In addition to (intertemporal) price discrimination, there's a parallel process of ramping up production. Especially with tech products, they can have bugs at the beginning, even with all the precautions during design/prototyping. So a massive launch at high production volume is more risky. (Lower production volume also implies higher unit price; see economies of scale.)
If you look how chipmakers tier their products, they almost always launch a flagship first, which usually high price and small market, then the tech "trickles down" to a more mass-market product (that usually has lower specs than the flagship). But as this mass-market product takes hold it also lowers the appeal of the flagship somewhat. So it is a more complicated process if you consider the company's entire product range than just a single (flagship) product.
For the first issue; I could locate some (fairly cited) "seminal" literature; but there's a lot of it out there--the topic has been massively researched:
- Stokey, Nancy L. 1979. "Intertemporal Price Discrimination." Quarterly Journal of Economics 94:355-71.
This paper covers both the basic case of heterogeneous consumers and the case when production unit cost falls over time, but it doesn't cover stuff like incomplete information or self-competition from alternative products.
As an aside, even temporary ("sales"/discount) low prices are related to the same issue of intertemporal price discrimination; see
- Sobel, J., 1984. "The timing of sales." The Review of Economic Studies 51, 353–368.
Another fairly cited theoretical paper is
- Besanko, D. & Winston, W. (1990), "Optimal Pricing Skimming by a Monopolist Facing Rational Consumers", Management Science 36(5), 555-567
It assumes that the producer doesn't know apriori the value that consumers attach to a product and that consumers anticipate future price drops, so there's an iterated game being played in which the producer discovers the prices that consumers are willing to pay. This model also leads to an optimal strategy of decreasing price over time, but with lesser profits for the producer than in more simplistic models.
Since I couldn't find an applied paper on actual devices (insofar), from an 2004/2006 applied paper on console video-games, which formulated a model inspired from actual strategies of producers, in which competition is apparently not much of a concern due to apparently weak substitutability of video game titles, and in which the unit production cost doesn't really change over time, but capturing the price premium from "hardcore" games is important. The model this paper uses is a more elaborate version of the Besanko & Winston model, but here is their empirical justification for applying such a model:
All video-games exhibit
consistent patterns of price cutting from their times of introduction. Our interviews with managers
in the industry revealed that much of the motivation for this price cutting arises from the desire to
sell to the segment of high valuation “hardcore gamers” initially, and to cut prices over time to sell to
the “mass market”. This closely parallels the price discrimination incentive. On the cost-side,
production of video-games is characterized by a constant marginal cost structure. Marginal costs
correspond to royalty fees paid by the game manufacturer to the hardware console manufacturer (i.e.
Sony), and also the costs of producing and packaging each CD-ROM title, both of which were
constant over the time-period of the data. Hence, falling marginal costs are unlikely to be an issue in
pricing over time. Further, competition from other games is also unlikely to be driving force behind
falling prices. Given the large number of games in the market (over 600 for the Sony Playstation
alone), and the fairly unique characteristics of each game, we find video-games to be weak substitutes for each other. The observed price data also reveal that the rates at which prices fall are
not explained by competitive conditions in the market, a feature corroborated by managers in the
industry. Hence, this industry forms an almost ideal setting to study the value of intertemporal
price discrimination policies in practice. Given the features of our empirical application, we work
with a monopolistic model of pricing that ignores competitive considerations.
[...] We learned that the typical rules-of-thumb used for pricing share
many similarities to our model. First, estimates are used to assess the evolution in the size of the
potential market. Our interviews also revealed that managers revise game-prices periodically,
cutting prices if sales are low, and keeping prices high if realized sales are high. We interpret this
heuristic as indicating that the total sales of the game is an important state variable for the firm’s
pricing decision. This is roughly consistent with the model since the theoretical state variables, the
segment sizes, are a function of cumulative sales of the game until that time period. The model
assumes that managers know the distribution of consumer types and can, therefore, translate the
observed sales of the game into segment sizes, which form the “payoff relevant” state variables for
the pricing decision. Managers are also aware that high willingness-to-pay “hardcore gamers”
sustain initial high prices, which have to be lowered once the game becomes “main-stream”. This adherers to the notion of price discrimination over time. As a reviewer pointed out, this heuristic
thumb-rule for price cutting is also consistent with managers cutting prices to reflect lower
valuations of consumers for older games. To the extent that such “novelty” effects are common across
segments, these are captured by our model via the game-specific shocks to utility [...].
A relevant question here is whether it is the retailer, rather than the manufacturer (as in our
model), that is initiating the observed price-cuts in the data. For instance, it could be that wholesale
prices from the manufacturer are constant, and the retailer is cutting prices over time due to reasons
unrelated to intertemporal price discrimination. For example, falling retail prices could arise from
retailers rapidly clearing inventory of low-selling games to free up shelf-space for new releases. Our
interviews with managers in the industry however, indicated that game manufacturers do
periodically initiate cuts to wholesales prices, which are mostly passed through to consumers by
retailers. Further, we found that the industry typically implements excess inventory return policies
within the manufacturer-retailer channel, whereby retailers can return unsold stocks of games back
to the manufacturer, making retail price-cutting to clear inventory a less compelling explanation.
Nevertheless, in the absence of retail-level inventory data, we are unable to rule out this explanation
Devices however aren't as unsubstituitable as video games. There exist fairly
sophisticated models (based on optimal stopping) of demand segmentation; e.g.
I found an analysis of the PC printer market from this perspective, which assumes that the prices fall exogenously (from the consumer perspective). But apparently the problem gets fiendishly complicated when such consumer models (i.e. with partially substitutable products along multiple dimensions) have to be combined with (iterated) supply-side price models; insofar I haven't found a paper that analyzes supply and pricing in the context of more complex demand models like this...
According to a 2005 review by Ryzin and Talluri (which summarizes the theoretical works I mentioned above in more technical terms, so it's a good read in that respect)
there was not much research on dynamic price models for multiple (competing) products; they only point out to one such (2004) paper on airfare pricing (by the same authors), in which the competing products are different classes (economy vs business). Also, airfares generally see a different intertemporal price discrimination pattern (relative to durable goods) with the price for a given airfare generally increasing over time due to business customers making last minute reservations/decisions. They also mentioned that multiple-product models are apparently very sensitive to the substitution function, so difficult to generalize. There was also no specific research mentioned on dynamic pricing oligopoly models, but an opinion piece was cited that such models might not even be too useful since the rational-behavior expectation might not hold for (a small number of) competitors; so monopoly models were the staple of dynamic pricing research up to that date. I haven't found a more recent review insofar.
The 1972 Coase conjecture (negative result), which Ubiquitous highlighted can indeed be consider seminal in the sense that it spurred the research on actual dynamic pricing models that do allow for the price discrimination that is often seen in practice; e.g. one book said
The seminal work on inter-temporal
sales is Coase (1972) which demonstrates that, given a durable product with an
infinite number of selling opportunities over time, a monopolist will eventually
decrease a product’s price to its marginal cost because consumers will anticipate
this decrease and will wait for discounts (the famous Coase conjecture). Numerous
papers that followed laid out conditions in which the Coase conjecture may not
hold (Stokey 1979, Besanko and Winston 1990, and DeGraba 1995).
The first two of these I've already mentioned above. The DeGraba paper is about the producer creating an artificial shortage in the first period. This strategy is also considered relevant to the consumer electronics and related (e.g. gaming) industries; the DeGraba paper exemplified it with some console videogames shortage at launch. In the (simplest) DeGraba model of buying frenzy, with a limited total number of potential customers (n):
the monopolist wants to sell to customers when they
are uninformed, because all uninformed customers have the same valuation for the good (equal to the common expected valuation). Thus, a single price could capture all of this willingness to pay from n - 1 customers. Once customers become informed they no
longer have homogeneous valuations, so the potential to capture all of this surplus with
a single price vanishes.
More recent works on buying frenzies exemplify them with Apple products (at launch). More generally the reason for the frenzy (for a rational consumer and in a dynamic [at least two-period] pricing model) is
buying frenzies occur when customers are sufficiently uncertain about their valuations of the product and when they discount the future sufficiently but not excessively. [...] such frenzies can have a significantly positive
effect on firm profits and partially recover the loss due to non-commitment to future prices. [...]
If customers are relatively informed about the product
and do not value the waiting option, the firm does
not have any incentives to ration demand. Buying
frenzies are more likely to occur when customers are
initially uncertain about their preferences for the
product and benefit from waiting to learn their preferences.
Such uncertainty is more likely with respect to
an innovative product that will match the needs of
some customers but not others which may explain
why frenzies are common for new electronic products
(e.g., the iPhone) than for similar, “me too” products
(e.g., Android phones) that are released later. It is
therefore reasonable to assume that customers have
more uncertainty about new products than about
knockoffs produced later.
There also some alternative behavioral-economics models for buying frenzies, which don't assume full rationality from the consumer. One of these, Papanastasiou et al. observes that
scarcity strategies appear to be employed even when prices are fixed
and availability in the long-term is ample
In their Bayesian-learning model (of consumers)
[induced] product scarcity may act as an effective substitute for dynamic pricing, allowing the firm to approximate dynamic-pricing outcomes while charging a fixed price
So it seems controversial that [producer induced] buying frenzies are (always) related to later price drops... which is why I left this issue for the end.