flowchart LR
Q(("Endowment ξ<br/>(unobserved)"))
R["Entry order r"]
S["Market share s"]
D["Survival D<br/>(sample filter)"]
Q -->|selection| R
Q --> S
R -->|"causal effect Δ (target)"| S
R --> D
Q --> D
S --> D
style Q fill:#f5f5f5,stroke:#999,stroke-dasharray: 4 3
style D stroke-dasharray: 4 3
26 Market Entry
When a firm brings a product to a market, when it arrives is a strategic variable as consequential as what it sells. Two firms with identical products, prices, and budgets can earn very different long-run shares solely because one entered first and the other entered third. The popular intuition—“first-mover advantage”—captures part of this, but only part: pioneers sometimes build durable franchises, sometimes are overtaken within a few years, and sometimes fail outright while a fast follower harvests the category they created. This chapter develops the constructs and methods needed to say which outcome a firm should expect, and why.
The chapter treats market entry as the joint outcome of three interacting decisions: the order in which firms enter relative to one another, the timing of entry relative to the market’s own development, and the responses entry provokes from incumbents and prospective entrants. We begin by defining order-of-entry effects precisely and separating the genuine causal advantage of pioneering from the survivorship and selection artifacts that have historically inflated estimates of it. We then formalize the mechanisms that generate, and erode, pioneer advantage—preemption, switching costs, demand learning, and network effects—and the offsetting forces that make the incumbent’s curse a real phenomenon. From there we move to entry-timing models that trade off the cost of being early against the cost of being late, to the game theory of entry deterrence and accommodation, and finally to a reproducible empirical workflow for estimating order-of-entry effects from share data. By the end, a reader should be able to define the effect formally, identify it from observational data without falling into the classic traps, and reason about the strategic response it implies.
Throughout, we connect entry to constructs developed elsewhere in the book: the brand-share persistence that makes early entry valuable (the branding material in Chapter 11), the diffusion processes that govern when a market is worth entering, and the signaling logic that explains both pioneer reputations and incumbent reactions. Entry is where positioning, diffusion, and competitive strategy meet.
26.1 Constructs: Order, Timing, and the Pioneer
Three distinct constructs are routinely conflated and must be separated before any measurement is possible.
Order of entry is the ordinal rank of a firm’s entry into a category: the pioneer is rank 1, the early follower ranks 2 and 3, and late entrants the remainder. Order is a relative, competitor-referenced quantity. Entry timing is the absolute position of entry on the market’s development clock—how far along the diffusion curve, technology cycle, or category life cycle the firm arrives. Two firms can share an order rank yet differ wildly in timing if they enter markets at different stages of maturity. Pioneer (or first-mover) advantage is the causal share or profit premium attributable to being first, holding the marketing mix fixed.
The third construct is the subtle one, because the pioneer differs from later entrants in many ways besides order, and a naive comparison attributes all of those differences to order. We therefore define the effect counterfactually.
Definition: order-of-entry effect
Let firm \(i\) enter category \(c\) at order rank \(r_i \in \{1,2,\dots\}\) and realize long-run market share \(s_i\). The order-of-entry effect is the causal contrast \[ \Delta_r \;=\; \mathbb{E}\!\left[s_i \mid \mathrm{do}(r_i = r)\right] \;-\; \mathbb{E}\!\left[s_i \mid \mathrm{do}(r_i = r+1)\right], \] the expected change in share from entering one position earlier, where \(\mathrm{do}(\cdot)\) denotes an intervention that sets the entry rank while holding the rest of the marketing mix and the firm’s intrinsic quality fixed. A pioneer advantage is the special case \(\sum_{k} \Delta_k\) accumulated from the pioneer’s rank to a reference follower’s rank.
The \(\mathrm{do}(\cdot)\) operator is doing heavy lifting. It is what distinguishes the order-of-entry effect from the order-of-entry gradient one observes in data, in which pioneers also tend to be the firms with better products, deeper pockets, and luckier timing. The empirical literature’s central methodological contribution is to make this distinction operational, and its central empirical finding is that the gradient overstates the effect—often severely.
26.1.1 The Empirical Generalization and Its Demolition
The early empirical consensus was striking in its regularity. Across consumer and industrial categories, market share appeared to decline systematically with order of entry. A widely cited functional form fits share as a power of rank, \[ s_r \;=\; s_1 \, r^{-\theta}, \qquad \theta > 0, \tag{26.1}\] so that the second entrant earns a fraction \(2^{-\theta}\) of the pioneer’s share, the third \(3^{-\theta}\), and so on. Estimating Equation 26.1 on the Assessor pretest database, Urban et al. (1986) recovered a share penalty on the order of a few percent of share per unit of rank, with the pioneer enjoying a substantial premium that survived controls for the marketing mix. Equation Equation 26.1 is attractive because it linearizes under logs, \[ \log s_r \;=\; \log s_1 \;-\; \theta \log r, \] making \(\theta\) a simple regression slope—the order-of-entry elasticity of share. The same convenience is its danger, as we will see.
The consensus was then sharply qualified. 2 re-examined the pioneer record using historical data—reconstructing the full population of entrants in 50 product categories from contemporaneous sources rather than from surviving firms’ own accounts—and reached a deflationary conclusion.
Almost half of the market pioneers failed, and the mean market share of the pioneers was much lower than earlier studies suggested. … Many “pioneers” in previous studies were not first to market but rather early leaders. Being first confers no guarantee of success; early market leadership, not pioneering, is the more reliable correlate of enduring share.
— paraphrasing 2
The reconciliation is not that pioneer advantage is fictional, but that the measured advantage was inflated by two artifacts, which we now formalize because they recur in every order-of-entry study.
26.2 Why the Gradient Lies: Survivorship and Selection
Two biases pull the observed rank–share gradient away from the causal effect \(\Delta_r\). Both are instances of conditioning on a collider or on a post-treatment outcome, and both bias the pioneer premium upward.
Survivorship bias. Cross-sectional studies typically sample firms alive at the time of measurement. Failed pioneers leave the sample; failed late entrants also leave, but pioneering and failure are correlated, so the surviving pioneers are a favorably selected subset. Formally, let \(D_i = 1\) if firm \(i\) survives to the measurement date. The estimand the analyst can compute is \(\mathbb{E}[s_i \mid r_i, D_i = 1]\), whereas the target is the unconditional \(\mathbb{E}[s_i \mid \mathrm{do}(r_i)]\). If survival depends on both order and the latent quality that also drives share, then \(D_i\) is a collider on the path \(r_i \to D_i \leftarrow \text{quality} \to s_i\), and conditioning on survival opens a spurious association between order and share. The historical-population design of 2 closes this path by sampling entrants regardless of survival; the resulting failure-inclusive pioneer share is far lower than the survivor-only figure.
Selection on unobservables (the “born-first vs. became-first” problem). Firms that pioneer are not randomly assigned to do so. They tend to possess unobserved complementary assets—superior R&D, marketing capability, or managerial foresight—that both lead them to enter early and independently raise their share. Write the share equation with an order term and an unobserved firm endowment \(\xi_i\), and the entry-order equation as \[ r_i \;=\; \delta' \mathbf{z}_i + \eta\, \xi_i + \nu_i . \] If \(\eta \neq 0\)—able firms self-select into early ranks—then \(\mathrm{Cov}(r_i, \xi_i) \neq 0\), the ordinary-least-squares (OLS) estimator of \(\beta\) in Equation 26.2 is inconsistent, and its bias has the sign of \(-\eta\) (since lower \(r\) means earlier), inflating the apparent pioneer premium. This is the order-of-entry version of the endogeneity that pervades marketing-mix estimation: the regressor is chosen by the firm in light of information the analyst does not see.
The two biases are distinct—survivorship is selection on the outcome, endogeneity is selection on an input—but they compound, and both inflate the same number. Boulding and Christen (2003) sharpen the substantive conclusion by moving from share to profit: they document a “sustainable pioneering advantage” that is real but conditional, residing not in mechanical share but in the pioneer’s accumulated learning and cost position, and they show that the profit advantage can persist even where the share advantage decays. The lesson for the analyst is that the dependent variable matters: share advantages erode faster than profit advantages, and studies that measure only share understate the durable component of pioneering.
1 summarizes the identification problem as a causal graph; the estimation section returns to how each path is closed.
26.3 Mechanisms of Pioneer Advantage
A causal pioneer advantage, where it exists, must run through an identifiable mechanism. Four are well established, and naming them matters because each predicts a different rate of erosion and a different defensive strategy.
Demand-side preemption of preferences and positioning. The pioneer defines the category prototype. Later entrants are evaluated relative to it, and the pioneer occupies the most preferred position in the perceptual space before rivals arrive, forcing them into less-preferred niches. Because the pioneer also shapes which attributes consumers regard as relevant, it can tilt the evaluative criteria toward dimensions on which it excels. This mechanism is closely tied to brand-share persistence: early-formed brand preferences are remarkably sticky, persisting for decades and even traveling with consumers when they migrate across regions (Bart J. Bronnenberg, Dhar, and Dubé 2009; B. J. Bronnenberg, Dubé, and Gentzkow 2012). The persistence of early advantage in B. J. Bronnenberg, Dubé, and Gentzkow (2012) is, in effect, a measured lower bound on demand-side preemption.
Switching costs and habit formation. Once consumers adopt the pioneer, the cost of switching—relearning, contract lock-in, lost loyalty benefits, or simply habit—shields its installed base from later entrants. A later entrant must compensate consumers for the switching cost as well as match the pioneer’s utility, which is equivalent to a price handicap.
Technological and learning-curve leadership. The pioneer accumulates production experience first and rides down the learning curve ahead of rivals, opening a unit-cost gap that a later entrant cannot close at equal volume. This is the cost-side counterpart of preference preemption and is the mechanism Boulding and Christen (2003) emphasize for the profit advantage.
Network effects. When a product’s value to a user rises with the number of other users, an early lead in installed base is self-reinforcing: the leader’s larger network makes its product more valuable, attracting still more users in a positive feedback loop that can tip the market (Katz and Shapiro 1985). Network effects are the strongest source of durable first-mover advantage—but, as Section 26.4 shows, they are also the mechanism most often defeated, because a sufficiently superior late entrant can dislodge an entrenched network leader when quality differences swamp the network premium (Tellis, Yin, and Niraj 2009).
Intuition before formalism
The four mechanisms share one logic: pioneering pays when the pioneer can build a stock—of preferences, of locked-in customers, of cumulative experience, of network members—that later entrants cannot replicate by spending a flow. Where no such stock accumulates, order is just a date on a calendar and confers no advantage. The strategic question is therefore always “what stock am I building, and how fast does it depreciate?”
26.3.1 A stock-accumulation formalization
The shared logic above can be written as a single state equation. Let \(K_{it}\) be the pioneer-advantage stock (installed base, accumulated experience, or preference capital) of firm \(i\), accumulating from the firm’s marketing and sales flow \(a_{it}\) and depreciating at rate \(\phi\): \[ K_{it} \;=\; (1-\phi)\,K_{i,t-1} \;+\; a_{it}. \tag{26.3}\] A pioneer enters at \(t=0\) with rivals absent and accumulates \(K\) unopposed until follower entry at \(t=\tau\). The pioneer’s head start is the stock it banks during \([0,\tau)\), and the durability of its advantage is governed entirely by \(\phi\). When \(\phi \to 0\) (preferences and networks that never decay), the head start is permanent; when \(\phi\) is large (fashion-driven categories, fast technical obsolescence), even a long monopoly period buys only a transient lead. Equation Equation 26.3 is the same perpetual-inventory logic used for brand capital in Chapter 11, and the parallel is not coincidental: pioneer advantage is a brand- and customer-capital stock built under temporary monopoly. The erosion of pioneer advantage, to which we now turn, is the statement that \(\phi > 0\) for most categories.
26.4 Erosion of First-Mover Advantage
First-mover advantage is not self-sustaining; it decays, and sometimes reverses. Four forces drive the erosion, mirroring the four mechanisms that create it.
Free-riding by followers. Later entrants observe the pioneer’s costly investments—market education, R&D dead ends, channel development—and imitate the winners while skipping the losers. The pioneer pays to resolve uncertainty that followers consume for free, so the follower’s expected cost of matching the pioneer’s position is strictly lower than the pioneer’s was.
Resolution of technological and market uncertainty. Pioneers commit under uncertainty about which technology and which positioning the market will reward. Followers enter after that uncertainty resolves and can target the realized winning configuration directly. This is the central tension of entry timing formalized in Section 26.6.
Incumbent inertia and the late-entrant quality play. A pioneer that has optimized around the first-generation technology and its installed base may be slow to adopt a superior new generation—the incumbent’s curse (Section 26.5). Where this happens, a late entrant wins not by being early but by being better. Tellis, Yin, and Niraj (2009) show directly that in high-technology markets network effects do not guarantee that the early or larger network wins: product quality frequently dominates, and a higher-quality late entrant can overturn an entrenched network-advantaged incumbent. This is the cleanest available evidence that even the strongest erosion-resistant mechanism—network effects—is defeasible by quality, and it is why “first but worse” is a losing position.
Category and positioning shifts. The attribute space the pioneer preempted can become obsolete when consumer preferences move or a disruptive technology redefines the category. The pioneer’s preference capital then depreciates rapidly (a large \(\phi\) in Equation 26.3), and incumbency in the old category is worth little in the new one.
Table 26.1 organizes the create-and-erode logic into a single reference.
| Mechanism (creates advantage) | Erosion force (destroys it) | Persists when… |
|---|---|---|
| Preference/prototype preemption | Preference shifts; repositioning | Tastes are stable; category well-defined |
| Switching costs / habit | Followers subsidize switching | Lock-in is high; relationships deep |
| Learning-curve cost lead | Followers free-ride on learning | Experience is proprietary, not leaky |
| Network effects | Quality-superior late entrant tips the market | Quality parity holds; standards lock in |
26.5 The Incumbent’s Curse and Incumbent Response
Order-of-entry effects are not solely the entrant’s story; they are co-produced by how incumbents respond. The starting point is a puzzle: incumbents have the resources, customers, and market knowledge to dominate the next generation, yet radical innovations disproportionately come from entrants. Chandy and Tellis (2000) document this incumbent’s curse across the histories of many product categories: incumbents and large firms introduce radical innovations far less often than their resource advantages would predict. The explanation is not capability but willingness: an incumbent’s prior investments in the current technology and its fear of cannibalizing existing sales create an internal disincentive to self-disrupt. The remedy is organizational—a demonstrated willingness to cannibalize existing products is the strongest internal driver of radical innovation (Chandy and Tellis 1998)—and it is precisely the absence of this willingness that opens the door for late entrants to overturn pioneers (Section 26.4).
When incumbents do respond to entry, the response takes recognizable forms. We distinguish deterrence (acting before entry to prevent it) from accommodation (reacting after entry to limit its damage). The line between them is the firm’s ability to commit credibly, which the next section formalizes.
26.5.1 Entry deterrence and credible commitment
The economics of entry deterrence turns on commitment. An incumbent would like to threaten a price war if an entrant comes, but a bare threat is not credible: once entry has occurred, fighting may be more costly to the incumbent than accommodating, so a rational entrant ignores the threat. Dixit (1980) resolves this by showing that the incumbent can make the threat credible through irreversible investment—capacity, in his model—installed before entry. The sunk investment changes the incumbent’s post-entry payoffs so that fighting becomes its best response, which deters the entrant in the first place. The mechanism is general: commitment value comes from foreclosing one’s own future options.
Formally, consider a two-stage game. The incumbent first chooses capacity (or any irreversible investment) \(k\) at sunk cost; the entrant then chooses to enter or not, anticipating the post-entry duopoly outcome that \(k\) induces. Let \(\pi^E(k)\) be the entrant’s post-entry profit net of its entry cost \(F\). The entrant stays out iff \[ \pi^E(k) \;<\; F, \] and the incumbent’s problem is to choose the smallest \(k\) that satisfies this inequality—entry-deterring capacity—and compare the profit from deterrence to the profit from accommodating entry at lower \(k\). Whether deterrence or accommodation is optimal depends on the entry cost \(F\), the size of the market, and how steeply the entrant’s profit falls in the incumbent’s commitment. The key qualitative result is that investment has strategic value beyond its direct production value: it shifts the rival’s beliefs about the incumbent’s future behavior.
Figure 26.2 renders the deterrence game as a decision tree.
flowchart TD
I["Incumbent: invest k?"]
I -->|"high k (commit)"| E1["Entrant: enter?"]
I -->|"low k (no commit)"| E2["Entrant: enter?"]
E1 -->|out| O1["Incumbent monopoly<br/>(deterrence succeeds)"]
E1 -->|in| F1["Incumbent fights<br/>(credible: entrant loses)"]
E2 -->|out| O2["Monopoly, but unstable"]
E2 -->|in| A2["Incumbent accommodates<br/>(duopoly)"]
style O1 fill:#e8f5e9,stroke:#43a047
style A2 fill:#fff3e0,stroke:#fb8c00
26.5.2 Signaling and preannouncement as entry tools
Commitment can also be communicated rather than built. A preannouncement—a public declaration of a forthcoming product before it ships—can shape the entry calculus of rivals and the expectations of customers. The signal is double-edged. Aimed at customers, it can stall their adoption of a rival’s product (a “vaporware” effect); aimed at rivals, it can warn them off a contested position. But the signal is costly when the promise is not kept: Sorescu, Shankar, and Kushwaha (2007) show that preannouncements move shareholder value, and that unfulfilled promises destroy it—“don’t make promises you can’t keep.” Preannouncement is therefore a credible signal only to the extent that reneging is punished, which ties it back to the costly-signaling logic developed for quality in Chapter 11: a signal works when high types can bear its cost and low types cannot. Specialization plays the same role on the entry margin, where committing to a single category credibly signals quality precisely because it forgoes profit elsewhere (Kalra and Li 2008).
26.5.3 Incumbent response to entry: empirical evidence
Beyond stylized games, entry provokes measurable incumbent reactions in real markets, and the magnitude depends on how distinctive the entrant is. In grocery retailing, the entry of an organic specialist store depresses category performance at incumbent generalist stores and sharpens consumers’ price sensitivity; incumbents can blunt the damage by reducing the entrant’s relative distinctiveness—broadening their own organic assortment, narrowing the price–quality gap, and shoring up authenticity, with premium organic lines proving more defensible than frequently promoted ones (Maesen and Lamey 2022). The general principle is that an incumbent’s optimal response is to contest the dimension on which the entrant differentiates, neutralizing the entrant’s positioning rather than competing head-on on price.
Entry decisions also interact with the channel through which a brand reaches the market. Bei and Gielens (2022) show that the choice between selling to a platform at wholesale (first-party) and selling on the platform directly (third-party) has asymmetric share consequences: first-party operation tends to depress a brand’s share—more so for brands that cannot establish trust amid many rival and rogue sellers—whereas third-party operation tends to raise share, especially for non-leading premium brands with prior direct-to-consumer experience. The mode of entry, not merely its timing, is a lever on the share a brand can ultimately hold.
26.6 Entry Timing: The Cost of Early versus Late
Order is relative to rivals; timing is relative to the market. The timing decision trades two opposing costs. Entering too early incurs the cost of an immature market—undeveloped demand, missing complements, unresolved technical standards—and the burden of educating customers, a cost later entrants free-ride upon. Entering too late forfeits the accumulated stock \(K\) in Equation 26.3 and cedes the preferred positions. The optimum balances the two.
A compact way to see the trade-off is to write the net value of entering at calendar time \(t\) as the discounted stream of profit from then on, minus the penalty for entering before the market is ready: \[ V(t) \;=\; \underbrace{\int_{t}^{\infty} e^{-\rho (u-t)}\, \pi\big(m(u),\, K(t)\big)\, du}_{\text{value of the captured market}} \;-\; \underbrace{C\big(m(t)\big)}_{\text{immaturity / education cost}}, \tag{26.4}\] where \(m(u)\) is market maturity (the share of ultimate demand realized by time \(u\), e.g., the cumulative diffusion curve), \(K(t)\) is the head-start stock the firm can bank by entering at \(t\) (declining in \(t\) as rivals arrive), \(\rho\) is the discount rate, and \(C(\cdot)\) is the immaturity cost, decreasing in maturity. The firm enters at the \(t^\star\) that maximizes \(V(t)\). Early in the life cycle \(K(t)\) is large but \(C(m(t))\) is also large and \(m(u)\) is low; later, the immaturity cost falls but so does the bankable head start. The first-order condition \(V'(t^\star)=0\) balances the marginal decay of the head start against the marginal fall in the cost of immaturity—a clean statement of why neither the earliest nor the latest entrant is generally optimal.
The maturity path \(m(u)\) is not exogenous to the firm’s information: timing is fundamentally about when uncertainty resolves. Adoption of a new technology follows rank, stock, and order effects in which heterogeneous adopters cross a profitability threshold at different times (Karshenas and Stoneman 1993), and the same logic governs when a category “takes off”—the abrupt acceleration that turns a niche into a mass market (Golder and Tellis 1997, 2004). A firm timing entry to a takeoff must forecast that inflection, not merely the level of current demand. Entering just before takeoff captures the head start with the immaturity cost nearly paid down; entering long before it pays the full education cost with deep discounting of distant returns.
Pioneering is not the same as being early
The constructs pull apart here. A firm can be the pioneer (rank 1) yet enter late in absolute terms if the category itself is young relative to consumer readiness; conversely a firm can be a late entrant (rank 4) yet arrive early on the maturity curve if the whole category is nascent. The empirical record of failed pioneers in 2 is in large part a record of firms that pioneered and mistimed—first to a market that was not yet ready, exhausting resources on education that a better-timed follower harvested.
26.7 Estimating Order-of-Entry Effects
We now make the identification problem of Section 26.1 operational. The goal is to estimate the order-of-entry elasticity \(\theta\) in Equation 26.1 without the survivorship and selection biases that inflate it.
Estimator and specification. The workhorse specification regresses log-share on log-rank with controls, \[ \log s_{ic} \;=\; \alpha_c \;-\; \theta \log r_{ic} \;+\; \boldsymbol{\gamma}'\mathbf{x}_{ic} \;+\; \varepsilon_{ic}, \tag{26.5}\] where \(i\) indexes brands, \(c\) indexes categories, \(\alpha_c\) is a category fixed effect (absorbing category-level demand and the mechanical constraint that shares sum to one), \(\mathbf{x}_{ic}\) collects marketing-mix and quality controls, and \(\theta\) is the order-of-entry elasticity. Under the assumptions below, OLS on Equation 26.5 consistently estimates \(\theta\).
Identifying assumptions and what breaks them.
- No selection on unobservables: \(\mathbb{E}[\varepsilon_{ic} \mid r_{ic}, \mathbf{x}_{ic}, \alpha_c] = 0\). This fails whenever an unobserved endowment \(\xi_i\) drives both early entry and high share (Equation 26.2). Remedy: an instrument \(z_i\) for entry order—shifting the order of entry without directly affecting share, such as the firm’s pre-existing presence in an adjacent category or a regulatory/technological event that gated entry timing—estimated by two-stage least squares. A valid instrument satisfies relevance (\(\mathrm{Cov}(z_i, r_i)\neq 0\)) and exclusion (\(\mathrm{Cov}(z_i, \varepsilon_i)=0\)); the exclusion restriction is the hard part and must be argued, not assumed.
-
No survivorship filtering: the sample includes entrants regardless of survival. Conditioning on survival reintroduces the collider bias of
- Remedy: historical-population sampling (Golder and Tellis 1993), or an explicit selection model (e.g., a survival equation estimated jointly with Equation 26.5).
- Correct functional form: Equation 26.1 imposes a constant elasticity. If the true gradient is convex or has a pioneer-specific discontinuity (a “rank-1 premium” beyond the smooth curve), the constant-\(\theta\) fit misattributes the discontinuity. Remedy: add a pioneer dummy and test it separately from the slope.
A reproducible illustration. The following simulation builds a population of entrants with a true order-of-entry effect and a confounding endowment, then shows (i) that the naive log-share-on-log-rank regression on survivors only overstates the effect, and (ii) that estimating on the full population recovers it. The data-generating process encodes exactly the biases formalized above, so the gap between the two estimates is the bias, isolated.
Code
set.seed(20260620)
n_cat <- 400 # number of categories
k_per <- 5 # entrants per category
theta <- 0.30 # TRUE order-of-entry elasticity (log share per log rank)
eta <- 0.60 # strength of selection: able firms enter earlier
sim <- do.call(rbind, lapply(seq_len(n_cat), function(cat) {
# latent firm endowment xi: drives BOTH early entry and high share
xi <- rnorm(k_per)
# entry order: firms with higher endowment tend to enter earlier (lower rank)
rank <- rank(-eta * xi + rnorm(k_per), ties.method = "first")
# log latent share: true negative order effect + endowment + noise
log_share_latent <- -theta * log(rank) + 0.8 * xi + rnorm(k_per, sd = 0.3)
share <- exp(log_share_latent)
share <- share / sum(share) # shares sum to one within category
# survival: better endowment AND earlier rank survive more often (collider)
p_surv <- plogis(1.0 + 1.2 * xi - 0.5 * rank)
survive <- rbinom(k_per, 1, p_surv)
data.frame(cat = cat, rank = rank, share = share,
xi = xi, survive = survive)
}))
# (1) NAIVE estimate: survivors only, no control for endowment
naive <- lm(log(share) ~ log(rank) + factor(cat),
data = subset(sim, survive == 1))
theta_naive <- -coef(naive)[["log(rank)"]]
# (2) FULL-POPULATION estimate: all entrants, controlling for the endowment
full <- lm(log(share) ~ log(rank) + xi + factor(cat), data = sim)
theta_full <- -coef(full)[["log(rank)"]]
cat(sprintf("True theta : %.3f\n", theta))
#> True theta : 0.300
cat(sprintf("Naive (survivors only): %.3f <- inflated\n", theta_naive))
#> Naive (survivors only): 0.600 <- inflated
cat(sprintf("Full population + xi : %.3f <- recovers truth\n", theta_full))
#> Full population + xi : 0.286 <- recovers truthThe naive survivor-only slope overstates the true elasticity because it conflates the genuine order effect with the survivorship and endowment biases; the full-population specification that observes the endowment recovers the truth. In real data the endowment is unobserved, which is why the credible designs are the historical-population sample (Golder and Tellis 1993) and an order-shifting instrument—the two remedies in the assumption list above. The simulation makes the abstract bias concrete: the number to distrust is the raw rank–share gradient.
26.8 Synthesis and Strategic Implications
The threads of this chapter converge on a small number of decision-relevant propositions. First, order is endogenous and survivor-filtered, so the raw rank–share gradient is a biased estimate of the value of entering early; the defensible estimate comes from failure-inclusive data or a credible instrument (Golder and Tellis 1993; Boulding and Christen 2003). Second, pioneering pays only when it builds a slowly depreciating stock—of preferences, switching costs, learning, or network members (Equation 26.3)—and the strategic question is always the size of \(\phi\), the depreciation rate. Third, the strongest erosion-resistant mechanism, network effects, is still defeasible by quality, so “first but worse” is not a tenable position (Tellis, Yin, and Niraj 2009). Fourth, incumbency is a liability as well as an asset: the incumbent’s curse (Chandy and Tellis 2000) and the cannibalization disincentive (Chandy and Tellis 1998) mean that the firm best placed to win the next generation often will not, which is precisely the opening late entrants exploit. Fifth, incumbents respond by contesting the entrant’s point of differentiation (Maesen and Lamey 2022) and, where they can commit irreversibly, by deterring entry outright (Dixit 1980); commitments communicated rather than built—preannouncements —work only when reneging is punished (Sorescu, Shankar, and Kushwaha 2007). Finally, timing is distinct from order (Equation 26.4): the optimal entrant arrives neither earliest nor latest but when the marginal decay of its head start just balances the marginal fall in the cost of market immaturity, an inflection tied to category takeoff (Golder and Tellis 1997; Karshenas and Stoneman 1993).
26.9 Key Takeaways
-
Separate the three constructs. Order (rank vs. rivals), timing (position on the maturity curve), and pioneer advantage (the causal premium for being first) are distinct; conflating them produces the “failed pioneer” paradox of
- Distrust the gradient. Survivorship bias (a collider) and selection on unobservables (endogenous order) both inflate the observed pioneer premium; identify the effect with failure-inclusive data or an order-shifting instrument (Section 26.7).
- Advantage = a stock with a depreciation rate. Pioneer advantage persists in proportion to how slowly its underlying stock—preferences, switching costs, learning, network—decays (Equation 26.3, Table 26.1).
- Quality defeats networks. Even network-effect leads can be overturned by a superior late entrant (Tellis, Yin, and Niraj 2009); incumbency invites the incumbent’s curse (Chandy and Tellis 2000).
- Respond on the entrant’s axis of differentiation, deter with credible irreversible commitment where feasible (Dixit 1980; Maesen and Lamey 2022), and treat the timing decision as a maturity-curve optimization, not a race to be first.
26.10 Further Reading
The pioneer-advantage debate is best read as a sequence: Urban et al. (1986) for the share-rewards generalization, 2 for its historical demolition, and Boulding and Christen (2003) for the profit-based reconciliation. For incumbent behavior, Chandy and Tellis (2000) and Chandy and Tellis (1998) develop the incumbent’s curse and its organizational remedy; Tellis, Yin, and Niraj (2009) supplies the quality-versus-network-effects evidence. The strategic economics of deterrence begins with Dixit (1980). Diffusion and takeoff, which govern entry timing, are developed in Golder and Tellis (1997), Golder and Tellis (2004), and Karshenas and Stoneman (1993), and connect to the diffusion treatment elsewhere in the book. The persistence that makes early entry valuable is documented in Bart J. Bronnenberg, Dhar, and Dubé (2009) and B. J. Bronnenberg, Dubé, and Gentzkow (2012), and links directly to the branding material in Chapter 11.
Bei, Zhiling, and Katrijn Gielens. 2022. “EXPRESS: The One-Party Versus Third-Party Platform Conundrum: How Can Brands Thrive?” Journal of Marketing, July, 002224292211168. https://doi.org/10.1177/00222429221116803.
Boulding, William, and Markus Christen. 2003. “Sustainable Pioneering Advantage? Profit Implications of Market Entry Order.” Marketing Science 22 (3): 371–92. https://doi.org/10.1287/mksc.22.3.371.17736.
Bronnenberg, Bart J., Sanjay K. Dhar, and Jean-Pierre H. Dubé. 2009. “Brand History, Geography, and the Persistence of Brand Shares.” Journal of Political Economy 117 (1): 87–115. https://doi.org/10.1086/597301.
Bronnenberg, Bart J, Jean-Pierre H Dubé, and Matthew Gentzkow. 2012. “The Evolution of Brand Preferences: Evidence from Consumer Migration.” American Economic Review 102 (6): 2472–2508.
Chandy, Rajesh K., and Gerard J. Tellis. 1998. “Organizing for Radical Product Innovation: The Overlooked Role of Willingness to Cannibalize.” Journal of Marketing Research 35 (4): 474. https://doi.org/10.2307/3152166.
———. 2000. “The Incumbent’s Curse? Incumbency, Size, and Radical Product Innovation.” Journal of Marketing 64 (3): 1–17. https://doi.org/10.1509/jmkg.64.3.1.18033.
Dixit, Avinash. 1980. “The Role of Investment in Entry-Deterrence.” The Economic Journal 90 (357): 95. https://doi.org/10.2307/2231658.
Golder, Peter N., and Gerard J. Tellis. 1993. “Pioneer Advantage: Marketing Logic or Marketing Legend?” Journal of Marketing Research 30 (2): 158. https://doi.org/10.2307/3172825.
———. 1997. “Will It Ever Fly? Modeling the Takeoff of Really New Consumer Durables.” Marketing Science 16 (3): 256–70. https://doi.org/10.1287/mksc.16.3.256.
———. 2004. “Growing, Growing, Gone: Cascades, Diffusion, and Turning Points in the Product Life Cycle.” Marketing Science 23 (2): 207–18. https://doi.org/10.1287/mksc.1040.0057.
Kalra, Ajay, and Shibo Li. 2008. “Signaling Quality Through Specialization.” Marketing Science 27 (2): 168–84.
Karshenas, Massoud, and Paul L. Stoneman. 1993. “Rank, Stock, Order, and Epidemic Effects in the Diffusion of New Process Technologies: An Empirical Model.” The RAND Journal of Economics 24 (4): 503. https://doi.org/10.2307/2555742.
Katz, Michael L, and Carl Shapiro. 1985. “Network Externalities, Competition, and Compatibility.” The American Economic Review 75 (3): 424–40.
Maesen, Stijn, and Lien Lamey. 2022. “EXPRESS: The Impact of Organic Specialist Store Entry on Category Performance at Incumbent Stores.” Journal of Marketing, March, 002224292210909. https://doi.org/10.1177/00222429221090983.
Sorescu, Alina, Venkatesh Shankar, and Tarun Kushwaha. 2007. “New Product Preannouncements and Shareholder Value: Don’t Make Promises You Can’t Keep.” Journal of Marketing Research 44 (3): 468–89. https://doi.org/10.1509/jmkr.44.3.468.
Tellis, Gerard J., Eden Yin, and Rakesh Niraj. 2009. “Does Quality Win? Network Effects Versus Quality in High-Tech Markets.” Journal of Marketing Research 46 (2): 135–49. https://doi.org/10.1509/jmkr.46.2.135.
Urban, Glen L., Theresa Carter, Steven Gaskin, and Zofia Mucha. 1986. “Market Share Rewards to Pioneering Brands: An Empirical Analysis and Strategic Implications.” Management Science 32 (6): 645–59. https://doi.org/10.1287/mnsc.32.6.645.