Skip to main content

A macroeconomic analysis reveals the benefits for consumers of digital advertising

October 11, 2021

authors:

  • Jeremy Greenwood, Professor of Economics, University of Pennsylvania
  • Yueyuan Ma, University of Pennsylvania
  • Mehmet Yörükoğlu, Professor of Economics, Koç University, Istanbul

This blog article is derived from the authors’ research paper titled ‘You Will’: A Macro-economic Analysis of Digital Advertising, a project of the Economics of Digital Services initiative led by Penn’s Center for Technology, Innovation and Competition (CTIC) and The Warren Center for Network & Data Services. CTIC and The Warren Center are grateful to the John S. and James L. Knight Foundation for its generous support of the EODS initiative.

 __________

Digital advertising rose with the advent of the information age. It is delivered through a vast quantity of free media goods that complement consumers’ leisure. It can be better targeted to consumers who are likely to buy the advertised products. This article reviews an information-based model used to measure the change in consumer welfare due to the expansion of digital advertising and concludes that the welfare gain is large. Non-college-educated consumers benefit more from an increase in leisure associated with the provision of leisure-enhancing free media goods, while college-educated consumers gain more from the increased intensity of price competition. Advertising costs are disproportionately financed by college-educated consumers.

The rise of digital advertising

Advertising has been around for eons, and the composition of advertising spending changed as new vehicles for delivering ads cropped up. After the arrival of the printing press came newspapers and then magazines.

Digital advertising began to emerge in the information age. A descendent of direct mail advertising is email marketing. This started in 1978 with an ad sent by Digital Equipment Corporation via the Arpanet to 400 DEC computer users. Online advertising started in the 1990s. The first clickable ad was on Hotwired.com in 1994, then the online version of Wired magazine—it was part of AT&T’s “You will” campaign that prognosticated about the future in the information age.

Now people are surrounded by digital media goods–Facebook, Google, Pandora, Twitter, Wikipedia, and apps for dating, dieting, exercising, playing guitar, meditation, and so on. Often these products are financed through advertising or the sale of marketing information for advertising purposes. Figure 1 shows that newspaper and magazine advertising saw a decline with the rise of TV, while TV advertising fell with the advent of the Internet. It is interesting to note that advertising’s share of GDP has remained roughly constant in the postwar period at around two percent.

This is a test

Figure 1:  Advertising in the United States, 1935-2019. Advertising has consistently amounted to approximately 2 percent of GDP. Its composition has seen dramatic changes, however, as new mediums for communicating emerged. Sources: Douglas Galbi and AdAge.

The media products that deliver ads are often provided for free. Think about the free goods just from Google: Chrome, Google Search, Google Maps, Gmail, Google Drive, YouTube, etc. Figure 2 shows the number of apps available in Google Play Store. In 2019 this was a whopping 2.8 million. Interestingly, consumers spend little for these products. Less than 14 percent of Google users spent more than $10 per digital media in the Google Play Store, as the left panel illustrates. Since media goods are not sold, they do not directly show up in the national income accounts. Additionally, advertising expenditure is deducted off firms’ profits and consequently does not show up as a final expenditure in the national income accounts in the same way as physical investment spending does. So, advertising is not precisely measured in GDP. Also, even if GDP is adjusted for such things, GDP and the well-being of consumers are not the same thing.

Figure 2A: Applications in the Google Play Store, 2009-2019. Source: statista.

Figure 2B: Money spent by U.S. consumers on Google digital media products in 2017, presented in cumulative distribution form. Source: statista. 

Media goods largely complement and enrich consumers’ leisure time (Greenwood and Vandenbrouke 2008 and Kopecky 2011). Advertising closely tracks consumers’ uses of time, as shown by the close relationship between spending by advertisers and the amount of time that consumers spend on various mediums (Figure 3). The welfare improvement from the higher quality of leisure has not received much attention by economists.

Figure 3:  U.S. advertising spending vs time spent by consumers by medium, 2018. Source: statista.

A theory of information-based advertising

To measure the contribution of advertising to the consumer well-being and to explore potential distribution effects, an information-based model is constructed where firms must advertise to sell goods. (See Greenwood et al. 2021 for more detail.) Firms send ads to consumers to inform them about the prices, while consumers decide whether to purchase products from firms based on the advertised prices.

Prices can differ across firms for the identical product due to information frictions, or the fact that consumers are not aware of all the different prices charged by firms. Consumers receive a random number of ads and then buy the product at the lowest price in the ads they receive, given that a purchase lies within their budget and maximizes their well-being. Suppose that there are two types of consumers differing in their income and therefore differing in the highest price they are willing to pay–call this their choke price.

Assume that ads cannot be directed to a particular type of consumer, as with undirected advertising. Figure 4 shows an example of the advertising and purchasing process. Consumer 1 receives five ads with prices ranging from 1.01 to 1.12 as seen in the green lines. As the lowest price (1.01) is below her choke price (1.06, as indicated by the dashed blue line), she will purchase the good at price 1.01. Similarly, consumer 2 will purchase the good at price 1.03. Consumer 3 does not receive any ad so will not buy the good. Consumer 4 receives two ads, but the prices in both ads exceed her willingness to pay (or her choke price). Therefore, she won’t buy the good.

Figure 4:  An example illustrating the advertising and purchasing process.

The advertised and transacted price distributions for this example are shown in Figure 5. Observe that 50 percent of transactions are consummated at a price less than or equal to 1.01, and 100 percent are at a price less than or equal to 1.03. Compare this to the fact that only 10 percent of advertised prices are at 1.01 or below, and only 30 percent are less than or equal to 1.03. That is, the transacted price distribution lies to the left of the advertised one, illustrating that when consumers purchase products, they do so at the lowest price in their information set.

Figure 5:  Advertised and transacted price distributions. The diagram shows on the vertical axis the fraction of either advertised or transacted prices that are less than or equal to (≤) the price indicated on the horizontal axis.

Simulating information-based advertising

A computer model along the above lines is simulated in Greenwood et al. (2021). Advertising is executed via the provision of free media goods. These media goods increase the enjoyment from leisure. Consumers choose which goods to consume based on the ads they receive and how much time to spend on leisure and work.

There are two modes of advertising: traditional and digital. Digital advertising costs less than traditional advertising. It also delivers more free media goods per message received by consumers. Firms choose how many ads of each mode to send out and the price they charge in the ads. In the benchmark Greenwood et al. (2021) framework, ads cannot be directed toward consumers based on their income. That is, advertising is undirected. (The case of directed advertising will be covered after this.)

The Greenwood et al. (2021) computer model is matched up with some stylized facts from the U.S. data: the average price markup over (marginal) production cost, the ratio of advertising expenses to consumption expenditure, the click-through rate for digital advertising, the growth in the ratio of spending on digital advertising relative to traditional advertising, the ratio of college to non-college earnings, and the rise in the time spent on leisure that was connected with media for both non-college- and college-educated people (the counterparts of low-income and high-income consumers). Interestingly, the framework is consistent with the recent decrease in hours worked for the non-college educated relative to the college educated.

The equilibrium distribution of prices in the model is displayed in Figure 6. There is a discrete price jump in prices as firms need to recover profits from a drop in potential consumers because low-income consumers will not purchase a good at a price higher than p(τ). This happens because ads are undirected (i.e., ads not directed based on income).

Figure 6:  The cumulative distribution functions for both advertised prices, P(p), and transacted prices that obtain in the calibrated benchmark equilibrium for 2018. Source: Greenwood et al. (2021).

The advertising equilibrium is not efficient. First, free media goods are underprovided. When firms make their advertising decisions, they do not consider the enjoyment that consumers realize from the free media goods. All firms care about is their profits. Second, some advertising is wasteful in the sense that ads are sent to consumers who cannot afford to purchase the good at the posted price. A second-best tax-cum-subsidy policy that overcomes these inefficiencies is developed. Part of this policy involves subsidizing the provision of media goods and taxing advertisements.

The provision of free media goods via advertising is connected to a large increase in welfare or in the well-being of consumers. As displayed in Table 1, the increasing share of digital advertising from 2003 to 2018 leads to an increase in the well-being (or welfare) for the non-college and college educated by 2.5 and 2.7 percent, respectively, measured in terms of consumption. The boost in welfare (or consumer well-being) is measured by the equivalent variation (EV), which is the amount of extra consumption that you would have to give a person in 2003 (when there was no digital advertising) to make them as well off as in 2018 (when there was digital advertising). For both groups of individuals, there is a significant increase in welfare due to the expansion of free media goods connected with digital advertising. The non-college educated realize a significant gain in welfare from their rise in leisure. The welfare gain from the increase in leisure is mostly offset by a decline in non-college educated consumption. The college educated enjoy a smaller improvement in welfare from the rise in leisure. Their decline in consumption is negligible. The reduced work effort by the college educated is counteracted by a reduction in prices at the upper end of the price distribution stimulated by increased competition. The tax-cum-subsidy policy (not shown) that overcomes the inefficiencies associated with advertising has a small impact on the well-being of consumers, which is swamped by the benefit from the free provision of media goods. Since the gain in consumer well-being from the tax-cum-subsidy policy is relatively small, this implies that amount of advertising observed in the United States might be close to being socially optimal.

Table 1:  The welfare gains from the expansion of free media goods arising from the advent of digital advertising. The welfare gains are measured in terms of consumption. These welfare gains are decomposed into the effects that digital advertising had on regular consumption, media goods provision, and leisure. Source: Greenwood et al. (2021).

So who is implicitly paying for the provision of free media goods? It is college-educated consumers. They bear 73 percent of the total advertising cost while representing 35 percent of the population. The college educated pay more than their share because they buy goods at higher prices where the markups (price less production cost) are larger. The high-price markets are thinner because fewer people can afford to buy goods in these markets. The competitive environment renders expected profits the same across firms. Therefore, goods sold on thinner markets must bring higher net revenue (markups) to firms.[i] The relationship between the probability of a sale (thickness of the market) and the markup is plotted in the right panel of

Figure 7A: Because markups cover the costs of advertising, this implies that consumers who buy goods in the thinner markets pay more for the advertising costs. This panel provides further illustration. It shows that the fraction of a firm’s sales revenue made up by advertising costs is an increasing function of their selling price, suggesting that consumers who buy goods at a higher price contribute more to the advertising costs thus stimulating the provision of media goods.

Figure 7B: The relationship between the probability of a sale at price p and a firm’s price markup, p-γ, where γ represents the marginal cost of producing the good. Source: Greenwood et al. (2021).

Advertisers now collect vast amounts of information on consumers. In an extension of the model, an economy is considered where advertising can be directed only toward those consumers who will potentially buy the product but that anyone can enjoy the free media goods used to disseminate the ads. The price distribution no longer has a discrete jump, as Figure 8 shows. With directed advertising, the college educated transact at a much more favorable price distribution, as shown by the leftward shift. This occurs because the cost of reaching these consumers is now lower because advertising can be directed. Thus, there is less wasteful advertising expenditure.

Figure 8: The cumulative distribution functions for transacted prices under both directed and undirected advertising. The college educated purchase from a much better price distribution when advertising is directed. For the non-college educated the two prices distributions are virtually identical. Source: Greenwood et al. (2021).

The competitive equilibrium with directed advertising is compared with the undirected case for the year 2018. As displayed in Table 2, there is a slightly smaller supply of free media goods in the world with directed advertising because there is less advertising. This (negligibly) hurts those consumers who would not have bought high-priced products in the economy with undirected advertising. It benefits those consumers who bought high-priced goods in the economy with undirected advertising because now there is more price competition, which results in lower prices and increased consumption.

Table 2: The welfare gains from a move toward directed advertising. Source: Greenwood et al. (2021).

A hybrid model is entertained where all digital advertising is directed and traditional advertising is undirected. After recalibrating, the hybrid model delivers similar results to the baseline one, with just undirected advertising, when digital advertising increases its share in total advertising spending over time. Non-college-educated workers enjoy a welfare gain between 2003 and 2018 of 2.0 percent while for the college educated the percent is 2.8. Compared with the baseline model, the skilled fare relatively better than the unskilled with the shift toward directed advertising.

Closing

In sum, digital advertising is less expensive for firms than traditional advertising and can be better targeted toward consumers who are interested in buying the advertised product or service. It provides benefits to consumers in terms of free media goods and increased price competition. The cost of digital advertising is disproportionately born by consumers in the upper end of the income distribution.

References

Greenwood, Jeremy, Yueyuan Ma, and Mehmet Yorukoglu. 2021 “‘You Will:’ A Macroeconomic Analysis of Digital Advertising.” Working Paper No. w28537, National Bureau of Economic Research. https://www.nber.org/papers/w28537

Greenwood, Jeremy and Guillaume Vandenbroucke. 2008. “Hours Worked (Long-Run Trends).” The New Palgrave Dictionary of Economics, v. 4, 2nd edition, edited by Lawrence E. Blume and Steven N. Durlauf, (New York, N.Y.: Palgrave Macmillan): 75-81. https://ideas.repec.org/p/eag/rereps/10.html

Kopecky, Karen A. 2011. “The Trend in Retirement.” International Economic Review 52 (2): 287-316. https://www.jstor.org/stable/23016634

[i] Expected profits across firms must be equalized in a competitive market. For two firms, 1 and 2, the following holds: , where is the probability that an ad with price can generate a sale; is the marginal cost of producing a good, implying that is the price markup; is the number of ads sent out by a firm; and is the cost of these ads. It must then transpire that expected net revenue across the two firms must also be equalized: . This occurs because the firm with the highest net revenue can always pick the same level of advertising as the other firm. This would lead to the firm with the highest net revenue having higher profits, contradicting the competitive market assumption. Therefore, generically, = constant, where constant is the common level of expected net revenue across all firms. Thus, the probability of a sale and price markups are inversely related; i.e., thin markets have high markups.