flowchart LR A["Culturally<br/>constituted world<br/><i>(status, class, gender,<br/>lifestyle)</i>"] -->|public roles<br/>& performances| B["Celebrity<br/>endorser"] B -->|advertising| C["Brand /<br/>product"] C -->|consumption<br/>rituals| D["Consumer's<br/>extended self"] classDef world fill:#eef,stroke:#557; classDef person fill:#efe,stroke:#575; class A world; class D person;
16 Celebrity Endorsement
A celebrity endorser is an individual who enjoys public recognition and lends that recognition to a consumer good by appearing with it in advertising (McCracken 1989). Firms spend heavily on this practice because a famous face appears to do something a product claim cannot: it imports ready-made meaning— status, competence, glamour, rebelliousness—into a brand that would otherwise have to manufacture those associations slowly and at great cost. The endorsement contract is, in effect, the rental of an accumulated reputation. The central empirical questions are therefore the same ones we ask of any intangible asset: through what mechanism does the asset create value, how is that value measured, and what is the risk that it evaporates?
This chapter treats celebrity endorsement as a problem with three faces. The first is psychological: why does an endorser change what a consumer thinks and does? Two classical models—source credibility and source attractiveness— locate the answer in properties of the source (expertise, trust, likability, similarity). A third, the meaning-transfer model, relocates it in culture: the endorser is a conduit that moves symbolic meaning from the culturally constituted world into the brand and finally into the consumer’s own identity project (McCracken 1986, 1989). The second face is strategic: which celebrity should endorse which brand? The match-up hypothesis answers that congruence between endorser and product governs effectiveness, and we give it a formal, testable structure. The third face is financial: endorsement contracts are firm-level investments whose announcement moves stock prices, whose returns can be estimated by event study, and whose downside—endorser risk—is the possibility that a scandal contaminates every brand the celebrity touches.
By the end of the chapter the reader should be able to state each model’s construct definitions formally, specify and estimate a match-up interaction, design an event study that isolates the firm-value effect of an endorsement (and of an endorser’s disgrace), and reason quantitatively about whether an endorsement is worth its price. Throughout, intuition leads and formalism follows immediately, in the house manner. Celebrity endorsement is best understood as a special case of the broader machinery developed in Chapter 11 (brands as networks of meaning and as priced firm assets) and Chapter 23 (event-study valuation of marketing actions); we draw on both.
16.1 Conceptual Foundations
The persuasion literature offers two distinct reasons a message recipient might be moved by who delivers a message rather than what the message says. The distinction matters because the two routes have different boundary conditions and different vulnerabilities, and because the dominant modern model—meaning transfer—rejects both as incomplete.
16.1.1 The Source Credibility Model
The source credibility model holds that a communicator’s persuasiveness rises with the audience’s perception that the source both knows the truth and is willing to tell it. Following the formalization that McCracken (1989) inherits from the Hovland tradition, credibility has two components. Expertise (or expertness) is the perceived ability of the source to make valid assertions—the degree to which the endorser is seen as a legitimate source of knowledge about the product class. Trustworthiness is the perceived willingness of the source to make valid assertions—the audience’s confidence that the endorser is reporting honestly rather than shilling. A message is most effective when the source scores high on both: a credentialed dermatologist endorsing a skin cream is persuasive to the extent the audience believes she knows skin and is not merely paid to say so.
An endorser is effective to the extent the audience perceives the source as able to make valid assertions (expertise) and willing to make them (trustworthiness); credibility is the conjunction of the two. — after the source-credibility tradition synthesized in McCracken (1989)
The model’s empirical workhorse is the Ohanian source scale, which operationalizes expertise, trustworthiness, and attractiveness as reflective multi-item semantic-differential batteries.1 Its appeal is that it yields a clean mediation story: endorser characteristics raise perceived credibility, which raises attitude toward the ad (\(A_{ad}\)), which raises attitude toward the brand (\(A_b\)) and purchase intention. Its limitation, which motivated the meaning-transfer turn, is that credibility ratings poorly predict why a non-expert celebrity—an actor with no plausible product knowledge—can nonetheless sell cars, watches, and soft drinks.
16.1.2 The Source Attractiveness Model
The source attractiveness model locates persuasion not in epistemic authority but in the audience’s affinity for the source. It decomposes attractiveness into three perceived properties (McCracken 1989). Familiarity is knowledge of the source through exposure—the sheer accumulated awareness that makes a face feel known. Likability is affection for the source arising from physical appearance and behavior. Similarity is a supposed resemblance between source and receiver. The model predicts that liked, familiar, and similar sources persuade more, operating through identification: the receiver accepts influence because doing so sustains a satisfying self-defining relationship with the source.
The attractiveness model rationalizes the use of physically beautiful endorsers and explains a robust laboratory regularity—attractive sources lift ad and brand attitudes—but it shares the credibility model’s defect. Treating attractiveness as a generic positive trait predicts that the most beautiful celebrity should sell everything equally well, which is false: an Olympic sprinter sells running shoes better than perfume, and a supermodel the reverse. The data demand a model in which the content of the celebrity’s image, not merely its valence, determines what the celebrity can sell. That is the contribution of meaning transfer.
16.1.3 The Meaning-Transfer Model
McCracken (1989) argues that both classical models are too thin: they reduce a culturally rich figure to a few scalar traits and cannot explain image-specific effectiveness. The meaning-transfer model replaces source properties with the symbolic meanings a celebrity carries—status, class, gender, age, personality, and lifestyle—accumulated across the roles, contexts, and performances in which the public has encountered the person. The celebrity is valuable precisely because these meanings are sharply drawn and already established in the audience’s mind.
Meaning moves in three stages, sketched in Figure 16.1. In the first stage, cultural meaning resides in the culturally constituted world and becomes attached to celebrities through their public lives. In the second, advertising transfers a selected subset of that meaning from celebrity to product: the ad’s creative task is to make the audience see the same meanings in the brand that they already see in the endorser. In the third, consumption rituals transfer meaning from the product to the consumer, who acquires and displays the good to construct and signal an identity. This third stage rests on the same premise as the rest of Chapter 11—that consumers consume to build the extended self (Belk 1988)—and is enacted through the possession, exchange, grooming, and divestment rituals through which the broader cultural system moves meaning into goods (McCracken 1986).
Four modes of endorsement govern how explicitly the celebrity is bound to the product within the ad, and thus how directly meaning is invited to transfer (McCracken 1989). In the explicit mode the endorser states “I endorse this product”; in the implicit mode, “I use this product”; in the imperative mode, “you should use this product”; and in the copresent mode the celebrity merely appears alongside the product, asserting nothing. The copresent mode is the purest meaning-transfer device—no claim is made, yet the juxtaposition licenses the audience to read the celebrity’s meanings into the brand—and it is precisely the mode the credibility and attractiveness models cannot explain, because no assertion is made whose validity expertise or trust could underwrite.
The three models are best read as nested in explanatory ambition rather than as rivals, as Table 16.1 summarizes. Credibility and attractiveness identify which source traits matter; meaning transfer explains why a given trait matters for a given brand and supplies the bridge to the match-up hypothesis below.
| Model | Locus of persuasion | Primitive constructs | Unresolved anomaly |
|---|---|---|---|
| Source credibility | Epistemic authority of the source | Expertise; trustworthiness | Non-expert celebrities still sell |
| Source attractiveness | Audience affinity for the source | Familiarity; likability; similarity | Beauty does not sell all categories equally |
| Meaning transfer | Cultural meanings the source carries | Status, class, gender, age, lifestyle | Requires congruence logic (match-up) |
16.2 The Match-Up Hypothesis
If a celebrity’s value is the specific meaning she carries, then effectiveness must depend on the fit between that meaning and the brand. The match-up hypothesis states exactly this: an endorser is more effective the more congruent the endorser’s image is with the endorsed product. The hypothesis is the strategic operationalization of meaning transfer—it tells the manager not merely to hire a credible or attractive celebrity but to hire one whose accumulated meanings match the meanings the brand wishes to own.
16.2.1 Intuition and Formalization
The intuition is a congruence-as-interaction claim: endorser quality and brand relevance are complements, not substitutes. A high-meaning endorser adds little to a brand the meaning does not fit, and a perfectly matched but low-profile endorser has little meaning to transfer. Let \(y\) denote a response outcome (attitude toward the brand, purchase intention, or choice). Let \(E\) index a property of the endorser (e.g., perceived attractiveness or expertise, mean- centered) and let \(M\) denote the match—the congruence between the endorser’s image and the product category, also mean-centered. The match-up hypothesis is the claim that the two interact:
\[ y = \beta_0 + \beta_1 E + \beta_2 M + \beta_3\,(E \times M) + \varepsilon, \qquad H_{\text{match-up}}:\ \beta_3 > 0. \tag{16.1}\]
The substantive content is entirely in \(\beta_3\). A positive interaction means the marginal effect of endorser quality, \(\partial y / \partial E = \beta_1 + \beta_3 M\), grows with match: the same attractive celebrity moves the outcome more for a beauty product than for a power tool. The main effects \(\beta_1\) and \(\beta_2\) are, on the match-up logic, of secondary interest; a model that omits the product term and reports only that “attractive endorsers work” has not tested the hypothesis at all. This is the same interaction-versus-main-effect distinction that governs brand-extension fit in Chapter 11, and it carries the same warning: congruence is a relational property and must be modeled relationally.
16.2.2 A Reproducible Test
Identification of \(\beta_3\) in an experiment is clean because \(E\) and \(M\) are assigned, not measured: a \(2 \times 2\) design crosses a high- versus low- attractiveness endorser with a high- versus low-match product, and randomization guarantees the regressors are orthogonal to \(\varepsilon\), so least squares recovers the interaction without confounding. The estimator is ordinary least squares (OLS); the identifying assumption is random assignment of the endorser– product pairing (which makes \(\mathbb{E}[\varepsilon \mid E, M] = 0\)); and what breaks identification in observational endorsement data is that firms choose matched endorsers, so \(M\) correlates with unobserved brand quality and \(\beta_3\) absorbs selection. The simulation below generates data under Equation 16.1 with a known positive interaction and recovers it.
Code
set.seed(1989) # year of McCracken's meaning-transfer paper
n_per_cell <- 60
design <- expand.grid(
endorser = c(-0.5, 0.5), # low / high attractiveness (effect-coded)
match = c(-0.5, 0.5), # low / high product match
rep = seq_len(n_per_cell)
)
# True data-generating process: positive match-up interaction (beta_3 = 1.4)
b0 <- 3.0; b1 <- 0.4; b2 <- 0.6; b3 <- 1.4
design$y <- with(design,
b0 + b1 * endorser + b2 * match + b3 * (endorser * match) +
rnorm(nrow(design), sd = 1))
fit <- lm(y ~ endorser * match, data = design)
round(summary(fit)$coefficients, 3)
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 2.937 0.066 44.695 0.000
#> endorser 0.587 0.131 4.464 0.000
#> match 0.376 0.131 2.864 0.005
#> endorser:match 1.249 0.263 4.750 0.000
# Cell means make the interaction visible: the endorser "lift" is large only
# when the product matches.
cell <- aggregate(y ~ endorser + match, data = design, FUN = mean)
lift_match <- with(cell, y[endorser == 0.5 & match == 0.5] -
y[endorser == -0.5 & match == 0.5])
lift_mismatch <- with(cell, y[endorser == 0.5 & match == -0.5] -
y[endorser == -0.5 & match == -0.5])
cat("Endorser lift when matched: ", round(lift_match, 2), "\n")
#> Endorser lift when matched: 1.21
cat("Endorser lift when mismatched: ", round(lift_mismatch, 2), "\n")
#> Endorser lift when mismatched: -0.04The estimated endorser:match coefficient recovers the planted \(\beta_3\), and the two “lift” quantities make the substantive claim concrete: switching to the attractive endorser helps far more when the product matches than when it does not. A non-significant interaction—not a significant main effect—is what would falsify the hypothesis.
16.2.3 Scope, Mechanism, and Boundaries
The match-up effect is not unconditional, and the modern literature has mapped its boundaries. The processing route matters: under the elaboration-likelihood logic of Petty, Cacioppo, and Schumann (1983) and Petty and Cacioppo (1986), source cues such as attractiveness operate as peripheral cues when involvement is low and the audience is not scrutinizing arguments, but as substantive evidence (or as a distraction) when involvement is high. Congruence therefore interacts with involvement, and an endorser who helps a low-involvement impulse purchase may be irrelevant or counterproductive for a high-involvement considered one. The mechanism also runs partly through transfer of affect: positive feelings conditioned to the celebrity can become attached to the brand through co-occurrence, a classical-conditioning account that complements meaning transfer and that, like it, predicts congruence-dependent effects rather than uniform lift. Finally, the multiple-product / multiple-endorser problem complicates the simple match-up story: a celebrity who endorses many unrelated brands dilutes the distinctiveness of the meaning available for transfer to any one of them, weakening congruence even when each pairing is individually well matched.
Two cross-cultural moderators deserve emphasis because they bound external validity. Power-distance belief—the extent to which a consumer accepts hierarchy— moderates the celebrity effect: power-distance beliefs strengthen the positive effect of celebrity (versus non-celebrity) endorsers on advertising evaluations, because celebrities are high-status sources whose influence is more readily accepted by those who endorse hierarchical social arrangements (Winterich, Gangwar, and Grewal 2018). This is a meaning-transfer result in disguise—status is one of the cultural meanings a celebrity carries, and its persuasive force depends on whether the audience values status. Self-construal further conditions which connection matters, paralleling the brand-attachment findings of Swaminathan, Page, and Gürhan-Canli (2007).
16.3 Influencers as Endorsers
The economics of endorsement have been reshaped by social-media influencers— individuals who accumulate audiences on platforms and monetize that attention through paid brand content. Influencers occupy an intermediate position between the classical celebrity and the ordinary peer: they are more similar to their followers than a film star (raising the attractiveness model’s similarity lever) yet command genuine reach, and their persuasiveness depends sharply on perceived authenticity. Sokolova, Seenivasan, and Thomas (2020) show that parasocial interaction and the perceived credibility of the influencer drive purchase intention, with engagement operating through identification rather than expertise. Disclosure of the commercial relationship is the new trust frontier: sponsorship transparency interacts with follower trust in ways the original trustworthiness construct anticipated but could not have specified. Influencer marketing is now an industry of its own, with intermediaries matching brands to creators (Nistor and Selove 2024), and even mega-influencer and celebrity-creator effects have been documented in firm-outcome data (Tian, Dew, and Iyengar 2024; Leung et al. 2022). The constructs are old; the measurement environment—observable, high-frequency, and platform-mediated—is new.
16.4 Endorser Risk
The asset a firm rents in an endorsement is a person, and persons misbehave. The defining hazard of celebrity endorsement is therefore endorser risk: the brand inherits not only the celebrity’s favorable meanings but also her liabilities, and a scandal can transfer negative meaning into the brand through exactly the channel of Figure 16.1 that was supposed to transfer positive meaning. A serious treatment of endorsement must price this downside, not merely the upside.
Several features make endorser risk distinctive. It is idiosyncratic to the person and largely outside the firm’s control, so it cannot be diversified within a single endorsement. It is transferable: meaning transfer is symmetric, and the same congruence that amplifies positive transfer amplifies the contamination when the endorser’s image sours. And it is correlated across a celebrity’s portfolio—one scandal simultaneously damages every brand the celebrity endorses, which is why endorsement contracts contain morality clauses permitting termination on disrepute. The negative information need not even be deserved: under the attribution logic that governs brand crises in Chapter 11, audiences who view character as stable (entity theorists) generalize a transgression to a global judgment of the person, and that judgment flows to the brand.
The timing of harm is subtle. A sleeper effect can arise when a discounting cue—“this source is untrustworthy”—decays faster than the message itself, so persuasion rises over time as the audience forgets why it discounted the source; symmetrically, the damage from a tainted endorser can dissipate as the association weakens, so the net effect of a scandal is dynamic rather than a one-time level shift (Wu et al. 2015). Managers who read a scandal’s immediate stock-price reaction as the whole story may therefore mis-estimate the long-run cost, in either direction.
16.5 Endorsement and Firm Value
Endorsement contracts are firm-level investments, and capital markets price them. The natural identification strategy is the event study developed in Chapter 23: an endorsement announcement (or an endorser scandal) is a discrete, dated event whose firm-value consequence is read from the abnormal stock return in a tight window around the announcement, after the normal return is removed.
16.5.1 The Event-Study Estimator
For firm \(i\) on trading day \(t\), the abnormal return is the realized return minus its expectation under a market model estimated on a clean pre-event window:
\[ \mathrm{AR}_{it} = R_{it} - \big(\hat\alpha_i + \hat\beta_i R_{mt}\big), \qquad \mathrm{CAR}_i(W) = \sum_{t \in W} \mathrm{AR}_{it}, \tag{16.2}\]
where \(R_{mt}\) is the market return, \((\hat\alpha_i, \hat\beta_i)\) are OLS estimates from the estimation window, and \(\mathrm{CAR}_i(W)\) is the cumulative abnormal return over a short event window \(W\) (often \([-1, +1]\) days). The method is the standard apparatus of McWilliams and Siegel (1997) and Srinivasan and Bharadwaj (2004) and is applied to brand-asset transactions in Chapter 11.
Three assumptions carry the inference, and each is a place identification can break. Market efficiency: prices impound the news quickly, so a short window captures the full effect; if the news leaks, the window must widen and power falls. No confounding: the window must be scrubbed of contemporaneous firm-specific news—earnings, splits, executive changes, buybacks, dividend changes—because the event study cannot separate two events sharing a window (Srinivasan and Bharadwaj 2004). Correct normal-return model: a misspecified benchmark (\(\hat\alpha, \hat\beta\)) contaminates every \(\mathrm{AR}\); thin trading and event-induced variance further bias the standard errors. The estimator answers a sharp question—by how much did this announcement change shareholder wealth?—only to the extent these assumptions hold.
16.5.2 A Reproducible Event Study
The following example simulates daily returns for a firm, injects a positive endorsement-announcement effect on the event day, estimates the market model on the pre-event window, and aggregates abnormal returns over a three-day event window. The structure mirrors exactly what one would run on CRSP data.
Code
set.seed(2024)
est_days <- 120 # estimation window length
event_win <- -1:1 # 3-day event window around the announcement
n_days <- est_days + length(event_win)
# True market model: R_it = alpha + beta * R_mt + e_it, with a +1.8% jump on day 0
alpha <- 0.0002; beta <- 1.1; sigma <- 0.012
r_m <- rnorm(n_days, mean = 0.0004, sd = 0.009)
shock <- rep(0, n_days); shock[est_days + which(event_win == 0)] <- 0.018
r_i <- alpha + beta * r_m + shock + rnorm(n_days, sd = sigma)
# Estimate the market model on the clean pre-event window only
est_idx <- seq_len(est_days)
mm <- lm(r_i[est_idx] ~ r_m[est_idx])
ab <- coef(mm)[1]; bt <- coef(mm)[2]
# Abnormal returns over the event window
ev_idx <- (est_days + 1):n_days
AR <- r_i[ev_idx] - (ab + bt * r_m[ev_idx])
CAR <- sum(AR)
# Test statistic: CAR scaled by the estimation-window residual SD
sd_ar <- sd(residuals(mm))
t_stat <- CAR / (sd_ar * sqrt(length(ev_idx)))
cat("Estimated alpha, beta:", round(ab, 5), round(bt, 3), "\n")
#> Estimated alpha, beta: 0.00197 1.069
cat("Daily abnormal returns:", paste(round(AR, 4), collapse = ", "), "\n")
#> Daily abnormal returns: 0.0118, 0.0052, 0.0206
cat("CAR over [-1,+1]: ", round(CAR, 4),
sprintf("(%.2f%%)", 100 * CAR), "\n")
#> CAR over [-1,+1]: 0.0376 (3.76%)
cat("t-statistic: ", round(t_stat, 2), "\n")
#> t-statistic: 1.94The recovered \(\mathrm{CAR}\) concentrates on the event day and is statistically distinguishable from zero, reproducing the planted announcement effect. An endorser-scandal study is the same machinery with the sign reversed and the event redefined as the disclosure date.
16.5.3 What the Evidence Shows
Applied work using this design finds that endorsement is, on average, value creating but heavily conditional. Announcements of celebrity-endorsement contracts are associated with positive abnormal returns and with gains in sales, and the effect is larger when the endorser is well matched and athletically or reputationally distinguished (Agrawal and Jaffe 2000; Elberse and Eliashberg 2003). Athlete endorsements in particular raise the sponsoring firm’s sales and stock returns, with the lift tracking the athlete’s on-field success and visibility (Elberse and Eliashberg 2003). The downside is equally real and is priced: negative information about an endorser—arrest, scandal, doping—transfers to the brand and to firm value, and the contamination is stronger where the celebrity–brand link is tighter, the mirror image of the match-up amplification (Chuluun, Prevost, and Upadhyay 2017). Endorsement, like co-branding in Chapter 11, is thus a reputation-coupling decision: the firm gains access to the celebrity’s reputational capital but assumes a share of its variance.
Table 16.2 frames the trade-off as a simple expected-value problem and foreshadows the quantitative reasoning a manager should bring to a contract.
| Component | Driver | Match-up role |
|---|---|---|
| Upside (successful campaign) | Match x reach x meaning fit | Raises upside (amplifies positive transfer) |
| Downside (endorser scandal) | Scandal probability x contamination severity | Raises severity (amplifies negative transfer) |
| Contract cost | Negotiated fee + activation spend | Higher for higher-meaning endorsers |
| Net expected value | E[upside] - E[downside] - cost | Match-up enters with opposite signs |
16.5.4 A Back-of-the-Envelope Valuation
The table’s logic is worth making numerical, because it disciplines the romance of a marquee signing. Treat the endorsement as a one-period lottery: with probability \(1-q\) the campaign succeeds and adds value \(V^{+}\); with probability \(q\) a scandal strikes and subtracts \(V^{-}\); the contract costs \(C\). The endorsement is ex-ante value creating only if
\[ (1-q)\,V^{+} \;-\; q\,V^{-} \;-\; C \;>\; 0. \tag{16.3}\]
Match-up enters Equation 16.3 with opposite signs on \(V^{+}\) and \(V^{-}\): a tighter celebrity–brand coupling raises the success payoff \(V^{+}\) but also raises the contamination loss \(V^{-}\), so the optimal match is interior, not maximal, once risk is priced. The snippet below evaluates the criterion and the breakeven scandal probability.
Code
V_up <- 40e6 # firm-value gain if the campaign succeeds
V_down <- 120e6 # firm-value loss if the endorser is disgraced
C <- 15e6 # contract + activation cost
q <- 0.05 # annual scandal probability
ev <- (1 - q) * V_up - q * V_down - C
q_breakeven <- (V_up - C) / (V_up + V_down)
cat("Expected value of the endorsement: $",
format(round(ev), big.mark = ","), "\n", sep = "")
#> Expected value of the endorsement: $1.7e+07
cat("Breakeven scandal probability: ",
sprintf("%.1f%%", 100 * q_breakeven), "\n")
#> Breakeven scandal probability: 15.6%The breakeven probability—the scandal rate at which the endorsement is exactly worth its cost—falls as the contamination loss \(V^{-}\) rises, which is why brands with the most to lose (those whose match is tightest and whose equity is largest) are the most sensitive to endorser conduct and the most aggressive about morality clauses. The model is deliberately spare, but it captures the governing tension: the very congruence the match-up hypothesis prescribes is also what makes a scandal expensive.
16.6 Pitfalls and Identification
The recurring threat in endorsement research is selection. Firms do not pair celebrities with brands at random; they hire matched, attractive, credible endorsers precisely for valuable brands and well-funded campaigns. In observational data the match variable \(M\) in Equation 16.1 is therefore correlated with unobserved brand quality and campaign budget, so a naive regression overstates the causal effect of congruence and the event-study \(\mathrm{CAR}\) in Equation 16.2 blends the endorsement’s effect with the information the announcement reveals about the firm’s prospects. Experiments solve this by assigning the pairing; event studies mitigate it by differencing against the market and scrubbing confounds, but neither fully removes the firm’s choice of when and whom to announce.
A second pitfall is construct conflation. Credibility, attractiveness, and meaning are distinct, and a study that measures only generic “endorser favorability” cannot adjudicate among the three models or test the match-up interaction, which requires a product-relative match measure. A third is dynamics: the sleeper effect (Wu et al. 2015) and the gradual decay of meaning mean that a single post-event snapshot can mislead; effects should be traced over a horizon. A fourth, specific to influencers, is disclosure endogeneity: sponsored posts are selectively disclosed and selectively produced, so the apparent effect of an influencer confounds the creator’s choice of which products to promote with the persuasion itself (Nistor and Selove 2024). Each pitfall is an instance of the general principle that runs through this book— state the assignment mechanism before interpreting the estimate.
16.7 Key Takeaways
- A celebrity endorser rents an accumulated reputation; the three classical models locate its persuasive force in credibility (expertise, trust), attractiveness (familiarity, likability, similarity), and—most completely— meaning transfer of specific cultural meanings from celebrity to brand to consumer (McCracken 1986, 1989).
- The match-up hypothesis is the testable, interaction-based (\(\beta_3 > 0\) in Equation 16.1) operationalization of meaning transfer; a significant main effect without a significant interaction does not test it.
- Endorser influence is moderated by processing route (Petty and Cacioppo 1986), cultural values such as power distance (Winterich, Gangwar, and Grewal 2018), and—for influencers—perceived authenticity and disclosure (Sokolova, Seenivasan, and Thomas 2020).
- Endorser risk is the symmetric, transferable, portfolio-correlated downside of meaning transfer; tighter match amplifies both the upside and the contamination, so the value-maximizing match is interior once risk is priced (Equation 16.3).
- Endorsement contracts and scandals are firm-level events whose value effect is estimated by event study (Equation 16.2); the evidence shows positive average returns to well-matched, high-reputation endorsements and significant losses to endorser scandals (Agrawal and Jaffe 2000; Elberse and Eliashberg 2003; Chuluun, Prevost, and Upadhyay 2017), subject always to the selection caveat that firms choose whom, and when, to announce.
Agrawal, Anup, and Jeffrey F. Jaffe. 2000. “The Post-Merger Performance Puzzle.” In, 7–41. Emerald (MCB UP ). https://doi.org/10.1016/s1479-361x(00)01002-4.
Belk, Russell W. 1988. “Possessions and the Extended Self.” Journal of Consumer Research 15 (2): 139. https://doi.org/10.1086/209154.
Chuluun, Tuugi, Andrew Prevost, and Arun Upadhyay. 2017. “Firm Network Structure and Innovation.” Journal of Corporate Finance 44: 193–214.
Elberse, Anita, and Jehoshua Eliashberg. 2003. “Demand and Supply Dynamics for Sequentially Released Products in International Markets: The Case of Motion Pictures.” Marketing Science 22 (3): 329–54. https://doi.org/10.1287/mksc.22.3.329.17740.
Leung, Fine F., Flora F. Gu, Yiwei Li, Jonathan Z. Zhang, and Robert W. Palmatier. 2022. “EXPRESS: Influencer Marketing Effectiveness.” Journal of Marketing, May, 002224292211028. https://doi.org/10.1177/00222429221102889.
McCracken, Grant. 1986. “Culture and Consumption: A Theoretical Account of the Structure and Movement of the Cultural Meaning of Consumer Goods.” Journal of Consumer Research 13 (1): 71. https://doi.org/10.1086/209048.
———. 1989. “Who Is the Celebrity Endorser? Cultural Foundations of the Endorsement Process.” Journal of Consumer Research 16 (3): 310. https://doi.org/10.1086/209217.
McWilliams, Abagail, and Donald Siegel. 1997. “Event Studies In Management Research: Theoretical And Empirical Issues.” Academy of Management Journal 40 (3): 626–57. https://doi.org/10.5465/257056.
Nistor, Cristina, and Matthew Selove. 2024. “Influencers: The Power of Comments.” Marketing Science.
Petty, Richard E., and John T. Cacioppo. 1986. “The Elaboration Likelihood Model of Persuasion.” In, 1–24. Springer New York. https://doi.org/10.1007/978-1-4612-4964-1_1.
Petty, Richard E., John T. Cacioppo, and David Schumann. 1983. “Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement.” Journal of Consumer Research 10 (2): 135. https://doi.org/10.1086/208954.
Sokolova, Tatiana, Satheesh Seenivasan, and Manoj Thomas. 2020. “The Left-Digit Bias: When and Why Are Consumers Penny Wise and Pound Foolish?” Journal of Marketing Research 57 (4): 771–88. https://doi.org/10.1177/0022243720932532.
Srinivasan, Raji, and Sundar Bharadwaj. 2004. “Event Studies in Marketing Strategy Research.” Assessing Marketing Strategy Performance 2004: 9–28.
Swaminathan, Vanitha, Karen L. Page, and Zeynep Gürhan-Canli. 2007. ““My” Brand or “Our” Brand: The Effects of Brand Relationship Dimensions and Self-Construal on Brand Evaluations.” Journal of Consumer Research 34 (2): 248–59. https://doi.org/10.1086/518539.
Tian, Zijun, Ryan Dew, and Raghuram Iyengar. 2024. “Mega or Micro? Influencer Selection Using Follower Elasticity.” Journal of Marketing Research 61 (3): 472–95.
Winterich, Karen Page, Manish Gangwar, and Rajdeep Grewal. 2018. “When Celebrities Count: Power Distance Beliefs and Celebrity Endorsements.” Journal of Marketing 82 (3): 70–86. https://doi.org/10.1509/jm.16.0169.
Wu, Qingsheng, Xueming Luo, Rebecca J Slotegraaf, and Jaakko Aspara. 2015. “Sleeping with Competitors: The Impact of NPD Phases on Stock Market Reactions to Horizontal Collaboration.” Journal of the Academy of Marketing Science 43: 490–511.
The 15-item credibility scale (attractiveness, trustworthiness, expertise) is the field’s most-used instrument; we describe its structure rather than reproduce it, since the items are reflective indicators of three latent factors and the construct logic is what matters here. ↩︎