we would find either insignificant or negative associations. Poor matching is the extent to which expenses are not matched against the resulting revenues and is modeled as noise in the economic relation of advancing expenses to earn revenues. Poor matching can arise for several reasons, such as fixed costs, poor traceability of costs, and accounting rules.

We would argue that an insignificant association suggests poor traceability of costs, while a negative association between advertising spending and sales is an indication of advertising spending behaving like fixed costs. In other words, once an optimal level of advertising spending is reached, any additional advertising spending per unit will be associated with a decrease in sales.

If advertising expenditures and contemporaneous sales are properly matched, we predict finding a positive association. On the other hand, if they are imperfectly matched, we predict finding an insignificant or a negative association.

Data

We obtain monthly advertising information from a proprietary database constructed by CreativeClub, a media-tracking company. Our sample covers 2004 through 2014 (data availability).² The data are gathered by continuously monitoring multiple media vehicles and collecting information about observed advertisements. The media vehicles covered include (i) television (cable, network, and cinema),³ (ii) the Internet, (iii) press (newspapers, local magazines, consumer magazines, and business magazines), and (iv) outdoor (e.g., billboards, displays).⁴ CreativeClub is a UK-based advertising database run by Ebiquity (formerly known as Thomson Intermedia), which is a provider of independent marketing analytics. Our initial sample from CreativeClub includes 180 unique firms, which translates into 7,920 quarterly observations.

The translation of media spots into monetary amounts is done by surveying media vehicles for their rates at various points during the quarter. For example, for television, the rates used to estimate advertising expenditures are supplied primarily by the networks and, in some instances, by advertising agencies and advertisers. The database covers all media spending that appears in the tracked vehicles and is reported by brand. For example, the media spending by Unilever is reported separately by its brands, which include Knorr, Dove, and many others. We aggregate the advertising outlays of all brands that belong to a particular firm. We emphasize that our data cover only media spending and does not include other forms of advertising, such as the production costs of advertising and promotional materials.

Creative Club covers a select number of companies that are operating in the United Kingdom. They cover various media vehicles (e.g., TV, Internet) and the ads published in them. Later, they assign these to a company. A significant portion of their coverage includes private companies. Because we cannot control for firms' internal decision making on their marketing budget, we focus on firms for which running ads is a creative choice rather than a budgetary decision. To operationalize this choice, we make a broad assumption that firms with high market values have the financial means to use every media vehicle, should they choose to. To make sure that the choice of running ads is a creative rather than a budgetary decision and that firms in this study's sample can afford to run ads, we only include firms that have been a constituent of the Financial Times Stock Exchange (FTSE) 100 Index at least once during the sample period.⁵ We acknowledge that the data acquired with CreativeClub presents its unique limitations, since our sample selection heavily relies on what they are offering. Notwithstanding this limitation, we still believe the data on advertising expenditures to offer unique insights that can be used by regulators and researchers in other research questions.

Next, we collect the data on firm performance and other firm characteristics from the quarterly Compustat Global database.⁶ To ensure that firms in our sample do not drop out unexpectedly, we require each firm in our sample to be available for at least two consecutive years. We note that we do not always have the necessary quarterly data on Compustat Global. Our final sample covers 88 publicly listed firms in the United Kingdom and a total of 2,039 firm-quarter observations for the time period from 2004 to 2014. All monetary variables are denoted in British pounds.

Table 1 reports the detailed distribution of our sample based on year and 2-digit Global Industry Classification Standard (GICS) industries. Except for the year 2004, more than 50% of our sample is represented in each year. Beyond the year 2008, more than 80% of our sample is included in each year. Firms in the consumer discretionary and consumer staples sectors are most widely represented in the sample. On the other hand, firms in the information technology and materials sectors are the least represented sectors. Our sample does not include any firms in the real estate and financial sectors.

Approximately 60% of the unique firms in our sample use at least one of the four advertising media vehicles. When a firm does not report expenditures on one of the media vehicles, we assign zero for that variable. Since we include firms that are part of the FTSE 100 Index, we assume that not using every media vehicle is not a budgetary decision.

Variables

We define the dependent variable as sales of firm i, quarter (t), scaled by total assets, denoted as . The primary interest variables are advertising expenditures on four different media vehicles. We include (i) , measured as TV and cinema advertising expenditures of firm i, quarter t;⁷ (ii) , measured as Internet advertising expenditures of firm i, quarter t; (iii) , measured as press advertising expenditures of firm i, quarter t; and (iv) , measured as outdoor advertising expenditures of firm i, quarter t. Each advertising variable is scaled by total assets.

We include , measured as private consumption growth, which measures the growth in private consumption in the United Kingdom while adjusting for inflation, in quarter t. Private consumption represents the total monetary value of consumer purchases of goods and services. It includes all purchases made by consumers, such as food, housing (rents), energy, clothing, health, leisure, education, communication, transport, and hotels and restaurant services. It also includes durable goods (such as cars), but not households' purchases of dwellings, which are counted as household investments. We also include , measured as the Consumer Price Index, which is a price index that measures the average cost of goods and services purchased by the typical household in the United Kingdom, in quarter t.

Our study includes firm size (denoted as SIZE), calculated as the natural logarithm of a firm's total employees, debt ratio (denoted as DEBT_RATIO), calculated as total short-term and long-term debt as a percentage of total assets, and cash ratio (denoted as CASH_RATIO), calculated as the total cash balance as a percentage of current liabilities. Last, our study includes the book-to-market ratio, BTM.

Univariate Analyses

Table 2 reports summary statistics for our sample firms. The sales to total assets ratio, on average, is 23%, with a median of 20%. Firm size, which is the natural logarithm of total employees, has a mean of 9.34 and a median of 9.51. The average book-to-market ratio of firms in our sample is 0.73, with a median of 0.55, which is comparable to the sample used in Cohen and Zarowin (2010). The average debt ratio is 62%, while the average cash ratio is 43%.

On average, total advertising expenses constitute 0.17% of total assets. Advertising expenditures on TV, Internet, press, and outdoor are about 0.11%, 0.01%, 0.04%, and 0.02% of total assets, respectively. Advertising expenditures on TV constitute about 42% of a firm's total advertising expenditures. Advertising expenditures on press and outdoor each constitute more than 20% of total advertising expenditures.

Table 3 presents Pearson's (lower part of the diagonal) and Spearman's (top part of the diagonal) correlation matrix with pairwise correlations between our variables. We do not see any problems with strong collinearity among our explanatory variables used in our empirical models. In addition, the variance inflation factors in our subsequent econometric models range from 2.1 to 8.15 per variable; these are well within the acceptable ranges (under 10), eliminating concerns over multicollinearity.

Advertising Expenditures and Sales

We aim to identify the ceteris paribus impact of advertising expenditures on the current sales performance of the firms. We model the sales performance at t as a function of current advertising expenditures on media vehicles (i.e., TV, Internet, press, outdoor), firm characteristics, and macroeconomic indicators. We estimate a variety of regressions of the following form:

(1)

The dependent variable is , which is sales of firm i, in quarter (t), scaled by total assets. Advertising expenditures are aggregated over a quarter. The primary interest variables are advertising expenditures on four different media vehicles. We include , , , and , representing the advertising expenditures on TV, Internet, press, and outdoor media vehicles of firm i, quarter (t), scaled by total assets.

To control for macroeconomic factors that might influence a firm's sales performance, we include private consumption growth, , and the Consumer Price Index, in the United Kingdom.

Next, we include control variables to control firm characteristics that might drive sales. First, we include firm size, SIZE. While small firms often have behavioral advantages, in many activities they are at a disadvantage with respect to sales and costs. Sales and costs have effects of scale, scope, experience, and learning. Small firms generally produce small volumes (scale) of few products (scope). They have not been in business long and thereby have little benefit from economies of experience. They have limited capacity for the acquisition of knowledge (learning). Therefore, firm size plays a role in predicting a firm's sales performance, since firm size will influence scale, scope, experience, and learning within a firm's operations.

Last, we include DEBT_RATIO and CASH_RATIO. A firm's debt ratio may explain aggressive sales efforts, while the cash ratio would capture whether a firm is offering sales on account. Accrual-based accounting recognizes revenues when earned, regardless of when cash is received. Since the firms in our sample are large firms, we expect cash collection to be less than timely; therefore, we predict a negative association between sales and cash ratio. All of the variables are defined in detail in section 3.

In the model, captures the random error term. Our model includes industry and time fixed effects to control for other unobserved industry-specific and time-specific heterogeneity and to lessen remaining heteroskedasticity concerns. The identifying assumption for this model is that fixed effects pick up time-invariant sources of unobserved heterogeneity, which might induce an endogeneity bias, and that such a specification again places less burden on the control function approach because it serves only to condition out the time-varying sources of unobserved variation that correlate with the error term. To account for potential dependence among observations, all our statistical tests are based on clustered standard errors at the firm level (Petersen 2009).

Our focus in this study is the contemporaneous relationship between sales and advertising expenditures. Clearly, studying a contemporaneous phenomenon is an empirical issue. Our study's goal is not to capture brand awareness or other brand components. We examine the economic gains of advertising expenditures in the short term. There are two reasons why we focus on the contemporaneous relationship.

First, our primary goal is to capture revenue and advertising expenditure matching. Under IFRS, the current accounting treatment for advertising costs is to expense them immediately. The rationale behind this is that advertising, which is designed to elicit immediate sales, is to be expensed as incurred or when the advertisement first appears. Advertising that can be expected to result in benefits in future periods should be capitalized as an intangible asset, as long as the benefits are measurable and a company makes certain disclosures. The immediate expensing procedure has been followed by most companies for decades. Therefore, from an accounting perspective, the benefits resulting from the association between advertising costs and sales is assumed to be observed in the short term.

Second, on the basis of studies using data for shorter periods, Clarke concludes that “the duration of cumulative advertising effect on sales is between 3 and 15 months; thus this effect is a short-term (about a year or less) phenomenon” (1976). Several studies offer further support for this conclusion. Ashley et al. (1980), Boyd and Seldon (1990), and Seldon and Doroodian (1989) all offer evidence that the effect of advertising on sales is largely depreciated within a year (if not less). Further, Leone (1995) presents empirical support for the generalization that, on average, the effect of advertising on sales is largely depreciated within six to nine months. While there is consensus that the impact of advertising on sales is short-term, there is no consensus on when it disappears. We take a conservative approach and assume that the impact of advertising on sales dissipates in three months.

Table 4 presents multivariate evidence regarding the impact of total advertising expenditures on sales. Table 5 uses advertising expenditures on each media vehicle as the primary interest variables. In each table, to ease the interpretation of the coefficients, the first column presents standardized estimates. Therefore, each coefficient in the table describes the impact of a one standard deviation increase in the independent variable on sales. The second column reports the t-statistics.

Table 4 reveals that total advertising expenditures are positively associated with sales and significant at a 1% level. This is consistent with the literature that there is a positive association between sales and total advertising expenditures. Further, the positive sign indicates that the economic benefit of advertising is captured in the current sales performance.

In Table 5, we investigate the impact of each media outlet on sales. This model considers the possibility that the association between sales and advertising expenditures varies by the media vehicle. Table 5 presents five models. Model (1) includes all media vehicles (i.e., TV, INTERNET PRESS, OUTDOOR). Models (2) to (5) include each advertising media vehicle variable individually to see whether our results are driven by multicollinearity.

As discussed in section 2, we expect to find positive associations if there is proper matching between contemporaneous sales and advertising expenditures. On the other hand, we expect to find insignificant or negative associations if there is any evidence of imperfect matching between advertising expenditures and sales. Model (1) reveals that advertising expenditures on TV and outdoor are positively associated with sales, significant at 1%. On the other hand, advertising expenditures on Internet and press are both insignificant in our model.

Our results suggest that economic benefits gained from TV and outdoor advertising expenditures are observed in the current sales performance of a firm. However, current sales do not reflect the benefits gained by press and Internet advertising expenditures. The lack of results for press and Internet advertising expenditures does not necessarily suggest a lack of economic benefits for these media vehicles, as the economic benefits may be observed in the future. However, our results imply that the one-size fit matching principle may not be appropriate for advertising expenditures when different media vehicles are used. Models (2) to (5) reveal results consistent with our main model.

The coefficients of the control variables in Tables 4 and 5 are consistent with each other. As expected, changes in macroeconomic factors explain some of the variation in sales performance. We find PCG and CPI to be positively associated with sales, significant at a 5% level. We find the coefficient on SIZE to be significantly negative in both models. These results contradict our prediction that larger firms are likely to have more economies of experience, which can be used to enhance sales growth. In contrast, our results support the idea that smaller firms can react more quickly because they are less burdened by bureaucracy and have more freedom to enhance sales growth (Uhlaner et al. 2013). Finally, we report insignificant coefficients on DEBT_RATIO, and significantly negative coefficients on CASH_RATIO.

Growth Opportunities

One plausible interpretation for the positive association between advertising expenditures and sales is that the economic benefit of advertising is observed in the current sales performance, suggesting proper matching between revenues and expenses. On the other hand, a lack of results would imply that either advertising is an intangible asset for which benefits will be collected in the future or there are cost inefficiencies within the firm and there will be no economic benefits. In this section, we explore the conditions under which the associations we find are stronger or weaker.

If advertising activities create economic benefits in the current period, then the power of this association should vary when firms have more incentives to exhibit an aggressive sales effort. We investigate this issue by conditioning our analysis on the median book-to-market ratio. We use the book-to-market ratio as a proxy for growth opportunities at stake. A low book-to-market ratio indicates growth opportunities, suggesting that firms will have incentives for aggressive sales efforts. We split our sample into two based on whether a firm's book-to-market ratio is above the sample median for the quarter.

Table 6 shows the results of this analysis. Model (1) shows the results when a firm's book-to-market ratio is above the sample median for the quarter, while model (2) shows the results when the ratio is below the sample median for the quarter. Both models reveal that the coefficients on TV and outdoor advertising are significantly positive. These results are consistent with our main findings. In model (1) of Table 6, when the book-to-market ratio is above the median, we find that the coefficient on press advertising is positive, significant at a 10% level. On the other hand, in model (2), when the book-to-market ratio is below the median, the coefficient on press advertising is negative, significant at 5%. We do not find results for Internet advertising.

Our results suggest that when there are more growth opportunities, press advertising behaves like fixed costs, evidenced by the negative association. Fixed costs may indicate that there is an optimum level of spending. Once the optimum level is reached, the spending per unit will increase while the output level remains constant. This is plausible, because for growing firms, the economic benefits of advertising spending may take more time to be realized, while more mature firms are likely to realize these benefits more quickly given their established customer base.

Overall, the results in Table 6 supplement our findings in Table 5. Collectively, they show that there is a mismatch between advertising spending and contemporaneous sales. We show that this mismatch is driven by various advertising media vehicles. We also show that the growth opportunities exacerbate the existing mismatch between advertising expenditures and sales.

Industry Sectors and Size-Based Differences

Firms may pursue different strategies to compete in the marketplace. Prior research finds significant variation in advertising and promotion intensity across consumer versus industrial sector firms (Chauvin and Hirschey 1993; Graham and Frankenberger 2000). In certain sectors (e.g., consumer goods and consumer services), firms target large audiences and often rely on advertising in pursuing low costs and differentiation strategies for their brands and services.

Similarly, another consideration in the relevance of advertising expenditures is firm size. By virtue of their size, large firms may be better equipped than small firms to afford large outlays on advertising and may benefit from economies of scale and scope in advertising. As a result, advertising expenditures are more likely to be effective for relatively larger firms (Chauvin and Hirschey 1993; Hirschey and Spencer 1992; Shah et al. 2009).

More recently, Shah et al. (2021) find industry sector and size-based differences in the association between advertising expenditures and firms' future performance. Specifically, they find that advertising has a positive and statistically significant impact on firms' future earnings for both large-firm and small-firm subsamples. However, their findings show that the size of the coefficient of advertising for the large-firm subsample is larger than that of the small-firm subsample. Lastly, they show that advertising has a positive influence for firms in the consumer goods and consumer services sectors, while this association becomes insignificant for the industrial and technology sectors.

Following the findings in the literature, we revisit our primary findings. We split our sample based on industry sectors and size. First, we split our sample into two: firms in the consumer discretionary and consumer staples sectors and all other firms.⁸ We report our findings in Table 7. Model (1) shows the results for the consumer discretionary and consumer staples sectors, while model (2) shows the results for all other sectors. Both models reveal that the coefficients on TV are significantly positive, while the coefficients on PRESS are insignificant. Our results show a negative coefficient on INTERNET, while model (2) shows a positive coefficient on INTERNET. Both of these coefficients are significant at 10%. Only model (1) reveals a significantly positive coefficient on OUTDOOR. Our results suggest that firms, regardless of the sector, benefit from using TV as an advertising media vehicle, and the economic benefits can be observed in the current sales performance. However, the (potential) benefits and costs of expenditures on other media vehicles largely depend on the sector.

Next, we split our sample based on whether a firm's size is above the sample median for the quarter. We report our findings in Table 8. Model (1) shows the results when the SIZE variable is above the sample median for the quarter, while model (2) shows the results when the SIZE variable is below the sample median for the quarter. Our findings are consistent and similar to those reported in Table 6. To the extent that SIZE may capture growth opportunities, this is not surprising.

Collectively, Tables 7 and 8 show that while advertising expenditures bring economic benefits, revenue-expense matching is not fully achieved. To the extent that significantly negative associations reveal poor matching, we report that firm size and sectors play a role in how economic benefits from various advertising vehicles match with the current sales and advertising expenditures on media vehicles. At the very least, our results suggest that there is noise over the economic benefits gained. Our results support our main argument that revenue-expense matching varies by media vehicles.

Robustness Tests

In this section, we discuss three sensitivity analyses to ensure the robustness of our results. First, we repeat our main findings in Table 5, using one-period lagged independent variables. Using the one-period lagged independent variables addresses endogeneity concerns and rules out reverse causality (to some extent). We report our findings in Table 9. The coefficient estimates show that advertising expenditures on TV and outdoor are positively associated with sales, significant at 1%. On the other hand, advertising expenditures on Internet and press are both insignificant in our model. Even when we use lagged independent variables, our findings remain consistent.

Second, we consider the possibility that the economic benefits of press and outdoor advertising may be observed with delay, while TV and Internet advertising may be recognized more quickly. One possible explanation for this is that television and Internet ads directly reach the media user and command their attention by interrupting the audience's real-time activities or by taking advantage of their obligations to be in a given place. On the other hand, press and outdoor media vehicles are not as intrusive because users of these media can observe ads in their own time. They allow for highly self-selected ad experiences because media users have the liberty to skip ads that hold little or no interest for them or focus only on the ones that they have interest in or are already familiar with. We revise our primary model to include one-period lagged variables of press and outdoor advertising, while we use current values of TV and Internet advertising. We report our findings in Table 10. Our findings are virtually the same as those reported in Table 10.

Third, we run our model by only including nonzero advertising variables. We report our results in Table 11. Our sample size significantly drops, and we only have 461 observations. Despite the small sample size, our findings remain qualitatively similar. The coefficients on TV and OUTDOOR are positive, significant at 1%. The coefficient on INTERNET remains insignificant. These results are consistent with our earlier results. Surprisingly, we find the coefficient on PRESS to be significantly positive (at the 1% level). We note that firms that use all four media vehicles in the same quarter-year tend to have larger book-to-market ratios. Thus, these results support what we report in Table 6.