Consumer Privacy Choices and (Un)Targeted Advertising Along the Purchase Journey

Abstract

Advertising to a consumer provides potentially useful information to the consumer and moves them along the purchase journey, and tracking the consumer's online activities enables an advertiser to infer the consumer's purchase journey state and target repeat ads accordingly. However, many consumers dislike being tracked, and, furthermore, repeat advertising may lead to ad wearout. The authors develop a model with consumers, an advertiser, and an ad network to investigate, under the preceding considerations, the impact of regulations that endow consumers with the choice to opt in to or out of online tracking. The authors find that, if ad effectiveness is intermediate, opting in to tracking decreases ad repetition; otherwise, opting in increases ad repetition. To make an opt-in decision, a consumer weighs the cost of ad wearout from repeat ads against the benefit of obtaining potentially relevant product information from them, and the consumer opts in to tracking if either ad effectiveness is intermediate or sensitivity to ad wearout is low. This opt-in pattern creates counterintuitive implications; for instance, higher ad effectiveness, even though it implies higher ad valuation for the advertiser, may reduce repeat ads and the ad network's profit. Under regulation that requires consumer consent for tracking, the results shed light on when and why consumers give such consent, and provide useful insights for practitioners and policy makers.

Keywords

consumer purchase journey ad wearout consumer tracking privacy regulations

Before making a purchase decision, a consumer typically goes through a “purchase journey” (e.g., represented by the canonical awareness–interest–desire–action “purchase funnel” model; Kotler and Keller 2012; Lavidge and Steiner 1961). Firms use advertising to convey product information to a consumer to move them along their journey (Kotler and Keller 2012; Todri, Ghose, and Singh 2020). Tracking of consumers’ online activities, which refers to the collection and analysis of consumers’ activity data on particular websites and across websites, has enabled advertisers to make real-time inferences about consumers’ purchase journey states, based on which firms can target ads (Hoban and Bucklin 2015; Johnson, Lewis, and Nubbemeyer 2017; Lambrecht and Tucker 2013; Sahni, Narayanan, and Kalyanam 2019). However, showing consumers multiple ads can also lead to a negative consumer response, in a phenomenon known as “ad wearout” or “ad annoyance” (Aaker and Bruzzone 1985; Campbell and Keller 2003; Chae, Bruno, and Feinberg 2019; Inskin Media 2019; Tourn 2019; Todri, Ghose, and Singh 2020). Furthermore, widespread adoption of consumer tracking has deepened consumers’ concerns about their online privacy (Auxier et al. 2019); partially in response to this, regulations have been passed to require firms to obtain explicit consent from consumers to use their data (e.g., the European Union's General Data Protection Regulation [GDPR] in 2018 and the California Consumer Privacy Act in 2020), and browsers (Grustniy 2022) and app stores (Apple 2021) provide options to consumers to opt out of third-party tracking.

The impact of privacy regulations on the advertising industry is a topic of ongoing debate among practitioners, academics, and policy makers. On the one hand, regulations will limit advertisers’ targeting capabilities if consumers opt out of tracking. On the other hand, there is evidence suggesting that despite consumers’ stated aversion toward tracking, they still opt in at substantial rates to being tracked; for example, Godinho de Matos and Adjerid (2022) find post-GDPR opt-in rates of approximately 64% for consumers with an existing relationship with a firm, and after Apple introduced its App Tracking Transparency policy, 40%–50% of consumers still opted in to being tracked (AppsFlyer 2021).

In this article, we study consumers’ opt-in/out choices under regulation that requires their consent for using their data, and how this affects their exposure to marketing information and their purchase incidence. We also study the implications for the ad ecosystem; for instance, we examine the repeat advertising strategy of the advertiser, and the revenue generated from ads for the ad network. For our study, we develop a multiperiod game theory model of informational advertising with consumers who move through their purchase journey stages, an advertiser selling a product, and an ad network enabling the advertiser to advertise to the consumers. Our approach captures the nuanced strategic interactions between the key players in the ad ecosystem (advertisers, ad networks, and consumers), which allows us to evaluate the implications of privacy regulations.

A policy that requires consumer consent before using data allows consumers to make a trade-off between avoiding repetitive ads and learning product information. Although the common debate focuses on the consumer's trade-off in isolation without regard to the equilibrium reactions in the advertising market (by the advertisers and the ad network), we explain how and under what conditions endowing consumers’ privacy choices can lead to more effective targeting of ads (with less repetition) and possibly less consumption by consumers. Specifically, when consumers choose to opt in, an advertiser can infer their level of product interest generated by the impact of past ads and avoid sending repetitive ads if the marginal impact on purchase likelihood will be low. A key condition of this result is that ads have intermediate effectiveness in terms of drawing consumers’ attention. Alternatively, when ads are extremely effective at drawing consumers’ attention, advertisers employ a spamlike approach, sending repeat ads to opt-in consumers who have not yet purchased the product. In contrast, when ads have low effectiveness, advertisers show few repeat ads irrespective of consumers’ privacy choices.

In accordance with this, we find that consumers opt in to tracking either if the ad's potential to attract attention is intermediate or if consumers have low sensitivity to ad wearout (such that repeat ads do not significantly annoy them and they prefer to benefit from the informational value of ads). Overall, endowing consumers with privacy choices weakly increases their net surplus but weakly decreases their ad exposure and product consumption. Interestingly, these findings imply that the ad network's profit and the advertiser's sales may decrease with higher ad effectiveness; this is because when ad effectiveness increases from a low value, consumers opt out of tracking, which leads to fewer ads and reduced product sales.

Although in the main model we focus on the instrumental implications of consumers’ privacy choices (referring to the indirect impact of allowing tracking on utility due to the actions of other agents, in this case the advertiser and the ad network), in an extension to the main model we also incorporate consumers’ heterogeneous intrinsic valuations of privacy (referring to the value of maintaining privacy for its own sake; Becker 1980; Farrell 2012; Lin 2022; Posner 1981). Specifically, we assume that consumers incur an intrinsic disutility from the very act of allowing the ad network to track their online behaviors, as doing so compromises their online privacy, but they may still do so if the expected instrumental benefits are sufficiently large. This creates discrepancies in privacy choices across consumers with low versus high intrinsic values of privacy; however, the insights from the main model continue to hold qualitatively.

In another extension, we allow the ad network to infer not only the purchase journey state of opt-in consumers but also their product category interest. Interestingly, we find that product preference inference is a double blessing for opt-in consumers: opt-in consumers not only obtain higher expected product surplus but also see fewer repeat ads. Intuitively, higher product category matching means that opt-in consumers are more likely to attend to ads and move down the purchase journey, which, combined with funnel state inference, lowers expected ad repetition.

Overall, our research analyzes the impact of privacy regulations on consumers who balance the double-edged nature of repeat ads (information vs. annoyance), as well as on other players in the advertising ecosystem. We derive which privacy choices consumers make and why,¹ how these choices impact their ad exposure and product consumption, and how they influence advertisers’ ad strategies. Finally, we provide useful insights for regulators who balance the implications of privacy regulations for multiple participants in the advertising ecosystem (i.e., consumers, advertisers, and ad networks).

Our research relates to the literature on targeted advertising and online privacy. Extant literature on targeted advertising investigates various implications of targeting. For example, it examines the impact of targeting on ad supply, ad prices, ad strategies, ad intensity, and adoption of ad avoidance tools (Athey and Gans 2010; Bergemann and Bonatti 2011; Esteban, Gil, and Hernandez 2001; Gritckevich, Katona, and Sarvary 2022; Iyer, Soberman, and Miguel Villas-Boas 2005; Johnson 2013; Shen and Villas-Boas 2018; Shin and Yu 2021). We extend the existing literature on dynamic, targeted advertising in a novel and important way by modeling advertising along the consumer purchase journey and investigating an understudied link between cross-period ads. This allows us to study funnel-state-dependent ad effects and ad strategies in a setting in which ad exposures endogenously create interim funnel-state heterogeneity across consumers.

We also contribute to the growing literature on consumer privacy in an online setting. A closely related study to ours is that of D’Annunzio and Russo (2020), who similarly examine a setting in which consumers can endogenously decide whether to be tracked. However, our research is different in several important ways. First, D’Annunzio and Russo show that tracking increases the number of ads if there are sufficiently many “multi-homers” (i.e., consumers who visit multiple publishers). This result is primarily based on the assumption that “tracking increases the expected revenue from ads that hit each multi-homer … as well as the value of additional impressions on [multi-homers]” (D’Annunzio and Russo 2020, p. 5044). In contrast, our research demonstrates that this is not necessarily true under purchase funnel considerations. Specifically, purchase funnel tracking may reduce the intensity of ads as lower-funnel-state consumers can be better identified and excluded from targeting. Such divergent outcomes from those of D’Annunzio and Russo highlight the significance of modeling the consumers’ purchase journey. Second, D’Annunzio and Russo assume that the number of ads a consumer is exposed to is independent of the consumer’s privacy choice and that ads always reduce consumer utility. In contrast, we explicitly model the changes in advertising intensity due to consumer trackability, which critically influence a consumer's utility, and allow repeat ads to provide informational value to consumers.

Another closely related study is Choi, Jerath, and Sarvary (2020), in which we build a generic model of price discrimination with advertising that does not consider the purchase funnel. Even though in both studies consumers make an opt-in/out decision for being tracked, in the 2020 study this decision influences the firms’ postadvertising price competition among multiple advertisers. In contrast, in the current study, consumers’ opt-in/out decisions influence whether firms can track consumers’ purchase funnel states and whether firms show repeat advertising. Due to these differences, the two studies focus on different aspects of marketing and provide different results. For instance, our 2020 study shows that consumers opt out of tracking when their type can be inferred accurately because they anticipate severe price discrimination. In contrast, the current study shows that consumers may opt in under accurate funnel state inference in anticipation of reduced ad repetition due to more efficient targeting.

A number of other studies are related to consumer privacy and how firms may use revealed consumer information. Several studies focus on the idea that if firms have greater information on consumers, then they can price discriminate better. To mitigate this, consumers can make implicit privacy decisions, such as strategically timing their purchases (Taylor 2004; Villas-Boas 2004), or explicit privacy decisions such as (dis)allowing sharing of their data (Anderson, Baik, and Larson 2019; Conitzer, Taylor, and Wagman 2012; Ichihashi 2020; Ke and Sudhir 2020; Montes, Sand-Zantman, and Valletti 2019). In our research, we take pricing as exogenous and focus on targeted or untargeted repeat advertising strategy based on consumers opting in to or out of tracking, respectively. De Cornière and De Nijs (2016) study a setting in which an ad network may disclose consumer information to advertisers; this is different from our study, where the consumers control the sharing of their personal information. Finally, as we do not consider advertiser competition in our study, the mechanisms driving our results are orthogonal to those studied in some previous research on privacy, such as market thickness (Rafieian and Yoganarasimhan 2021) and market structure (Campbell, Goldfarb, and Tucker 2015). We refer the reader to Choi and Jerath (2022) for more comprehensive discussions on these and other studies related to the interplay between consumer privacy and online advertising.

The rest of the article is organized as follows. In the next section, we describe our model. Following this, we present our analysis and results on the consumers’ opt-in versus opt-out choices under different conditions, and the impact on the advertising ecosystem. We then analyze various extensions that assess the robustness of the main insights and generate new insights. Finally, we conclude with a discussion. Proofs for the main results are provided in the Appendix, while additional analyses are provided in the Web Appendix.

Model

In this section, we describe our model. The model consists of three entities—a consumer, an advertiser, and an ad network—interacting over two time periods. In each time period, a consumer visits a web page and the ad network enables an advertiser to show an ad to the consumer. We next describe the components of the model in detail.

Consumer

We consider a single representative consumer who visits two web pages, one in each Period $t \in {1, 2}$ .² Both web pages are covered by the ad network. The consumer is exposed to at most one ad for the advertiser's product per page visit. Ads may provide useful information to the consumer and thereby move the consumer along the purchase journey.

We consider a stylized purchase journey, which we interchangeably refer to as the purchase funnel, consisting of two states: top funnel state (T-state) and bottom funnel state (B-state). A funnel state is defined by the level of product information a consumer possesses. A consumer in T-state does not know how well the advertiser’s product fits their preferences; that is, the consumer lacks product match information. In contrast, a consumer in B-state possesses product match information that allows them to discern product fit.

We assume that ads influence a consumer's progression down the purchase funnel. An ad may or may not attract a consumer's attention; specifically, an ad draws the consumer's attention with probability (hereafter, w.p.) $μ \in [0, 1]$ . Also, the consumer may or may not be interested in the category of the advertised product; specifically, a consumer is interested in the product category w.p. $λ \in [0, 1]$ . Therefore, a T-state consumer obtains product match information from an advertiser's ad and moves down to B-state if the ad attracts the consumer’s attention (which happens w.p. μ) and if the consumer is interested in the advertised product category (which happens independently w.p. λ).³

Note that μ denotes the ad's potential to attract consumer attention (which may include factors such as prominence of the ad's location and appeal of ad design characteristics; Pieters, Wedel, and Zhang 2007; Schwartz, Bradlow, and Fader 2017; Wedel and Pieters 2000; Zhang, Wedel, and Pieters 2009); for simplicity, we refer to μ as ad effectiveness. In contrast, λ captures the ex ante probability that a consumer will be interested in the category of the product being advertised, which is realized by a consumer after seeing the advertiser's ad.⁴ Ads have no effect on a B-state consumer, as this consumer already has product match information, or on a T-state consumer who, after seeing previous ads, is not interested in the product category.

The following summarizes the effect of the first ad exposure on a T-state consumer, in particular, how the ad probabilistically redistributes a T-state consumer along the purchase funnel and creates interim heterogeneity (see Figure 1). W.p. μ, an ad shown to a T-state consumer draws the consumer's attention. If the consumer is interested in the product category (w.p. λ), the consumer obtains product match information and moves down to B-state. If the consumer is not interested in the product category (w.p. 1 − λ), the consumer does not obtain (alternatively, ignores) product match information and stays in T-state. W.p. 1 − μ, an ad shown to a T-state consumer fails to attract the consumer’s attention; in this case, no product match information is conveyed, and the consumer stays in T-state. Therefore, the first ad shown to a T-state consumer creates three consumer “segments”: a consumer who moves down to B-state after obtaining product match information, a consumer who remains in T-state because they are uninterested in the product category, and a consumer who remains in T-state because they did not see the ad. As we discuss subsequently, the advertiser's successive ad strategies in Period 2 will have different effects on each of these segments.

Figure 1.

Effect of First Ad on a Consumer in the Top Funnel State.

The consumer makes a product purchase decision. If the product category does not match (which occurs w.p. 1 − λ), the consumer does not buy the product. If the product category matches (which occurs w.p. λ), the consumer purchases the product if the product match is sufficiently good.⁵ A consumer who obtains product match information from an ad realizes a product match value V ∼ U[0,1]. A B-state consumer who has product match information purchases if and only if v ≥ p, where p is the exogenous product price.⁶ The consumer's expected product surplus is $\int_{p}^{1} (v - p) dv = \frac{1}{2} (1 - p)^{2}$ . We assume that $E [v] - p \leq 0$ , that is, $p \geq \frac{1}{2}$ , such that even if the consumer is interested in the advertised product category, they do not purchase the product prior to obtaining product match information.

The consumer also makes a privacy choice of whether or not to allow the ad network to track their web page visits and purchase activity, from which the purchase journey stage of the consumer can be inferred. If a consumer opts in to tracking, then the ad network can infer their purchase journey stage and sell targeted ads based on the inferred state. If the consumer opts out of tracking, the ad network cannot infer their purchase journey stage and sells untargeted ads. We denote the consumer's privacy decision by $x \in {0, 1}$ , where x = 1 if the consumer opts in to tracking, and x = 0 if the consumer opts out of tracking. We refer to this choice interchangeably as the consumer's privacy choice, tracking choice, or opt-in choice.

When making their privacy choice, the consumer considers how opting in to or out of tracking affects (1) their expected product surplus, and (2) the intensity of repeat ads that they expect to see. As discussed previously, ads benefit the consumer in expectation by providing product match information that aids their product purchase decision. However, we assume that if the consumer sees an ad that they have previously seen, then they experience ad wearout, which causes disutility. Overall, a consumer's expected utility is

E [U (x)] = E [S (x)] - η E [Q (x)],

(1)

where S(x) is their product surplus, Q(x) is the total number of repeat ads they see, and η > 0 is the ad wearout parameter (i.e., the disutility the consumer incurs per unit of ad repetition).⁷^,⁸

Depending on the consumer's privacy choice x, different numbers of ads may be shown to the consumer such that the ad repetition intensity, Q(x), potentially changes with x. This in turn affects $E [S (x)]$ through the effects of ad repetition on the consumer's product match knowledge. In summary, a consumer's opt-in decision depends on the intensity of repeat ads and product surplus the consumer expects as a result of the opt-in decision.⁹

Advertiser

Depending on whether the consumer can be tracked, and what the ad network sells, the advertiser can buy different types of ads. If tracking is not possible, then the advertiser can only buy untargeted impressions (e.g., ads displayed to all website visitors independent of their browsing histories). In particular, even if an ad is shown in Period 1 and probabilistically redistributes the consumer along the funnel, the advertiser cannot target ads in Period 2 on the basis of the consumer's inferred funnel state and must make ad purchase decisions based on the expected distribution of the consumer in the different stages of the funnel based on the Period 1 ad.

However, if tracking is possible, the ad network can sell and the advertiser can buy ad impressions on the basis of inferences about the consumer's funnel state. The advertiser can specify the target audience such that ads are shown only to the consumer who meets some prespecified criteria that correlate with their funnel state. In each period, the advertiser decides, conditional on the consumer's inferred funnel state, the bid amount for each ad impression. We assume funnel states to be perfectly observable for an opt-in consumer; we relax this assumption in an extension presented in the Web Appendix and show that the qualitative insights continue to hold. Note that because an ad that is served to a consumer only attracts the consumer's attention w.p. μ, the number of ads shown by the advertiser differs from the number of ads seen by the consumer.

We normalize the advertiser's marginal cost of production of the product to zero. The advertiser maximizes its net profit across two periods, which is the difference of its product sales margin and advertising costs.

Ad Network

The ad network sells ad impressions to advertisers via second-price auctions.¹⁰^,¹¹ It sets a reserve price in each Period $t \in {1, 2}$ for each of the different types of ad impressions sold (e.g., ads targeted to a T-state consumer).¹² The ad network maximizes its revenue from ad sales across two periods. We assume that if an ad slot is not sold to the advertiser, then it is offered in a different channel to a mass of advertisers (e.g., via ad exchanges) where it is sold at price b. We hereinafter refer to the focal advertiser, who bids prior to the remnant inventory sale, simply as “the advertiser,” and the focal advertiser's ad simply as “the ad.”¹³

Timing

The timing of the game is as follows.

Period 0: The consumer decides whether or not to opt in to being tracked.

Period 1: The ad network sets the reserve price for the Period 1 ad for the opt-in consumer and for the opt-out consumer. The advertiser bids for ad impression. The ad is shown and probabilistically attracts the attention of the consumer, who then realizes product category match, according to which the consumer transitions through the funnel or not. Product purchases are realized.

Period 2: The ad network sets the reserve price for a targeted ad (based on inferred funnel states) for the opt-in consumer, and for an untargeted ad for the opt-out consumer. The advertiser bids for ad impression. The ad is shown and probabilistically attracts the attention of the consumer, who then realizes product category match, according to which the consumer transitions through the funnel or not. A consumer who sees ads in both periods suffers ad wearout.

We solve for the subgame-perfect equilibrium of the game.

Analysis

In this section, first, we describe the impact of the Period 1 ad. Next, we analyze the subgame for the opt-out consumer and then the subgame for the opt-in consumer. In each subgame, we describe the advertiser's and the ad network's strategies and then characterize $E [S (.)]$ , the expected product surplus, and $E [Q (\cdot)]$ , the expected intensity of repeat ads. After solving the subgames for the opt-out and opt-in consumer, we analyze the consumer's opt-in choice. Recall from Equation 1 that $E [S (\cdot)]$ and $E [Q (\cdot)]$ jointly determine the consumer's opt-in choice in Period 0. We assume that $\underline{b}$ is not too large; that is,

\underline{b} \leq \frac{(1 - μ) λ μ}{1 - μ λ} (1 - p) p .

(2)

This is a technical assumption that ensures that the consumer's opt-in choice materially impacts the consumer’s utility.¹⁴

Impact of Period 1 Ad

Before an ad is shown in Period 1, the consumer is in T-state. However, showing the Period 1 ad probabilistically distributes the consumer across the funnel and also causes the consumer to purchase with some probability. Note that there is no wearout from the Period 1 ad as it is not a repeat ad.

The Period 1 ad informs the T-state consumer w.p. μλ about product match; the informed consumer moves down to B-state (because an ad is effective in capturing a consumer's attention w.p. μ, and a consumer is interested in the advertised product category w.p. λ and thus processes the ad's information). W.p. 1 − p, a B-state consumer realizes product match value greater than p and purchases, whereas w.p. p, the consumer realizes a match value less than p and hence does not purchase.

The consumer remains in T-state w.p. 1 − μλ after the Period 1 ad is shown. This is because an ad fails to capture a consumer's attention w.p. 1 − μ, and the consumer sees the ad but ignores its informational content w.p. μ(1 − λ) because they are not interested in the product category. In other words, the Period 1 ad creates interim heterogeneity, as described in the following assertion.

Assertion 1.

Before the Period 1 ad is shown, the consumer is in T-state. After the Period 1 ad is shown, and before Period 2 starts:

1.W.p. 1 − μλ, the consumer is in T-state.

a.W.p. 1 − μ, the Period 1 ad does not attract the T-state consumer's attention.

b.W.p. μ(1 − λ), the Period 1 ad attracts the T-state consumer's attention, but the consumer is not interested in the category.

2.W.p. μλ, the consumer is in B-state. The Period 1 ad attracts the B-state consumer's attention, and the consumer is interested in the product category.

a.W.p. μλp, the B-state consumer realizes a low match value and therefore does not purchase.

b.W.p. μλ(1 − p), the B-state consumer realizes a high match value and therefore purchases.

The advertiser can benefit from showing a repeat ad in Period 2 only to the consumer in Segment 1(a) in Assertion 1 because the Period 1 ad did not attract the consumer's attention; however, w.p. 1 − λ, this consumer will turn out not to be interested in the product category. Also, this consumer will not experience ad wearout from being served a repeat ad in Period 2 because the Period 1 ad did not attract their attention in the first place. The consumer in Segment 1(b) in Assertion 1 will experience wearout from being served a repeat ad in Period 2 if it attracts their attention because the Period 1 ad attracted their attention and they were not interested in the product category. The consumer in Segments 2(a) and 2(b) in Assertion 1 will not be affected informationally by a repeat ad in Period 2, as they already have the product match information that an ad can supply. However, they do experience wearout from a repeat ad because they had already paid attention to, and processed, the first ad.

The interim funnel state heterogeneity created by the Period 1 ad is important for both the advertiser and the consumer. From the advertiser's perspective, funnel state heterogeneity influences repeat ad strategy because the effectiveness of repeat ads depends on the target consumer's funnel state. From the consumer's perspective, their funnel state affects whether they are worn out by repeat ads or not; therefore, the consumer will consider expected interim funnel state heterogeneity in making their opt-in choice. Since opting in to tracking allows for more efficient targeting of repeat ads along the purchase funnel, one may intuit that both the advertiser and consumers would benefit from tracking. Interestingly, we subsequently show that this is not always the case; under certain conditions, the consumer opts out of tracking, which decreases the advertiser's targeting efficiency for repeat ads.¹⁵

The Period 1 analysis for the opt-out and opt-in consumer is the same, and as described previously. We proceed to the Period 2 analysis for the opt-out and opt-in consumer. For an opt-in consumer (i.e., one who has opted in to being tracked), the advertiser and ad network can observe, at the start of Period 2, whether the consumer has made a purchase or not, and if not then whether the consumer is in T-state or B-state. For an opt-out consumer (i.e., one who has opted out of being tracked), the advertiser and ad network cannot observe anything about the consumer’s purchase or funnel state.

Analysis for the Opt-Out Consumer

Consider the Period 2 subgame for the opt-out consumer in which the Period 1 ad is shown. The distribution of the consumer at the start of Period 2 is as in Assertion 1; however, the ad network cannot track the opt-out consumer's funnel state. Therefore, in Period 2, it sells untargeted ads based on the expected interim heterogeneity generated by the Period 1 ad. Since the Period 2 ad may only impact the T-state consumer who is interested in the product category, the advertiser's valuation, and hence its bid, for the Period 2 ad shown to the opt-out consumer is

b_{0} = P {T - state, interested} μ (1 - p) p = (1 - μ) λ μ (1 - p) p,

(3)

where

P {T - state, interested}

is the probability that the opt-out consumer is in T-state and is interested in the advertiser's product category, μ(1 − p) is the probability of the ad inducing the consumer in Period 2 to realize a sufficiently high product match value and purchase, and p is the advertiser's margin. Recall that advertising has no effect on a B-state consumer or on the consumer who is not interested in the product category.

The ad network anticipates the advertiser's bid in Equation 3 and sets the reserve price for the Period 2 ad accordingly. If $\underline{b} \leq (1 - μ) λ μ (1 - p) p$ , where $\underline{b}$ is the “outside bid,” then the ad network's profit-maximizing reserve price, as per Equation 3, is $(1 - μ) λ μ (1 - p) p$ ; otherwise, it is $\underline{b}$ . Therefore, the reserve price for the Period 2 ad shown to the opt-out consumer is $max {\underline{b}, (1 - μ) λ μ (1 - p) p}$ , and the advertiser's Period 2 ad is shown if and only if $\underline{b} \leq (1 - μ) λ μ (1 - p) p$ .¹⁶ The following lemma summarizes the implications of the Period 2 ad (with an ad shown to the consumer in Period 1) for the opt-out consumer's instrumental values.

Lemma 1.

An opt-out consumer's expected intensity of repeat ads and expected product surplus (across the two periods), respectively, are

E [Q (0)] = {\begin{matrix} μ^{2} & if \underline{b} \leq \frac{1}{4} λ (1 - p) p and \underline{μ} < μ \leq \bar{μ} \\ 0 & otherwise \end{matrix},

(4)

\begin{aligned} and E [S (0)] = & \frac{1}{2} λ μ (1 - p)^{2} \\ \times {\begin{matrix} 2 - μ & if \underline{b} \leq \frac{1}{4} λ (1 - p) p and \underline{μ} < μ \leq \bar{μ} \\ 1 & otherwise \end{matrix}, \end{aligned}

(5)

where

\underline{μ}

and

\bar{μ}

are as defined in the proof.

Lemma 1 shows that the opt-out consumer expects to see repeat ads if and only if ad effectiveness is intermediate (see the dashed line in Figure 2, Panel A). For the parameter values b = .03 and λ = p = .5 used in Figure 2, Panel A, the intermediate range corresponds to the central 20% quantile of the range of ad effectiveness (i.e., when ad effectiveness is between 40% and 60%). Intuitively, if μ is either small or large, then (1 − μ)λμ(1 − p)—the average incremental sales generated by the Period 2 ad—is low. This is because if μ is small, the Period 2 ad is not effective in attracting consumer attention, whereas if μ is large, the Period 1 ad diminishes the likelihood of the consumer being in T-state, and this saturation effect of the Period 1 ad under large μ renders successive advertising less valuable. However, if μ is intermediate—large enough that the Period 2 ad sufficiently captures the consumer's attention and small enough that the Period 1 ad does not saturate—the advertiser shows a repeat ad to the opt-out consumer in Period 2.

Figure 2.

Subgame Outcomes for Opt-Out and Opt-In Consumers; b = .03, λ = p = .5.

The relationship between the ad's potential to attract attention (i.e., its effectiveness) and the opt-out consumer's expected product surplus is driven by two effects. First, higher probability of capturing a consumer's attention implies that the consumer is more likely to benefit from the product information conveyed by the ad, leading to higher expected product surplus for $μ \leq \bar{μ}$ , as illustrated by the dashed line in Figure 2, Panel B. Second, the expected product surplus also changes discontinuously as the advertising regime shifts in μ. For large μ, the saturation effect induces a shift in advertising regime: in the neighborhood of $\bar{μ}$ in Figure 2, Panel B, repeat ads are shown for $μ < \bar{μ}$ but not for $μ > \bar{μ}$ . The reduction in ad repetition may prevent a consumer from obtaining potentially valuable product information from a repeat ad in the event that the consumer misses seeing the first ad.

Analysis for the Opt-In Consumer

Consider the Period 2 subgame for the opt-in consumer in which a Period 1 ad is shown. The distribution of the consumer at the start of Period 2 is as in Assertion 1. In contrast to the case of the opt-out consumer, the advertiser can condition its Period 2 bid, and the ad network its reserve price, on the opt-in consumer's inferred purchase and funnel state.

Since only a T-state consumer who is interested in the product category is potentially influenced by advertising, the expected marginal value of repeat advertising is positive only for a T-state consumer. A B-state consumer who is already informed of product match value is not impacted by successive ad exposure. The advertiser's valuation, and hence its Period 2 bid, for a T-state consumer impression is

b_{T} = \frac{(1 - μ) λ}{1 - μ + μ (1 - λ)} μ (1 - p) p = \frac{(1 - μ) λ}{1 - μ λ} μ (1 - p) p,

(6)

where

\frac{(1 - μ) λ}{1 - μ + μ (1 - λ)}

is the probability that, conditional on the consumer being in T-state, the consumer missed the first ad exposure but is interested in the product category. The advertiser's valuation for showing an ad to a consumer who is either in B-state or in T-state and uninterested in the product category is 0. The ad network anticipates the bid in Equation 6 and sets reserve price

max {\underline{b}, \frac{(1 - μ) λ}{1 - μ λ} μ (1 - p) p}

, which equals

\frac{(1 - μ) λ}{1 - μ λ} μ (1 - p) p

per Assumption 2. This implies that the advertiser's Period 2 ad is always shown to a T-state opt-in consumer.

Lemma 2.

The opt-in consumer's expected intensity of repeat ads and expected product surplus (across the two periods), respectively, are

E [Q (1)] = μ^{2} (1 - λ) and E [S (1)] = \frac{1}{2} λ μ (1 - p)^{2} (2 - μ) .

(7)

The expected intensity of repeat ads for the opt-in consumer increases in μ (see the solid line in Figure 2, Panel A). The intuition for why this trend is different from that for the opt-out consumer lies in the ad network's inability to distinguish between the two types of T-state consumer: the consumer who remains in T-state because they missed the first ad, and the consumer who saw the ad but ignored the informational content because they are not interested in the product category. As the ad's potential to attract attention increases, repeat ads are more likely mistargeted to the latter consumer, who saw the first ad but remained in T-state due to product category mismatch. As the consumer ex ante does not know whether the ad matches their product category interest, the opt-in consumer's expected ad repetition increases in μ.

The expected product surplus for the opt-in consumer increases in μ (see the solid line in Figure 2, Panel B). This is because tracking enables the advertiser to target a repeat ad to the T-state consumer who may benefit from another opportunity to obtain potentially valuable product information.

The Consumer's Privacy Choice and Its Economic Impact

We now turn to the consumer's equilibrium privacy choice. Given the previously noted forces that affect consumer utility, we characterize the consumer's equilibrium privacy choice under different market conditions. Furthermore, we highlight the economic ramifications of the consumer’s privacy choice on the strategies and payoffs of the advertiser and the ad network.

Comparing Lemmas 1 and 2, we characterize the effects of the consumer's opting in to tracking on the two components of consumer utility. First, opting in to tracking has nuanced effects on the expected intensity of ad repetition. Opting in decreases ad repetition if the ad is moderately effective in capturing consumer attention, and increases ad repetition otherwise (see Figure 2a). If the ad's potential to attract attention is intermediate, then repeat ads are shown to the opt-out consumer, whereas for the opt-in consumer, repeat ads are targeted only to a T-state consumer. However, if μ is either small or large, then the advertiser shows no repeat ads to the opt-out consumer, as explained previously, but targets repeat ads to the T-state opt-in consumer. In other words, increased targeting efficiency afforded by trackability of the opt-in consumer motivates the advertiser to always show repeat ads to the T-state opt-in consumer. Although this benefits the opt-in consumer who realizes a product category match but missed the first ad exposure, it wears out the consumer who is not interested in the product category.

Second, opting in to tracking weakly increases the consumer's expected product surplus (see Figure 2, Panel B). This is because tracking increases targeting efficiency, which enables the advertiser to target a repeat ad to the T-state consumer. This provides the consumer who missed the first ad another opportunity to obtain potentially relevant product information.

The following proposition summarizes the effects of opting in to tracking on ad repetition and product surplus, which are the two components of consumer utility.

Proposition 1 (impact of opting in on components of consumer utility).

The following hold:

Opting in to tracking decreases repeat ads if the ad's potential to attract attention is intermediate, and increases repeat ads otherwise; that is, $E [Q (1)] < E [Q (0)]$ if $\underline{b} \leq \frac{1}{4} λ (1 - p) p$ and $\underline{μ} < μ \leq \bar{μ}$ , and $E [Q (1)] > E [Q (0)]$ otherwise.

Opting in to tracking weakly increases the consumer's expected product surplus; that is, $E [S (1)] \geq E [S (0)]$ .

An important takeaway from Proposition 1 is that if μ is either small or large, the consumer faces a trade-off between more intense ad repetition and higher product surplus by opting in. If an opt-in consumer realizes that they are not interested in the advertised product category, they may see more repeat ads than if they had opted out, which exacerbates wearout. However, if the opt-in consumer is interested in the product category, then repeat ads provide this consumer with additional opportunities to receive valuable product information that they may have missed from earlier ad exposures. Therefore, if μ is either small or large, the consumer's sensitivity to ad wearout critically determines their net utility of (dis)allowing tracking.

Therefore, the consumer considers two potentially countervailing effects of tracking on their utility to make their opt-in choice. If the ad's potential for attracting consumer attention is high, then opting in to tracking reduces ad repetition without hurting their product surplus. In this case, the consumer opts in to tracking. In contrast, if the ad's potential to attract attention is low, then opting in to tracking exacerbates wearout but also increases product surplus. In this case, the trade-off depends on how sensitive the consumer is to ad wearout. Define $η^{*} \equiv \frac{λ {(1 - p)}^{2} (1 - μ)}{2 μ (1 - λ)}$ ; the following proposition states the conditions under which the consumer opts in to or out of tracking.

Proposition 2 (consumer opt-in pattern).

The consumer opts in to tracking if either (a) $\underline{b} \leq \frac{1}{4} λ (1 - p) p$ and $\underline{μ} < μ \leq \bar{μ}$ , or (b) $η < η^{*}$ ; otherwise, the consumer opts out of tracking.

Proposition 2 shows that the consumer opts in to tracking under two conditions: either if the ad's potential to attract attention is intermediate, or if the consumer is sufficiently averse to seeing repeat ads (see Figure 3). The intuition is as follows. If ads are moderately effective in attracting consumer attention, then the consumer expects the advertiser to show untargeted repeat ads to opt-out consumers, even if the consumer is already informationally saturated by the first ad. In contrast, the consumer expects fewer repeat ads if the consumer opts in to tracking because finer targetability allows the advertiser to reduce wasteful ad repetition. Therefore, the mitigation of ad wearout incentivizes the consumer to opt in to tracking.

Figure 3.

Consumer Opt-In Behavior; b = .03, λ = p = .5.

The consumer also opts in to tracking if their ad wearout sensitivity is sufficiently low. Recall from Proposition 1 that if ad effectiveness is either low or high, the consumer faces a trade-off between higher ad repetition and higher product surplus. Therefore, if the consumer does not mind seeing repeat ads, then the surplus gains from the informational value of repeat ads motivate the consumer to opt in to tracking.

Finally, Proposition 2 sheds light on an interesting relationship between an ad's potential to attract attention and the consumer's opt-in behavior. If the consumer's wearout sensitivity is high (i.e., η > η*), then their opt-in behavior may be nonmonotonic in ad effectiveness (see Figure 3). Specifically, the consumer opts out for small and large μ, but opts in for intermediate μ. If the consumer is sensitive to ad wearout, their privacy choice is largely determined by the ad repetition implications. Since the consumer expects opting in to tracking to decrease ad repetition for intermediate μ and increase it for extreme μ (see Proposition 1), the consumer opts in to tracking if and only if μ is intermediate.

Overall, Proposition 2 shows that the consumer's privacy choice depends crucially on the ad's potential to attract attention and the consumer's sensitivity to ad wearout. The ad's potential to attract attention determines the intensity of repeat ads, and wearout sensitivity motivates the consumer to use tracking choices to mitigate wearout effects. Before proceeding, we clarify here that we assume a single representative consumer to deliver core insights cleanly. In an extension, we model heterogeneity in consumers’ intrinsic privacy valuations and find that consumers’ privacy choices vary across those with low versus high intrinsic privacy valuations.

Next, we analyze the advertiser's ad repetition strategy in a regime where the consumer can make privacy choices. We focus our discussion on low ranges of ad effectiveness (i.e., $μ \leq \bar{μ}$ ), which are arguably more aligned with industry estimates.¹⁷ We state the pattern in the following proposition.

Proposition 3 (advertiser’s repeat ad strategy).

Suppose $μ \leq \bar{μ}$ , and let $\underline{η} < \bar{η}$ be as defined in the proof. As ad effectiveness increases, the intensity of repeat ads shown may increase or decrease, depending on the ad effectiveness and the consumer's sensitivity to ad wearout. Specifically,

if $η \leq \underline{η}$ , the intensity of repeat ads decreases in μ;

if $\underline{η} < η \leq \bar{η}$ , the intensity of repeat ads first weakly decreases in μ, increases discontinuously at $μ = \underline{μ}$ , and then decreases in μ; and

if $\bar{η} < η$ , the intensity of repeat ads discontinuously increases at $μ = \underline{μ}$ , and then decreases in μ.

In Figure 4, we illustrate the pattern implied in Part 2 of Proposition 3.¹⁸ For $μ \leq \underline{μ}$ , the advertiser serves weakly fewer repeat ads as ad effectiveness increases. Let $\underline{\underline{μ}} \equiv \frac{{(1 - p)}^{2} λ}{2 η (1 - λ) + {(1 - p)}^{2} λ} < \underline{μ}$ and consider $μ < \underline{\underline{μ}}$ , where the consumer will have opted in and can be targeted on the basis of the funnel state in Period 2. As ad effectiveness increases, the consumer will remain in T-state with a lower probability at the end of Period 1, such that fewer repeat ads need to be shown overall in Period 2. Next, consider $\underline{\underline{μ}} \leq μ < \underline{μ}$ . Here, the consumer will have opted out, such that the inability to target repeat ads induces the advertiser to withdraw from repeat advertising. At the threshold $μ = \underline{μ}$ , however, the consumer will have again opted in, and the same logic holds as for $μ < \underline{\underline{μ}}$ . A main insight from this analysis is that the advertiser's repeat ad strategy is nontrivial when the consumer can make privacy choices.

Figure 4.

Ad Repetition Intensity Under Endogenous Consumer Privacy Choice; b = .03, λ = p = .5, η = .25.

We next investigate the economic consequences of the consumer's privacy choice. First, we study the effect of the consumer's endogenous opt-in decision on the ad network's profit. Second, we compare our results with the preregulation benchmark in which the consumer is opted in to tracking by default and highlight the impact on the ad ecosystem of endowing the consumer with privacy choices.

The ad network's profit is

Π (x) = μ λ (1 - p) p + {\begin{matrix} (1 - μ) λ μ (1 - p) p + μ λ \underline{b} & if x = 1 (i . e ., consumer opts in), \\ max {(1 - μ) λ μ (1 - p) p, \underline{b}} & if x = 0 (i . e ., consumer opts out), \end{matrix}

(8)

where the first-period payoff (first summand in Equation 8) is the same across the consumer's opt-in choice, and the second-period payoff (second summand in Equation 8) varies due to the different information structure and thus different ad targetability. It can be shown that the ad network's profit is weakly greater under consumer opt-in than under consumer opt-out: Π(1) ≥ Π(0).¹⁹ This is intuitive because ad targeting efficiency increases as the opt-in consumer discloses information; this allows the ad network to extract a larger surplus from the advertiser. Combined with the consumer's equilibrium privacy choice (see Proposition 2), we obtain the following equilibrium ad network profit under endogenous privacy choices:

Π^{*} = μ λ (1 - p) p + {\begin{matrix} (1 - μ) λ μ (1 - p) p + μ λ \underline{b} & if \underline{b} \leq \frac{1}{4} λ (1 - p) p and \underline{μ} < μ \leq \bar{μ}, or η < η^{*}, \\ max {(1 - μ) λ μ (1 - p) p, \underline{b}} & otherwise . \end{matrix}

(9)

From Equation 9, with thresholds

\underline{η}

and

\bar{η}

defined in the proof of Proposition 3, and defining

\underline{\underline{μ}} \equiv \frac{{(1 - p)}^{2} λ}{2 η (1 - λ) + {(1 - p)}^{2} λ} < \underline{μ}

, we obtain the following proposition.

Proposition 4 (ad network profit).

Suppose $μ \leq \bar{μ}$ . If $η \leq \underline{η}$ or $\bar{η} < η$ , then the ad network's profit increases in μ. If $\underline{η} < η \leq \bar{η}$ , then the ad network's profit increases in μ with a discontinuous drop at $μ = \underline{\underline{μ}}$ .

Lay intuition suggests that as ads become more effective in capturing consumer attention, advertisers would bid higher for ad impressions such that the ad network's revenue would increase. Proposition 4 shows that this intuition is not necessarily true in a regime where consumers make endogenous privacy choices. Specifically, Figure 5 illustrates this for the case when the consumer's wearout sensitivity is intermediate (i.e., $\underline{η} < η \leq \bar{η}$ ); that is, the ad network's profit discontinuously decreases in the ad's effectiveness at $μ = \underline{\underline{μ}}$ . This is because at $μ = \underline{\underline{μ}}$ , the consumer changes from opting in to opting out as ad effectiveness increases, such that the decline in targeting efficiency lowers the ad network's profit.

Figure 5.

Ad Network Profit Under Endogenous Consumer Privacy Choices; b = .03, λ = p = .5, η = .25.

The preceding discussion provides important insights for marketers adapting to the new business landscape in which data ownership is shifting from firms to consumers. To the extent that ad networks can influence ad effectiveness (e.g., by helping advertisers optimize their ad copies through ad campaign and bidding optimization tools), these efforts should take into account the consumers’ sensitivity to ad repetition, which could materially impact their privacy choices and, therefore, the profit generated from ads.

Thus far, we have examined the consumer's endogenous privacy choice and its implications for advertiser's ad repetition strategy and the ad network's profit. An important question for policy makers is as follows: What is the impact on the ad ecosystem of granting consumers privacy choices? Specifically, what are the implications of privacy regulations that empower consumers with data privacy rights on consumer surplus, total ad revenue, and advertising outcomes? We address these questions by comparing the equilibrium results from the previous section with the benchmark case in which the consumer is opted in by default. Note that the all-opt-in benchmark is simply the subgame for the opt-in consumer, which was presented previously. From this comparison, we address these questions in the following proposition.

Proposition 5 (impact of switching from all-opt-in regime to consumer-privacy-choice regime).

Compared with the benchmark in which the consumer is opted in to tracking, endowing the consumer with privacy choices

weakly decreases the expected intensity of repeat ads,

weakly increases consumer surplus while weakly decreasing product consumption, and

weakly decreases total ad revenue.

First, empowering the consumer with privacy choices changes the advertiser's strategy compared with the all-opt-in benchmark only if the consumer exercises their opt-out rights. This occurs if ad effectiveness is at the extremes and if the consumer is sensitive to ad wearout because the consumer expects higher targetability to raise ad repetition intensity for opt-in consumers. As the consumer opts out of tracking under these conditions, the lack of targetability induces the advertiser to withdraw from repeat advertising such that the expected intensity of ad repetition decreases. Second, privacy choices weakly increase consumer surplus, even though the consumer's product consumption weakly decreases. This is because the consumer's benefit of having privacy choices stems primarily from the alleviation of ad wearout. At the margin, a consumer will opt out of tracking when the additional expected surplus from an ad's information is equal to the additional wearout from the ad; therefore, if the consumer is forced to opt in, then their surplus decreases. Finally, the endowment of privacy choices weakly reduces the total ad revenue. As the wearout-sensitive consumer opts out of tracking, the advertiser–consumer matching efficiency decreases, which in turn reduces the total surplus.

Extensions

In this section, we consider two extensions of the main model that incorporate important and interesting aspects into the model, namely, the intrinsic value of privacy and the ability to infer product category interest for opt-in consumers. These extensions show the robustness of the insights obtained from the main model and provide additional insights.

Intrinsic Value of Privacy

We consider a situation with a unit mass of consumers in which consumers not only care about the instrumental implications of consenting to being tracked (i.e., the impact on ad repetition and product surplus) but also have intrinsic valuations of privacy; that is, they may want to protect privacy for its own sake (Becker 1980; Farrell 2012; Lin 2022; Posner 1981). We assume that if a consumer opts in to tracking, then independently of the instrumental values associated with the firm's reactions, this consumer experiences a disutility of θ, where θ is distributed across consumers according to a cumulative distribution function $F (\cdot)$ on $[0, \infty)$ . Heterogeneity in θ may stem from numerous factors, such as differences in consumers’ perceptions of how secure their shared data is and differences in what consumers believe constitutes personal information (Acquisti, Taylor, and Wagman 2016).

The consumer's expected utility from making privacy choice $x \in {0, 1}$ is

E [U (x)] = E [S (x)] - η E [Q (x)] - θ x .

(10)

We note that since the intrinsic value of privacy is heterogeneous, expected utilities in Equation 10 are different across consumers. Therefore, different consumers make different privacy choices, unlike in the main model.

Since the intrinsic privacy cost θ is incurred at the moment of the consumer's decision, the advertiser's and the ad network's strategies for opt-in and opt-out consumers remain unchanged. The key departure from the main model is that consumers weigh the instrumental utility $E [S (x)] - η E [Q (x)]$ against the intrinsic privacy cost θ. The following proposition characterizes the consumers’ equilibrium opt-in choices.

Proposition 6 (consumer opt-in pattern with intrinsic value of privacy).

Let I denote the mass of consumers who opt in to tracking. If $\underline{μ} < μ \leq \bar{μ}$ , then $I = F (η λ μ^{2})$ .

If $μ \leq \underline{μ}$ or $μ > \bar{μ}$ , then $I = {\begin{matrix} F [μ (\frac{1}{2} λ (1 - μ) {(1 - p)}^{2} - η μ (1 - λ))] & if η \leq η^{*}, \\ 0 & otherwise . \end{matrix}$

The equilibrium intensity of repeat ads is $I \cdot μ^{2} (1 - λ) + (1 - I) \cdot μ^{2} I {\underline{μ} < μ \leq \bar{μ}} .$

Comparing Propositions 2 and 6 shows that the key economic forces from the main model carry over. However, heterogeneous privacy costs imply that consumers with different privacy costs make different privacy choices. Specifically, if no consumers opt in in the main model, then no consumers opt in in this extended model, whereas if all consumers opt in in the main model, then only consumers with low enough intrinsic privacy cost opt in in this extended model. As an illustration, Figure 6 shows the fraction of consumers who opt in for a specific set of values of the model's parameters (in this plot, for $\underline{μ} < μ \leq \bar{μ}$ , all consumers opt in in the main model). The equilibrium intensity of repeat ads characterized in Proposition 6 is a convex combination of the ad repetition intensity for opt-out and opt-in consumers from the main model. Therefore, the qualitative pattern with respect to ad effectiveness carries over.

Figure 6.

Fraction of Opt-In Consumers; b = .03, λ = p = η = .5, F(θ) = θ, $θ \in [0, 1]$ .

Tracking and Product Category Targeting

In the main model, we assumed that the ad network was able to infer the funnel states of opt-in consumers and used this information to sell targeted ads to the advertiser. We extend the main model by allowing the ad network to infer, for the opt-in consumers, not only their ex post funnel state information after ads are shown but also their ex ante product category interest information before ads are shown in the first period. We assume that, for opt-in consumers, advertisers select themselves such that opt-in consumers have higher probability of seeing ads from product categories they are interested in than opt-out consumers. Specifically, opt-in consumers are interested in the advertised product category w.p. (1 + α)λ, whereas opt-out consumers are interested in the advertised product category w.p. λ, where $α \in [0, 1 / λ - 1]$ proxies the differential product interest targetability between opt-in and opt-out consumers.

Compared with the main model, the ability to infer consumers’ product category interest has two key implications for the advertiser: (1) the advertiser shows fewer repeat ads to opt-in consumers, and (2) the advertiser bids higher for repeat ads. First, since opt-in consumers are more likely to be interested in the product category than in the main model, they are less likely to remain in T-state; they are more likely to process the first ad and transition to B-state. Combined with the fact that the advertiser shows repeat ads only to T-state consumers, inference of product category interest reduces the intensity of repeat ads shown to opt-in consumers, compared with the main model. Second, higher likelihood of interest in the advertised product category implies higher ad valuation such that the advertiser's bid for repeat ads is greater than that in the main model. Consequently, the ad network's profit increases in α. We summarize these findings in the following proposition.

Proposition 7 (advertiser’s repeat ad strategy and ad network profit when category interest can be inferred).

The intensity of repeat ads weakly decreases in α, whereas the ad network's profit weakly increases in α.

Next, we specify how the inference of product category interest affects consumers’ opt-in choices. Define $η_{α}^{*} \equiv \frac{λ {(1 - p)}^{2} ((1 + α) (2 - μ) - 1)}{2 μ (1 - (1 + α) λ)}$ ; the following proposition summarizes the consumers, opt-in choices in this setting.

Proposition 8 (consumer opt-in pattern when category interest can be inferred).

Consumers opt in to tracking if either (a) $\underline{b} \leq \frac{1}{4} λ (1 - p) p$ and $\underline{μ} < μ \leq \bar{μ}$ , or (b) $η < η_{α}^{*}$ ; otherwise, consumers opt out of tracking.

For $α \in [0, 1 / λ - 1]$ , we have that $η_{α}^{*} \geq η^{*}$ (where $η^{*}$ is as in Proposition 2); that is, the parametric region for which consumers opt in to tracking when category interest can be inferred is larger than that in the main model. Furthermore, $η_{α}^{*}$ increases in α; that is, consumers have a stronger incentive to opt in to tracking as α increases.

The additional incentive for consumers to opt in to tracking when category interest can be inferred for opt-in consumers, and why it increases with α, is driven by two effects. The direct effect of higher α is that opt-in consumers receive higher product surplus because they are more likely to see ads for product categories that they are interested in. The indirect effect of higher α is that the higher product category interest motivates opt-in consumers to process ads such that they move down the purchase funnel. According to the funnel state inferences, the advertiser forgoes showing repeat ads to these consumers because they already obtained product match information (see Proposition 7). In summary, more accurate product category targeting provides double blessings to consumers that incentivize them to opt in to tracking: it increases the opt-in consumers’ expected product surplus and reduces the expected ad repetition.

Conclusion

Tracking consumers’ internet activities enables dynamic ad targeting as consumers traverse the purchase funnel for a product. Although tracking may increase the efficiency with which ads are matched to consumers in different funnel states, it may also affect advertisers’ ad strategies that indirectly impact consumers’ utilities. In this setting, we study the impact of regulations (e.g., the GDPR) that, motivated by privacy concerns, endow consumers with the choice to have their online activities be tracked or not. In particular, we develop a framework to analyze consumers’ opt-in/out decisions and their impact on the strategies and profits of ad networks and advertisers.

We obtain a number of insights from our analysis. First, opting in to tracking, compared with opting out of it, has mixed effects on the number of repeat ads shown to consumers. For instance, if the ad's potential to attract consumer attention is intermediate, then trackability through opting in reduces wasteful ad repetition. However, if the ad's potential to attract consumer attention is low or high, then trackability increases ad repetition; this is because the advertiser forgoes showing repeat ads to opt-out consumers due to the inability to identify only those consumers who may obtain information from them, whereas trackability of opt-in consumers enables such specific targeting.

Second, when making their tracking choices, consumers consider the previously described advertising implications and balance the useful product information they can expect to obtain from ads against the wearout they can expect to experience. If the ad's potential to attract attention is intermediate, then the reduced ad repetition under trackability motivates consumers to opt in, whereas in other cases, consumers opt out to mitigate ad wearout. If consumers have low sensitivity to wearout, then consumers weigh more heavily the benefit of obtaining potentially valuable product information from repeat ads than the cost of ad wearout. We find that endowing consumers with privacy choices weakly increases their net surplus while weakly decreasing product consumption.

Third, we generate implications for the advertiser and the ad network that sells the ads. For the advertiser, we derive optimal advertising strategies, such as which consumers it should target with repeat ads (when targeting is possible). For the ad network, we uncover a nuanced relationship between ad effectiveness and its profit. Specifically, due to consumers’ incentive to mitigate ad wearout, higher ad effectiveness may induce consumers to opt out of tracking. Therefore, even if higher ad effectiveness implies higher ad valuation for the advertiser, it may elicit privacy choices from consumers that undermine the ad network's tracking capability, which in turn reduces the ad network's profit.

We acknowledge several limitations of our research. First, it would be interesting to examine a more active role of publishers. One could consider publishers acting as information gateways and study the forces that affect the publishers’ incentives to disclose consumers’ information to the ad network or withhold it from the ad network. Second, future research could relax the exogenous consumer website visit assumption and investigate how consumers adjust their website visit decisions in tandem with their tracking choices. Third, although we allow heterogeneity in consumers’ ad responsiveness, we assume that ad effectiveness for top-funnel consumers is homogeneous (i.e., all equal to μ). An interesting avenue for future research could be to study the implications of heterogeneous ad effects and the advertiser's incentive to potentially screen consumers on the basis of this heterogeneity. Fourth, future research could consider the interesting and related phenomenon of skippable ads. A distinct feature of skippable ads is their interactivity (Dukes, Liu, and Shuai 2022), which may help advertisers glean information about consumers’ interests. To the extent that consumer's ad skip decisions also fall under the umbrella of personal data and thus cannot be tracked by firms unless consumers consent, advertisers and ad networks would learn disproportionately more about opt-in consumers’ preferences than about opt-out consumers’ preferences. Such asymmetry would amplify the discrepancy in targetability between opt-in and opt-out consumers.²⁰ Fifth, regulations such as the GDPR and the California Consumer Privacy Act and California Privacy Rights Act are consistent with the “rights and responsibilities” model. This model states that consumers should have certain privacy rights, such as being tracked only on opting in, and firms have certain responsibilities, such as protecting consumer data. The instrumental and intrinsic values of privacy in our model correspond to the responsibilities and the rights, respectively. However, we do not consider all rights and responsibilities covered in recent regulations, such as the ability of consumers to have their data shared with themselves or transferred to other parties, and the requirement that firms maintain data in a format that is easy to understand at the time of such data operations. Future work can consider these other aspects of regulation as well. Finally, although our model focuses on consumer heterogeneity in intrinsic privacy valuations, the framework could be extended to incorporate heterogeneous wearout sensitivities (Chae, Bruno, and Feinberg 2019). Under such formulation, the proportion of opt-in consumers would be shaped primarily by the trade-offs within the instrumental aspects of privacy; that is, between the consumer's expected product surplus and intensity of ad repetition.

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437221140052 - Supplemental material for Consumer Privacy Choices and (Un)Targeted Advertising Along the Purchase Journey

Supplemental material, sj-pdf-1-mrj-10.1177_00222437221140052 for Consumer Privacy Choices and (Un)Targeted Advertising Along the Purchase Journey by W. Jason Choi, Kinshuk Jerath and Miklos Sarvary in Journal of Marketing Research

Footnotes

Notes

Appendix: Proofs

Associate Editor

Anthony Dukes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

W. Jason Choi

References

Aaker

David A.

Bruzzone

Donald E.

(1985), “Causes of Irritation in Advertising,” Journal of Marketing, 49 (2), 47–57.

Acquisti

Alessandro

Taylor

Curtis R.

Wagman

Liad

(2016), “The Economics of Privacy,” Journal of Economic Literature, 54 (2), 442–92.

Anderson

Simon P.

Baik

Alicia

Larson

Nathan

(2019), “Prices, Poaching, and Privacy with Personalized Targeted Discounts,” working paper, University of Virginia.

Apple (2021), “AppTrackingTransparency Requirement Update,” (January 28), https://apple.co/3AV8Ubc.

AppsFlyer (2021), “The Impact of iOS 14+& ATT on the Mobile App Economy,” (accessed May 19, 2022), https://www.appsflyer.com/resources/reports/ios-14-att-dashboard/.

Athey

Susan

Gans

Joshua S.

(2010), “The Impact of Targeting Technology on Advertising Markets and Media Competition,” American Economic Review, 100 (2), 608–13.

Auxier

Brooke

Rainie

Lee

Anderson

Monica

Perrin

Andrew

Kumar

Madhu

Turner

Erica

(2019), “Americans and Privacy: Concerned, Confused and Feeling Lack of Control Over Their Personal Information,” Pew Research Center (November 15), https://pewrsr.ch/3PwDBrY.

Becker

Gary S.

(1980), “Privacy and Malfeasance: A Comment,” Journal of Legal Studies, 9 (4), 823–26.

Bergemann

Dirk

Bonatti

Alessandro

(2011), “Targeting in Advertising Markets: Implications for Offline Versus Online Media,” RAND Journal of Economics, 42 (3), 417–43.

10.

Bigler

Jason

(2019), “Rolling Out First Price Auctions to Google Ad Manager Partners,” Google (September 5), https://www.blog.google/products/admanager/rolling-out-first-price-auctions-google-ad-manager-partners/.

11.

Campbell

James

Goldfarb

Avi

Tucker

Catherine

(2015), “Privacy Regulation and Market Structure,” Journal of Economics & Management Strategy, 24 (1), 47–73.

12.

Campbell

Margaret C.

Keller

Kevin L.

(2003), “Brand Familiarity and Advertising Repetition Effect,” Journal of Consumer Research, 30 (2), 292–304.

13.

Chae

Inyoung

Bruno

Hernán A.

Feinberg

Fred M.

(2019), “Wearout or Weariness? Measuring Potential Negative Consequences of Online Ad Volume and Placement on Website Visits,” Journal of Marketing Research, 56 (1), 57–75.

14.

Choi

W. Jason

Jerath

Kinshuk

(2022), “Privacy and Consumer Empowerment in Online Advertising,” Foundations and Trends in Marketing, 15 (3), 153–212.

15.

Choi

W. Jason

Jerath

Kinshuk

Sarvary

Miklos

(2020), “Advertising and Price Competition Under Consumer Data Privacy Choices,” working paper, University of Maryland.

16.

Conitzer

Vincent

Taylor

Curtis R.

Wagman

Liad

(2012), “Hide and Seek: Costly Consumer Privacy in a Market with Repeat Purchases,” Marketing Science, 31 (2), 277–92.

17.

D’Annunzio

Anna

Russo

Antonio

(2020), “Ad Networks and Consumer Tracking,” Management Science, 66 (11), 5040–58.

18.

De Cornière

Alexandre

de Nijs

Romain

(2016), “Online Advertising and Privacy,” RAND Journal of Economics, 47 (650), 48–72.

19.

Dukes

Anthony

Liu

Qihong

Shuai

Jie

(2022), “Skippable Ads: Interactive Advertising on Digital Media Platforms,” Marketing Science, 41 (3), 528–47.

20.

eMarketer (2021), “Display Ad In-View Rates,” (accessed May 19, 2022), https://performance-na1.emarketer.com/channel/59e8a4a0bfce890eb411ec6a/58eb85f7dc3d370d90af591e.

21.

Esteban

Lola

Gil

Agustin

Hernandez

Jose M.

(2001), “Informative Advertising and Optimal Targeting in a Monopoly,” Journal of Industrial Economics, 49 (2), 161–80.

22.

Farrell

Joseph

(2012), “Can Privacy Be Just Another Good?” Journal on Telecommunications and High Technology Law, 10 (2), 251–64.

23.

Fisher

Lauren

(2019), “Programmatic Digital Display Ad Spending,” eMarketer (November 21), https://www.insiderintelligence.com/content/programmatic-digital-display-ad-spending.

24.

Godinho de Matos

Miguel

Adjerid

Idris

(2022), “Consumer Consent and Firm Targeting After GDPR: The Case of a Large Telecom Provider,” Management Science, 68 (5), 3330–78.

25.

Google (2022), “AdSense Revenue Share,” Google AdSense Help (accessed May 19, 2022), https://support.google.com/adsense/answer/180195.

26.

Gritckevich

Aleksandr

Katona

Zsolt

Sarvary

Miklos

(2022), “Ad Blocking,” Management Science, 68 (6), 4703–24.

27.

Grustniy

Leonid

(2022), “How to Block Cookies in Your Browser,” Kaspersky (January 31), https://usa.kaspersky.com/blog/how-to-block-cookies-in-chrome-safari-firefox-edge/26098/.

28.

Hoban

Paul R.

Bucklin

Randolph E.

(2015), “Effects of Internet Display Advertising in the Purchase Funnel: Model-Based Insights from a Randomized Field Experiment,” Journal of Marketing Research, 52 (3), 375–93.

29.

Ichihashi

Shota

(2020), “Online Privacy and Information Disclosure by Consumers,” American Economic Review, 110 (2), 569–95.

30.

Inskin Media (2019), “Environment Matters: Improving Online Brand Experiences,” (accessed October 18, 2020), https://www.inskinmedia.com/news/infographic-environment-matters-improving-online-brand-experiences.

31.

Iyer

Ganesh

Soberman

David

Villas-Boas

J. Miguel

(2005), “The Targeting of Advertising,” Marketing Science, 24 (3), 461–76.

32.

Jatain

Vishveshwar

(2021), “24 Best Ad Networks for Publishers,” AdPushup (accessed May 19, 2022), https://www.adpushup.com/blog/the-best-ad-networks-for-publishers/.

33.

Johnson

Justin P.

(2013), “Targeted Advertising and Advertising Avoidance,” RAND Journal of Economics, 44 (1), 128–44.

34.

Johnson

Garrett

Lewis

Randall A.

Nubbemeyer

Elmar

(2017), “The Online Display Ad Effectiveness Funnel & Carryover: Lessons from 432 Field Experiments,” working paper, Boston University.

35.

T. Tony

Sudhir

(2020), “Privacy Rights and Data Security: GDPR and Personal Data Driven Markets,” working paper, The Chinese University of Hong Kong.

36.

Kotler

Philip

Keller

Kevin L.

(2012), Marketing Management. Pearson.

37.

Lambrecht

Anja

Tucker

Catherine E.

(2013), “When Does Retargeting Work? Information Specificity in Online Advertising,” Journal of Marketing Research, 50 (5), 561–76.

38.

Lavidge

Robert J.

Steiner

Gary A.

(1961), “A Model for Predictive Measurements of Advertising Effectiveness,” Journal of Marketing, 25 (6), 59–62.

39.

Lin

Tesary

(2022), “Valuing Intrinsic and Instrumental Preferences for Privacy,” Marketing Science, 41 (4), 663–81.

40.

Montes

Rodrigo

Sand-Zantman

Wilfried

Valletti

Tommaso

(2019), “The Value of Personal Information in Online Markets with Endogenous Privacy,” Management Science, 65 (3), 1342–62.

41.

Pieters

Rik

Wedel

Michel

Zhang

Jie

(2007), “Optimal Feature Advertising Design Under Competitive Clutter,” Management Science, 53 (11), 1815–28.

42.

Posner

Richard A.

(1981), “The Economics of Privacy,” American Economic Review, 71 (2), 405–09.

43.

Rafieian

Omid

Yoganarasimhan

Hema

(2021), “Targeting and Privacy in Mobile Advertising,” Marketing Science, 40 (2), 193–218.

44.

Sahni

Navdeep S.

Narayanan

Sridhar

Kalyanam

Kirthi

(2019), “An Experimental Investigation of the Effects of Retargeted Advertising: The Role of Frequency and Timing,” Journal of Marketing Research, 56 (3), 401–18.

45.

Schwartz

Eric M.

Bradlow

Eric T.

Fader

Peter S.

(2017), “Customer Acquisition Via Display Advertising Using Multi-Armed Bandit Experiments,” Marketing Science, 36 (4), 500–522.

46.

Shen

Qiaowei

Villas-Boas

J. Miguel

(2018), “Behavior-Based Advertising,” Management Science, 64 (5), 2047–64.

47.

Shin

Jiwoong

Jungju

(2021), “Targeted Advertising and Consumer Inference,” Marketing Science, 40 (5), 900–922.

48.

Taylor

Curtis R.

(2004), “Consumer Privacy and the Market for Customer Information,” RAND Journal of Economics, 35 (4), 631–50.

49.

Todri

Vilma

Ghose

Anindya

Singh

Param Vir

(2020), “Trade-Offs in Online Advertising: Advertising Effectiveness and Annoyance Dynamics Across the Purchase Funnel,” Information Systems Research, 31 (1), 102–25.

50.

Tourn

Gastón

(2019), “Repetition Isn’t Authentic or Consistent—It’s Annoying,” Adweek, (October 24), https://www.adweek.com/brand-marketing/repetition-isnt-authentic-or-consistent-its-annoying/.

51.

Villas-Boas

J. Miguel

(2004), “Price Cycles in Markets with Customer Recognition,” RAND Journal of Economics, 35 (3), 486–501.

52.

Wedel

Michel

Pieters

Rik

(2000), “Eye Fixations on Advertisements and Memory for Brands: A Model and Findings,” Marketing Science, 19 (4), 297–312.

53.

Zhang

Jie

Wedel

Michel

Pieters

Rik

(2009), “Sales Effects of Attention to Feature Advertisements: A Bayesian Mediation Analysis,” Journal of Marketing Research, 46 (5), 669–81.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.42 MB